Forecasting doi: 10.3390/forecast8040075

Authors: László Vancsura Tibor Tatay Tivadar Zakár

Understanding which segments of the economy drive aggregate stock market movements is central to risk management. This study traces how the economic drivers of the Standard & Poor’s 500 Index (S&P 500) changed across two episodes: the 2020 COVID-19 shock and the 2025 technology-led period, using eleven sector indices and nine deep learning architectures. During COVID-19, forecasting power concentrated in Consumer Discretionary, Health Care, and Industrials before reorganizing sharply around Information Technology, consistent with a disruptive break. In 2025, Information Technology and market momentum dominated throughout, with no comparable reorganization, consistent with a gradual adjustment rather than a disruptive shift. This distinction, invisible from accuracy metrics alone (Gated Recurrent Unit: Mean Absolute Percentage Error = 3.41% and 2.16%), shows that information-concentration diagnostics can complement forecast-accuracy and risk-monitoring frameworks.

Forecasting doi: 10.3390/forecast8040074

Authors: Harrison E. Katz

Tourism-demand forecasting overwhelmingly targets aggregate volumes, leaving the question of where demand will come from largely unaddressed. This paper forecasts that question directly. We make three contributions. First, we apply Bayesian Dirichlet autoregressive moving average (BDARMA) models to guest origin composition in large-scale platform booking data, which to our knowledge is the first use of Bayesian compositional time-series methods on booking-origin shares across multiple global destination regions. Second, we introduce seasonal structure in the Dirichlet precision parameter, and we isolate its contribution through an ablation against an otherwise identical constant-precision specification: seasonal precision lowers mean absolute error in all four destination regions, by between 13% and 35%. Third, the forecast target is the monthly composition of bookings indexed by booking date rather than stay date, which makes it observable ahead of realized arrivals and therefore usable for decisions with long lead times. Using proprietary Airbnb reservation data spanning 2017–2025 across four destination regions, we document substantial pandemic-era shifts in booking composition with heterogeneous recovery patterns. In rolling-origin evaluation, BDARMA achieves the lowest forecast error for EMEA, the most compositionally diverse region, reducing mean absolute error by 27% relative to naïve forecasts (p<0.001). Performance elsewhere is mixed: simple benchmarks remain hard to beat, and exponential smoothing on isometric log-ratio-transformed data attains the lowest error averaged across the four regions. The EMEA pattern suggests that direct compositional modeling is most valuable where several origin markets hold material shares, although four destination regions are too few to establish this as a general rule. The methodology yields probabilistic forecasts of source market shares that can inform marketing allocation, concentration-risk monitoring, and forward-looking operational planning.

Forecasting doi: 10.3390/forecast8040073

Authors: Dler Hussein Kadir Diyar Muadh Khalil Azhin Muhammed Khudhur

This study develops a rigorous, leakage-free forecasting framework for monthly Robusta coffee prices using historical observations from January 1975 to December 2025. A comprehensive set of explanatory variables is constructed from lagged coffee prices, moving averages, logarithmic returns, rolling volatility, and exogenous variables such as the Oceanic Niño Index (ONI), the U.S. Dollar Index, and Brent crude oil prices. To ensure methodological fairness, all predictors are generated exclusively from information available at the forecast origin, and all competing models are evaluated under a unified expanding-window walk-forward validation framework. Seven forecasting models are compared: Naïve, Exponential Smoothing (ETS), ARIMA, ARIMAX, Extreme Gradient Boosting (XGBoost), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). Forecasting performance is evaluated using R2, RMSE, MAE, and MAPE, while Taylor diagrams and the Diebold–Mariano test are employed to assess model agreement and differences in predictive accuracy. The results show that XGBoost achieves the highest forecasting accuracy (R2 = 0.956, RMSE = 0.264), followed closely by the Naïve (R2 = 0.954, RMSE = 0.271) and ARIMA (R2 = 0.954, RMSE = 0.270) benchmarks, whereas ARIMAX and ETS provide comparable performance and the deep learning models (LSTM and GRU) produce substantially larger prediction errors. Feature importance analysis further indicates that the first lag of coffee price is the dominant predictor, accounting for approximately 94% of the predictive gain in XGBoost. Overall, the findings demonstrate that rigorous leakage-free validation is essential for reliable forecasting research and that, for monthly Robusta coffee prices, increased model complexity does not necessarily yield superior predictive performance.

Forecasting doi: 10.3390/forecast8040072

Authors: Elena Chianese Angelo Riccio

Air-quality forecasting models are often compared by architecture, although reported skill also depends on the pollutant, monitoring density, forecast horizon, predictor latency, validation design, and deployment objective. We conducted a systematic mapping review of 533 unique records published between 2000 and 15 June 2026; 409 met the forecasting eligibility criteria. The evidence was analysed in two layers: a metadata-derived map of the full corpus and a targeted full-text synthesis of representative studies. The non-exclusive metadata categories show that general machine learning or benchmark studies were most common (n=207), followed by recurrent deep learning (n=109), hybrid or decomposition methods (n=60), Transformer or attention models (n=51), tree ensembles (n=48), CNN/ConvLSTM models (n=43), classical statistical methods (n=26), graph neural networks (n=13), and physics-informed or CTM-coupled methods (n=7). These counts describe topical prevalence, not comparative effectiveness. The main contribution is a decision framework that links the forecasting setting to a defensible starting model, the evidence available for that model family, and the minimum validation needed to support temporal, spatial, or external generalisation. The synthesis favours transparent statistical and tabular baselines for short or sparse single-station records; spatial deep models only when network geometry and leave-site-out testing support them; and CTM-coupled postprocessing when operational physical fields are available at issue time. Diffusion and foundation models remain promising but unevenly validated for pollutant forecasting. A leakage-safe daily PM2.5 case study in Naples illustrates the practical consequence: model rankings change with the metric, and every fitted model underestimates the highest 5% of concentrations.

Forecasting doi: 10.3390/forecast8040071

Authors: Dohun Kim Seonghee Lee In-Hwan Lee

Forecasting research repeatedly finds that simple methods can match or beat complex ones out of sample. We test this in a safety-critical domain, near-real-time prediction of fire hazard state and structural damage, using a simulator-grounded benchmark: high-fidelity computational fluid dynamics (Fire Dynamics Simulator, FDS) provides reference data, an FDS-calibrated zone model (CFAST) generates a large corpus cheaply, and the temperature trajectories drive a finite-element model (OpenSees) and a HAZUS/Eurocode-informed damage rule. Under one protocol we compare simple, deep (PatchTST, TimesNet), and physics-informed (PINN, PIKAN) forecasters. Complex models do not dominate: a small corpus lets parsimonious models approach best accuracy, and physics helps mainly when data are scarce (crossover near twenty scenarios). We propose CD-PINN, which identifies a data-optimal reduced-order physics residual from the corpus by physics-guided regression over a candidate library and uses it as the physics constraint. This lifts a per-event physics model to the accuracy of corpus-trained forecasters while staying interpretable. On 26 laboratory-fire experiments, however, in-distribution rankings do not transfer: the large accuracy spread collapses to near-parity, so a leaderboard poorly predicts laboratory-fire accuracy. For downstream damage, we further show that the label definition, not the model class, sets the achievable ceiling.

Forecasting doi: 10.3390/forecast8040070

Authors: João Marreiros Ana de Jesus Mendes Marcela Castro Maria da Graça Costa Tiago Pinho

Accurate Estimated Time of Arrival (ETA) forecasting is essential for improving operational planning and decision-making in modern ports. While machine learning has significantly enhanced ETA prediction using Automatic Identification System data, the impact of data preprocessing on forecasting performance remains underexplored. This study investigates how AIS data preprocessing and machine learning model selection jointly affect ETA forecasting for short-sea shipping, using the Port of Sines as an empirical case study. A reproducible forecasting workflow was developed, integrating dataset construction, voyage selection, feature engineering, predictive modelling and performance evaluation. Three supervised machine learning algorithms, K-Nearest Neighbors, Random Forest Regression, and Multilayer Perceptron, were trained and compared under identical experimental conditions. The results show that forecasting performance depends not only on model selection but also on the quality of the modelling dataset. In particular, Random Forest Regression achieved the strongest and most consistent performance. It was found to be invariant to feature scaling, whereas scaling had a small negative effect on K-Nearest Neighbors and increased training variance for the Multilayer Perceptron. Although all three models achieved accurate ETA predictions, they exhibited different strengths regarding predictive performance, computational efficiency, and operational applicability. The proposed workflow contributes to the development of transparent and reproducible ETA forecasting methodologies and provides practical guidance for implementing AIS-based decision-support systems in short-sea port operations.

Forecasting doi: 10.3390/forecast8040069

Authors: Ogochukwu Ejike David Ndzi Muhammad Shakir

Rainfall prediction models often lose skill when transferred across regions, particularly in data-sparse settings where local recalibration is not feasible. This study investigates whether topographically analogous landscapes exhibit consistent patterns of model portability in next-day rainfall occurrence prediction across contrasting climates. Random Forest and Logistic Regression classifiers trained on multi-year daily atmospheric data were evaluated across three hydrogeomorphic classes, Alluvial/Valley, Delta/Marsh, and Coastal Plain, using paired temperate sites in the United States and tropical sites in Malaysia. Portability was quantified using Transferability, representing a model’s ability to export predictive skill, and Adaptability, representing a site’s receptivity to externally trained models. Performance, sensitivity, and stability metrics were synthesized within a Behaviour Grid to support systematic interpretation. Results reveal a robust terrain-driven hierarchy: Delta/Marsh models are the strongest exporters, Alluvial/Valley sites the most adaptable receivers, and Coastal Plain sites are stable generalists. This general hierarchy is preserved across climates and years when evaluated using Area Under the Receiver Operating Characteristics Curve, indicating strong terrain structuring of portability, although fixed-threshold performance metrics show sensitivity to temporal variability. These findings indicate that terrain class provides a useful basis for evaluating cross-regional model transfer, supporting more informed regionalisation and deployment decisions across heterogeneous environmental settings.

Forecasting doi: 10.3390/forecast8040068

Authors: Yu Chao Nur Fazidah Elias Yazrina Yahya Ruzzakiah Jenal Mo Fan

High classification performance in financial-risk early-warning research may reflect recovery of a constructed contemporaneous rating rule rather than prediction of an independent future outcome. This study distinguishes these two forms of evidence through a controlled synthetic-panel experiment situated in the context of higher education institutions (HEIs). Task 1 is a contemporaneous positive-control audit in which the four components defining a deterministic three-class rating are supplied to the classifiers. Task 2 uses predictors measured at year t−1 to classify the rating at year t through strict rolling-origin evaluation under prespecified Weak, Moderate, and Strong temporal-persistence conditions. In Task 1, the four fitted classifiers achieved held-out Macro-F1 values of 0.9968–1.0000, demonstrating near-complete recovery of the disclosed rating rule but providing no prospective forecasting evidence. In Task 2, the best mean annual Macro-F1/Macro-AUC increased from 0.480/0.696 under Weak persistence to 0.549/0.747 under Moderate persistence and 0.633/0.824 under Strong persistence. The Weak–Moderate–Strong ordering was observed for both primary metrics across all four fitted classifiers, while higher model complexity provided no consistent advantage within the disclosed synthetic mechanism. These findings confirm contemporaneous rule recoverability and sensitivity to deliberately embedded temporal persistence only within the controlled experiment. They do not establish predictive validity, transportability, or decision benefit in real HEIs. The study contributes a reproducible task-to-claim approach that aligns target construction, predictor–target overlap, information timing, rolling-origin evaluation, probability quality, and model performance with the inferences that the resulting evidence can legitimately support. Real-world validation would require source-indexed longitudinal HEI data, independently adjudicated post-origin outcomes, verified information-availability dates, external testing, and prospective decision evaluation.

Forecasting doi: 10.3390/forecast8040067

Authors: Cuicui Yu Wangyu Shen Erlong Zhao Shouyang Wang Pei Du

This paper proposes a novel variational mode decomposition (VMD)-enhanced ensemble framework for tourism demand forecasting, which achieves competitive predictive performance. The core innovation is a structured hybrid methodology that first employs VMD to decompose complex temporal signals, from which multiple interpretable features—encoding trend, volatility, and momentum characteristics—are extracted. These enriched features are then processed by an intelligently weighted ensemble of three diverse models: support vector regression (SVR), gradient boosting regression (GBR), and random forest (RF). An adaptive weighting mechanism dynamically optimizes the contribution of each base model, enhancing robustness and accuracy. Comprehensive experiments demonstrate the framework’s superior effectiveness. Compared to strong contemporary benchmarks such as CNN-LSTM, our approach achieves significant error reduction: an RMSE of 0.837 (16.29% improvement), an MAE of 0.584 (12.19% improvement), and a MAPE of 15.42% (13.38% improvement). These results validate that the integration of VMD-based feature engineering with an adaptive multi-model ensemble effectively captures complex temporal dynamics and mitigates individual model limitations.

Forecasting doi: 10.3390/forecast8040066

Authors: Juan R. Trapero Enrique Holgado de Frutos Francisco Ramos

High-performance computing based on general-purpose graphical processing units (GPUs) is a powerful tool for reducing computational time. In a context where big data is becoming increasingly relevant, GPUs may play a crucial role. This study analyzes the performance of GPUs by implementing probabilistic forecasts based on single exponential smoothing combined with simulated predictive distributions. In supply chain environments, companies must generate a large number of forecasts at the SKU level. Therefore, reducing computational time can provide a significant competitive advantage. Since forecasts are typically computed independently for each SKU, the problem is naturally parallelizable, making it well-suited for GPU computing. To the best of the authors’ knowledge, this is the first study to apply GPU computing to demand forecasting in a supply chain context. First, we show how to adapt typical probabilistic forecasting algorithms to a parallel computing framework. Then, real data from a manufacturing company are used to compare GPU and traditional CPU implementations.Theresults indicate that GPUs can deliver computational speedups ranging from 28 to 42 times relative to CPU-based implementations.

Forecasting doi: 10.3390/forecast8040065

Authors: Yan Egorov Dmitry Grigoriev Daniil Zharikov Anastasia Grigorieva

This study examines whether centralized-exchange flows improve one-day-ahead forecasts of SOL return direction. Daily on-chain analysis (OCA) predictors are constructed from more than 41 million transfers above 1 SOL involving 101 labeled centralized-exchange hot wallets. Seven feature sets are evaluated using CatBoost, XGBoost, Random Forest, LSTM, and BiLSTM across four expanding-origin holdouts, with Elastic Net Logistic Regression as a linear benchmark. Technical analysis (TA) achieves the strongest average classification and transaction-cost-adjusted trading performance. OCA performs weakly alone and provides no stable average improvement when added to Baseline or TA. Selected flow- and volume-related subblocks outperform unrestricted OCA on individual descriptive metrics, but none achieves positive average trading performance. Observable Solana exchange flows therefore provide no robust standalone or incremental forecasting advantage beyond conventional predictors.

Forecasting doi: 10.3390/forecast8040064

Authors: Tetiana Nechaieva Volodymyr Derii Artur Zaporozhets Viktor Denysov

The growing integration of variable renewable energy sources (VRES) challenges power system flexibility and may cause curtailment due to excess capacity, grid constraints, or operational and market factors. Power-to-Heat (PtH) technology can mitigate these issues by coupling electricity and district heating sectors, providing additional flexibility and supporting decarbonisation. This study develops a long-term generation capacity expansion model that integrates PtH and district heating system (DHS) operation to achieve balanced VRES penetration. The model includes DHS heat demand balances and links electricity and heat via thermal power plants, combined heat and power (CHP) plants, and PtH units. The methodology is applied to Ukraine’s Integrated Power System and district heating demand through 2040, employing typical daily load profiles discretised into six four-hour segments. Results demonstrate the feasibility of deploying PtH electric boilers during the non-heating season, when high RES and base load nuclear generation create surplus electricity. These boilers convert excess wind and solar power into thermal energy for district heating, displacing natural gas-fired technologies and simultaneously decarbonising electricity and heat supply.

Forecasting doi: 10.3390/forecast8040063

Authors: Adam Ashford Fahad Ayaz Muhammad Zeeshan Shakir Naeem Ramzan Michael Grebreslasie Serestina Viriri David Ndzi Natalie Dickinson Llinos Haf Spencer Mary Lynch Saloshni Naidoo

As extreme heat events increase in frequency, intensity, and duration due to climate change, forecasting these events has become vital for early warning systems, public health preparedness, and climate adaptation strategies, especially in parts of the world that are already subject to extreme heat, such as tropical regions. In recent years, machine learning (ML) has increasingly been applied to environmental and meteorological data to improve the prediction of heatwaves and extreme heat events. This scoping review examines global peer-reviewed literature on the application of ML techniques for extreme heat prediction using environmental variables. This includes heatwave prediction, environmental and meteorological predictors used in these models, and the geographical distribution of existing research. A total of 23 peer-reviewed studies meeting the inclusion criteria were included in the review, following the PRISMA-ScR guidelines. The findings indicate that artificial neural networks and random forest models were most frequently reported as high performing within individual studies. However, direct comparisons across studies are limited by heterogeneity in prediction targets, validation strategies, lead times, heatwave definitions, and performance metrics. Temperature-related variables, especially maximum temperature, were consistently identified as the most influential predictors across studies. Furthermore, the evidence base was heavily concentrated in Europe, Asia, and North America, with comparatively limited representation from low- and middle-income countries respective to population, despite these regions often experiencing disproportionate impacts of climate change and extreme heat exposure. By synthesising current evidence on ML-based heatwave prediction, associated environmental predictors, and geographical research trends, this review provides insights to support the development of more robust, context-aware, and globally representative heatwave forecasting frameworks.

Forecasting doi: 10.3390/forecast8040062

Authors: Farzaneh Esmaeilian Xinxuan Zhang Fatemeh Azizpourshoubi Marina Astitha Emmanouil Anagnostou

Reliable power systems are essential to modern life, as severe storms continue to disrupt grid stability and cause widespread outages. Predicting storm outages enables utilities and emergency managers to pre-stage resources and improve resilience. However, the several days of forecast lead time typically needed for preparedness significantly affect the accuracy of outage predictions. This study investigates the impact of forecast lead time on the error propagation of a Gradient Boosting Machine (GBM)-based outage prediction model (OPM) driven by Weather Research and Forecasting (WRF) model forecasts and analysis predictions. We evaluate three error-analysis scenarios: FFAP (forecast vs. analysis-based outage predictions), FFAO (forecast vs. actual outages), and LFAO (leave-one-storm-out forecast vs. actual outages). Model performance is compared using Mean Absolute Percentage Error (MAPE) and Centered Root-Mean-Square Error (CRMSE) across short (12 h–1 d), medium (2–3 d), and long (4–5 d) forecast lead-time categories, with the long category representing the upper end of the medium-range forecast window relevant to operational preparedness. The results show that forecast lead time substantially affects outage prediction accuracy, but the magnitude depends on the evaluation setup. In the controlled FFAP scenario, CRMSE increased by approximately 110% as lead time increased, from 259 to 543 outages, isolating the effect of weather forecast degradation. In the more operational LFAO scenario, CRMSE was already high at short lead times, increasing from 847 to 920 outages, indicating that model generalization error dominates once storms are unseen. Across scenarios, LFAO errors were 51% higher than FFAO errors at short lead times, highlighting the importance of testing outage models under unseen-event conditions. These results quantify how forecast degradation and model generalization jointly shape the reliability of outage prediction and provide practical guidance for lead-time-aware storm preparedness.

Forecasting doi: 10.3390/forecast8040061

Authors: Amgad Khamis Francesco Crespi David Sánchez

Accurate electricity price forecasting is essential for market participants seeking to optimise bidding and arbitrage strategies. This paper presents a week-ahead (168 h) hourly electricity price forecasting study for the Spanish day-ahead market. Nine competing models—two naïve baselines (a Seasonal Naïve and a Day-of-Week persistence), a Lasso-estimated auto-regressive (LEAR) statistical benchmark, and six machine- and deep-learning models (CatBoost, Random Forest, LSTM, GRU, CNN, and a hybrid CNN–LSTM)—are benchmarked; the two leading models, CNN–LSTM and CatBoost, are then compared under exogenous-feature configurations. The analysis is complemented by an ex-post Add-One-In and Leave-One-Out feature-importance analysis, a controlled comparison of weather-input scenarios, and a rolling battery-arbitrage backtest that translates forecast quality into economic value. Under an endogenous benchmark of weekly rolling origins across 2024 (with a rotating start weekday) and Diebold–Mariano testing, a recursive CatBoost and the hybrid CNN–LSTM are statistically indistinguishable and both significantly outperform a direct multi-horizon CatBoost; once an operational (forecasted) weather input is added, recursive CatBoost becomes significantly the most accurate while remaining simpler and more stable to train, a ranking confirmed on a fully out-of-sample 2025 year. Operational weather forecasts are found to be the best weather input, recovering about 84% of the perfect-foresight weather improvement over a no-weather baseline, with the advantage concentrated at longer lead times. Natural-gas-fired generation emerged as the dominant explanatory feature, consistent with the marginal-pricing mechanism governing the Spanish market. In a rolling battery-arbitrage backtest on the out-of-sample 2025 year, a deployable forecast-driven 4-h grid-scale unit (200 MW/800 MWh) captured about 89% of perfect-foresight value at a 168 h optimisation horizon and about 87% at 24 h; extending the horizon from 24 h to 168 h added about 2.4% of profit, an optimisation-horizon (look-ahead) effect bounded at +4.5% under perfect foresight.

Forecasting doi: 10.3390/forecast8040060

Authors: Sanelisiwe Amanda Nkomo Ehsan Ahmadi Reza Maihami

Short-term rental housing plays an important role in the housing market by increasing property utilization and generating income opportunities for property owners. This study investigates the key attributes associated with Airbnb occupancy rates using listing data from five major U.S. cities. Data mining and machine learning techniques, including Random Forest, XGBoost, Deep Neural Networks (DNN), and hierarchical cluster analysis, were applied to identify factors associated with occupancy rate variation. Random Forest achieved the best performance (R2 = 23.92%). Feature importance analysis identified price, location (city), listing capacity, amenities, and host response rate as the variables most strongly associated with occupancy rates. Cluster analysis supported these findings by identifying a dominant group of moderately priced listings with smaller accommodation capacity, more amenities, higher host responsiveness, and an average occupancy rate of 57%. These findings provide data-driven insights that may help property owners optimize listing performance and improve occupancy rates.

Forecasting doi: 10.3390/forecast8040059

Authors: Nader Naifar

This study examines whether daily supply-chain stress and geopolitical risk improve the forecasting of strategic commodity and clean energy market returns. Using daily data on aluminum, copper, nickel, and clean energy from 10 February 2015 to 27 February 2026, the analysis compares a baseline forecasting model based on conventional market controls with augmented specifications that incorporate supply-chain stress, geopolitical risk, and their joint effects. The empirical framework combines multiple machine-learning algorithms with SHAP-based explainability to evaluate both forecast performance and the relative importance of predictors. Formal Diebold-Mariano tests are also used to assess whether the forecasting gains from augmented specifications are statistically significant. A Model Confidence Set analysis is further used to identify statistically superior model groups across the full set of algorithm-specification combinations. The results show that disruption-related predictors contain asset-specific forecasting information, while the comparison across algorithms indicates that no single model uniformly dominates across all assets and loss functions. The forecasting gains from disruption-related predictors, however, are strongly asset-specific and statistically uneven. For aluminum returns, augmented specifications that include supply-chain stress and/or geopolitical risk significantly improve forecast accuracy relative to the baseline. For copper returns, the evidence is weaker and mainly associated with geopolitical risk. For nickel returns, the joint inclusion of supply-chain stress and geopolitical risk provides the greatest improvement. By contrast, clean energy returns remain more closely tied to conventional macro-financial conditions, with no statistically significant incremental gains from disruption-related variables. SHAP evidence further indicates that predictor importance is asset-specific rather than dominated by a single market factor across all assets. The findings highlight the importance of combining flexible forecasting methods with economically interpretable tools when evaluating disruption-sensitive commodity and clean energy markets.

Forecasting doi: 10.3390/forecast8040058

Authors: Yuantai Cui Hiroaki Fukunishi

In the highly volatile cryptocurrency market, trading decision support based on price prediction remains a challenging task. Although machine learning and deep learning techniques have been widely applied to cryptocurrency price prediction, many existing approaches rely on correlation-based black-box models, which limits interpretability and robustness. In this study, we employed a NOTEARS-Linear-based Prediction Model (NLBPM) that directly incorporated causal structures inferred through a causal discovery method as structural constraints within the prediction model. Unlike conventional approaches that focus primarily on minimizing prediction error, the NLBPM emphasized return maximization as its objective function, thereby prioritizing practical economic value. Using Bitcoin as a case study, we constructed a model to predict the direction of price movement four hours ahead and evaluated its performance using a rolling-window scheme with a one-month sliding window. Analysis of the inferred causal structures showed that the returns improved when trades were executed only during rolling-window trials in which specific directed edges to the target variable were detected. Based on this finding, we proposed a causal filter strategy that restricts trading to periods in which specific directed edges to the target variable are detected. In the data period analyzed in this study, the selected edge was the one from the opening price (Open) to the target variable. Backtesting experiments incorporating a transaction fee of 0.1% demonstrated that, while the benchmark LSTM model achieved a negative monthly average return of −3.20% and the NLBPM without filtering yielded −0.72%, the NLBPM with the Open filter attained a higher monthly average return of 10.35%. This study supports the usefulness of using inferred causal structure for cryptocurrency trading decision support.

Forecasting doi: 10.3390/forecast8040057

Authors: Israel Edem Agbehadji Ibidun Christiana Obagbuwa

South Africa faces significant challenges in monitoring air pollution from different provinces due to the sparse nature of the sensor network and heterogeneous pollutant sources. Notably, some provinces continue to record a limited amount of data on air pollution, thus making monitoring in those locations problematic. Fortunately, the capabilities of deep learning models to facilitate effective monitoring in data-scarce locations have been highlighted by researchers; however, these models within the context of transfer learning still lack transparency and uncertainty quantification. Using air pollutants and meteorological factors, this study proposes a transfer learning model for particulate matter (PM2.5) prediction in a data-scarce region. This transfer learning (TL) model leverages an adaptive Bi-directional Gated Recurrent Unit (adaBiGRU) with explainable artificial intelligence (xAI) and uncertainty quantification (UQ) to provide a novel uncertainty-aware adaptation transfer learning (UATL_adaBiGRU) model for a data-scarce location. Variant models based on the adaBiGRU technique, such as the temporal convolution network adaBiGRU (TCN-adaBiGRU) and domain-adversarial neural network adaBiGRU (DANNadaBiGRU), are presented as comparative models. The performance evaluation metrics are root mean squared, R2 score and mean squared error. The R2 score of pre-trained models in source domain is adaBiGRU (0.888), DANN_adaBiGRU (0.7788) and TCN_adaBiGRU (0.876). Furthermore, other comparative TL models include GRU (0.898), MLP (0.802) and adaptive LSTM (0.886). Afterwards, the pre-trained baseline model (adaBiGRU) was fine-tuned in the target domain dataset and the unpromising result contributed to the proposition of the UATL_adaBiGRU model for a data-scarce location, with R2 score of 0.9618. Uncertainty assessment metrics results were also presented for the proposed model. Ablation assessment demonstrates that each component of the UATL_adaBiGRU contributes to enhancing the predictive performance. Again, the Diebold–Mariano (DM) test statistic demonstrates a statistically significant difference between baseline model and UATL_adaBiGRU model. Finally, the local interpretable model-agnostic explanation highlights multi-scaled features as contributing towards the prediction of PM2.5 in the target domain. In view of this result, model fine-tuning is strongly recommended to enhance the robustness of the proposed uncertainty-aware adaption model in data-limited regions in South Africa.

Forecasting doi: 10.3390/forecast8040056

Authors: Betül Yılmaz Sucuoğlu Ömer Faruk Beyca Fuat Kosanoğlu

Intermittent sales patterns, prevalent in fast-fashion retail, pose a critical challenge for conventional forecasting methods. This study empirically compares one-stage and two-stage machine learning (ML) frameworks with classical benchmarks (Croston, SBA). The two-stage approach uses a Random Forest classifier for demand occurrence, followed by regression models (RF, GBM, XGBoost, LightGBM) for magnitude. Models are evaluated using weekly sales data from an Iraqi fashion retailer, incorporating rich exogenous features like product attributes, pricing, weather, and special events across 64 unique attribute-defined product group time series. Performance is assessed via a fixed 13-week holdout and rolling-origin cross-validation, with LSTM and Temporal Fusion Transformer (TFT) serving as deep learning benchmarks. Empirical findings show that machine learning configurations achieve superior WRMSSE accuracy, with two-stage models often outperforming one-stage counterparts, and both significantly surpassing classical and deep learning baselines. The Two-Stage XGBoost yielded the lowest WRMSSE, establishing the feature-engineered two-stage framework as the strongest overall for this intermittent retail setting. Furthermore, a detailed SHAP analysis elucidated the distinct feature contributions to demand occurrence versus demand magnitude, providing actionable insights for inventory management. This rigorous benchmarking analysis offers practical implications for inventory planning and demand management in volatile markets, highlighting the effectiveness of explicit demand occurrence modeling.

Forecasting doi: 10.3390/forecast8040055

Authors: Yassir El Bakkali Nissrine Krami Youssef Rochdi Achraf Boukaibat

The operational integration of renewable energy into contemporary power systems requires accurate and dependable wind power forecasting, particularly in multi-site settings with nonlinear temporal dynamics, inter-site dependence, and forecast uncertainty. Static site conditioning, conditional variable selection, dynamic graph learning, encoder–decoder temporal fusion, interpretable temporal attention, quantile regression, and post hoc split conformal calibration are all combined in this work to create DG-TFT-CQR, a global multi-site historical-power-based probabilistic forecasting framework. A representative eight-site subset of the AEMO 5 Minute Wind Power benchmark was used to evaluate the model under four different forecasting settings: H1, H3, H6, and H12. The proposed model demonstrated the most balanced probabilistic behavior and the strongest overall point-forecasting performance over these horizons among the compared baselines. The MAE/RMSE/R2 values for the point-forecasting results were 0.025490/0.043186/0.980096 at H1, 0.037241/0.062569/0.958221 at H3, 0.047917/0.079747/0.932133 at H6, and 0.062891/0.102751/0.887340 at H12. Additionally, the model preserved competitive interval sharpness while maintaining empirical coverage near the nominal 90% target. DG-TFT-CQR is the most robust balanced framework, with particularly evident advantages at H1 and H12, according to ablation, site-wise, daily case, statistical, and complexity analyses. In pairwise comparisons, H3 and H6 correspond to more mixed regimes. All things considered, the suggested approach offers a reliable and practically significant solution for multi-site wind power forecasting that takes uncertainty into account.

Forecasting doi: 10.3390/forecast8040054

Authors: Ntebogang Dinah Moroke

Emerging-market equity exchanges require regime forecasting systems that are continuous in time, robust to heavy-tailed distributions, and optimised against false alarms. No existing method addresses all three simultaneously, and no prior study has reported a crisis false-alarm rate on JSE equities. We propose S-NODE-ANF-RRC: a stochastic neural ODE within an Adaptive Neuro-Fuzzy Risk-Regime Clustering architecture, integrated by a Milstein scheme with Lyapunov-regularised dual-loss training. The system is evaluated as a one-step-ahead probabilistic forecaster (h=1 trading day) on 2696 daily observations across 17 JSE securities (March 2015–March 2026). Gaussian mixture clustering on raw features (kurtosis 54.8) inflates ARI by 1.3×; log-transformation corrects this artefact. Two operational profiles emerge: the N-ODE-ANF-RRC achieves the lowest cost (10,350 bp, 65.1% below GMM) and longest lead time (0.71 days); the S-NODE-ANF-RRC achieves the lowest false alarm rate among probabilistic architectures (FAR = 0.051), with a 42.0% cost reduction versus GMM (McNemar p=0.027, power 1−β=0.73; bootstrap CI [5250, 19,600] bp excludes zero). Ablation confirms drift, diffusion, and dual-loss as the minimum viable daily-frequency configuration.

Forecasting doi: 10.3390/forecast8040053

Authors: Kleber H. Santos Francisco Cribari-Neto

This paper proposes a generalized seasonal beta prime autoregressive moving average model with dynamic precision, denoted by BPSARMA, for modeling and forecasting positive-valued seasonal time series. The proposed framework extends the generalized BPARMA model by incorporating stochastic seasonal dynamics in the conditional mean through seasonal autoregressive and moving average components while allowing a flexible autoregressive structure for the conditional precision parameter, thereby accommodating time-varying uncertainty. The model also allows the inclusion of covariates and deterministic seasonal regressors. Parameter estimation is carried out by conditional maximum likelihood, and the main inferential and diagnostic tools are discussed. Monte Carlo simulations are conducted to examine the finite-sample behavior of the estimators and associated inference procedures. The practical usefulness of the proposed approach is illustrated through hydro-environmental time series applications, where its forecasting performance is evaluated using both in-sample and out-of-sample predictive measures. The empirical results indicate that the BPSARMA specification often provides competitive or superior forecasting accuracy relative to competing models, highlighting its usefulness for modeling and prediction in positive seasonal time series.

Forecasting doi: 10.3390/forecast8030052

Authors: Mark Sinclair Andrew J. Shepley Farshid Hajati

Price spikes are rare but economically significant events observed across electricity, financial, commodity, and cryptocurrency markets. Their abrupt magnitude, heavy-tailed distributions, and severe class imbalance make them difficult to forecast using conventional time-series methods. This systematic literature review, conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, synthesises recent deep learning approaches to forward-looking price-spike prediction and classification. Searches of Scopus, Web of Science, and IEEE Xplore identified studies published between 2020 and 2026. Following screening and full-text eligibility assessment of approximately 300 studies, only 20 met the inclusion criteria and were included in the final synthesis, comprising 19 peer-reviewed papers and one doctoral thesis. The review develops a structured taxonomy spanning spike definitions, task formulations, model architectures, input design, and evaluation practices. A central finding is that predictive performance is driven more by problem formulation, label construction, and evaluation design than by model architecture. While architectures have diversified to include recurrent networks, transformers, graph neural networks, and hybrid frameworks, improvements are often attributable to differences in how the prediction problem is defined rather than the models themselves. Key limitations stem from inconsistent spike definitions and insufficient treatment of class imbalance, leading to a misalignment between modelling objectives and evaluation practices, further exacerbated by the absence of standardised benchmarks. These issues hinder comparability and can lead to overstated model performance by masking poor detection of rare but economically critical spike events. The review therefore identifies clear directions for future research, including standardised spike labelling, adoption of rare-event-appropriate evaluation frameworks, and problem formulations that explicitly target extreme-event prediction.

Forecasting doi: 10.3390/forecast8030051

Authors: Malibongwe Cyprian Nyati Paul-Francois Muzindutsi Christian Tipoy

This study set out to investigate the interactions between business cycles, financial cycles and monetary policy in South Africa. Explicitly, the study aims to examine the role of financial factors in business cycle models and the possibility of a unified macroeconomic framework in South Africa. Further, the study assesses the effects of demand shocks, supply shocks, interest rate shocks, and financial shocks on macroeconomic fluctuations. The study applied an analytical approach integrating the Generalised Method of Moments and System Generalised Method of Moments with a Structural New Keynesian Dynamic Stochastic General Equilibrium framework. Accordingly, it was concluded that the financial cycle plays a significant role in business cycle models and is a main driver of macroeconomic fluctuations in South Africa. Further, a unified macroeconomic framework for monetary policy analysis that links the financial system to the real economy in South Africa possibly exists. This study contributes to the South African Reserve Bank’s efforts by deepening understanding of the interactions between the financial system and the real economy and their implications for monetary policy in South Africa. By comparing the standard Taylor rule with a finance-augmented Taylor rule in a DSGE framework, the study helps answer the question of whether financial stability should be adopted as a second objective of monetary policy.

Forecasting doi: 10.3390/forecast8030050

Authors: Chunxia Tian Roengchai Tansuchat Songsak Sriboonchitta

This study proposes a hybrid forecasting framework that integrates Kalman Filtering (KF), Markov Switching (MS), and nonlinear recurrent learning for stock-index prediction. The KF component smooths short-term price noise, the MS model identifies latent return–volatility regimes, and the LSTM/GRU components learn nonlinear temporal patterns from regime-conditioned information. The framework is evaluated using the CSI 300, S&P 500, and Nikkei 225 indices through forecasting-accuracy measures, Bootstrap Diebold–Mariano tests with Modified Bayes Factor evidence, out-of-sample trading simulations, and robustness checks. The empirical results show that regime conditioning is the primary source of forecasting and economic improvement. KF–MS–LSTM performs best for the CSI 300 and Standard MS performs strongest for the S&P 500, while KF–MS–LSTM and KF–MS–GRU are more competitive for the Nikkei 225. In contrast, models without regime information, including pure LSTM/GRU and the standalone Transformer, generally exhibit weaker forecasting and trading performance. The findings suggest that latent market-state information is more important than neural-network complexity alone for robust financial forecasting, while the incremental value of Kalman filtering and recurrent learning remains market dependent. Overall, the results support regime-aware forecasting as an interpretable and economically meaningful approach for stock-index prediction under heterogeneous market environments.

Forecasting doi: 10.3390/forecast8030049

Authors: Mehrshad Samadi Aydin Shishegaran Mina Torabi Zohreh Sheikh Khozani

The ski-jump spillway is an energy-dissipating structure that discharges extra water beyond the dam’s capacity. The scour process occurs below spillways due to the collision of the water jet with high energy. It is critical to acquire information on scour holes to improve the dam’s safety and related components. Machine learning (ML) techniques have successfully demonstrated their effectiveness for modeling scour in hydraulic engineering. The present research considers novel approaches of ML models for estimating the scour hole geometries below ski-jump bucket spillways. This study investigates the capability of two novel feature-engineering approaches, namely Stronger Variable Creator Machine (SVCM) and High Correlated Variables Creator Machine (HCVCM), along with Gene Expression Programming (GEP) and their hybrid forms (SVCM+GEP and HCVCM+GEP), which were employed to predict normalized scour depth, scour length, and scour width below ski-jump spillways. Statistical metrics, graphical analyses, the Rank Mean (RM) method, the cross-validation approach, and U95 index were used for the evaluation and reliability assessment of the proposed ML models. The results showed that hybrid ML models consistently outperformed individual algorithms. The results indicated that the SVCM+GEP method with RM=1.83 and 1.50 had the highest performance compared to other methods for the prediction of DsDw and LsDw, respectively. In addition, the HCVCM+GEP method with RM=1.33 was the best model for the prediction of WsDw. In comparison with the conventional regression-based equations and previously reported ML methods, the proposed hybrid approaches improved the prediction results. In addition, the cross-validation method confirmed the robustness and generalization capability of the suggested hybrid ML models. The superior performance of the hybrid models is attributed to their ability to capture complex nonlinear interactions among hydraulic and geometric variables. The developed SVCM/HCVCM+GEP models provide accurate approaches for predicting scour parameters in hydraulic structures.

Forecasting doi: 10.3390/forecast8030048

Authors: Salim Lahmiri Stelios Bekiros

Background: Lyapunov exponent has been used in many science and engineering problems to quantify chaos in systems and understand their nonlinear dynamics. In financial engineering and forecasting, evaluation of chaos in financial data helps determine whether the data are predictable and if profits can be generated. The purpose of this study is to examine presence of chaos in cryptocurrency markets. Methods: To examine chaos, Lyapunov exponent is computed from a set of 50 cryptocurrencies and statistical one-sided and two-sided Student-t tests are performed to check if on average the computed Lyapunov exponents are equal, less, or larger than zero. Results: The statistical results reveal strong evidence that prices, returns, and trading volume changes are all chaotic; hence, they show nonlinear and deterministic characteristics. Conclusions: Prices, returns, and trading volume changes in cryptocurrencies could be predicted in the short run; for instance, on a daily basis. In this regard, active traders and investors may implement predictive systems to generate daily profits.

Forecasting doi: 10.3390/forecast8030047

Authors: Luyanda Majenge Simiso Msomi Sakhile Mpungose

South Africa remains one of the world’s most coal-dependent electricity systems, with coal accounting for 81.57% of generation in 2023. Despite policy interventions to diversify the energy mix, structural change is slow to emerge. This study provides the first integrated, empirically calibrated forecast of South Africa’s coal-to-clean-energy transition using a unified modelling architecture that combines structural break analysis, Bayesian estimation, and an enhanced Monte Carlo simulation with dynamic volatility (10,000 stochastic pathways). The findings confirm a permanent structural break in 2011 that coincided with the implementation of REIPPPP, following which coal began a statistically significant and sustained decline of approximately 0.7–0.75% points per year. The simulation produced a full probability distribution for the transition year (2053) when coal share falls below 50%. This demonstrated that long-term uncertainty rises faster than linearly and that, under current conditions, deep decarbonisation milestones are unattainable before mid-century. Policy scenario experiments also demonstrated that accelerating the annual decline rate necessitates coordinated, synergistic policy portfolios rather than isolated interventions. These findings provide a transparent, uncertainty-explicit forecast of South Africa’s transition trajectory, as well as a decision-relevant evidence base for planning, regulation, and equitable transition implementation.

Forecasting doi: 10.3390/forecast8030046

Authors: R. Shankar A. Alroomi V. Bougioukos K. Nikolopoulos

This paper contributes to the stream of literature on extreme event modelling and forecasting by comparing various forecasting methods for predicting extreme movements in GDP and unemployment in the United States. The data were obtained from multiple open sources for the USA, including CNBC, the U.S. National Library of Medicine, the National Institutes of Health, the Centres for Disease Control and Prevention, the Bureau of Transportation Statistics site, Investing Com, the U.S. Bureau of Labour Statistics, Yahoo Finance, The Balance and Wikipedia. The research focuses on identifying the optimal forecasting method between Machine Learning and time-series forecasting algorithms, for predicting extreme values of GDP and unemployment, accounting for natural disasters and industrial and economic factors. The statistical and analytical insights derived from this study, if used judiciously, can inform policymaking and planning.

Forecasting doi: 10.3390/forecast8030045

Authors: Sabrina Haque Nuruzzaman Khan Delwar Akbar Susan Kinnear Azad Rahman

Export supply chains (ESCs) for perishable fruits, such as mangoes and avocados, are shaped by complex supply–demand dynamics and macroeconomic conditions. However, limited forecasting of these dynamics constrains strategic planning and investment in Australia’s horticultural sector. This study assesses the longitudinal growth and future potential of mango and avocado exports. To achieve this, the study identifies influential supply–demand dynamics and applies time-series forecasting to understand the export trends. Historical export–import data were analysed for mango and avocado from 1992 to 2024, including volume, value, per capita GDP (Australia and key importing nations), real exchange rate, and real interest rate. Holt’s exponential smoothing was used to forecast export trends, supported by unit root testing in RStudio 4.2.3 and model execution in SPSS version 30. ARIMA and ARIMAX models were applied to stationary variables to improve mango export forecasts. The results show that avocado exports follow a strong upward trajectory, while mango exports remain volatile due to logistical inefficiencies and informal trade disruptions. ARIMAX modelling confirmed that production and consumption volumes significantly enhance forecast accuracy. Macroeconomic trends, rising GDP, declining real interest rates, and stable real exchange rates further reinforce Australia’s competitive position in the destination markets. The long-run trends in export volume and value suggest that both the mango and avocado sectors hold potential for further export growth, although the higher volatility observed in the avocado series indicates that expansion should be approached cautiously. To sustain this growth, maintaining a balanced relationship between production capacity and export demand, particularly for commodities exhibiting higher volatility, will be essential for ensuring stable and efficient export performance over time.

Forecasting doi: 10.3390/forecast8030044

Authors: Rafael Magallanes-Quintanar Carlos Eric Galván-Tejada Jorge Isaac Galván-Tejada Santiago de Jesús Méndez-Gallegos Antonio García-Domínguez

Accurate long-horizon drought forecasting is essential for water resource management and early warning systems in semi-arid regions. This study evaluates five state-of-the-art Transformer architectures—Vanilla Transformer, Informer, Autoformer, Temporal Fusion Transformer (TFT), and PatchTST—for 24-month forecasting of the Standardized Precipitation Index (SPI-12) across four climatically homogeneous regions of Zacatecas, Mexico (Semi-arid, Highlands, Mountains, and Canyons). Models were trained on monthly precipitation data from 1965–2022 and evaluated on an independent test period (2023–2024) using MAE, RMSE, Pearson correlation, and the Diebold–Mariano test. The results show that PatchTST achieved the best overall performance in three of the four regions, significantly outperforming the other models in most cases. The Vanilla Transformer performed best in the less variable Highlands region. These findings demonstrate that the model’s suitability is strongly dependent on regional climatic characteristics. PatchTST’s patch-based approach proved particularly effective for capturing complex temporal dependencies in highly variable semi-arid environments. This study highlights the potential of Transformer architectures, especially PatchTST, to improve long-horizon SPI forecasting and strengthen operational drought monitoring systems in water-scarce regions.

Forecasting doi: 10.3390/forecast8030043

Authors: Montchai Pinitjitsamut

This study examines whether decomposition-based deep learning forecasts of daily changes in natural rubber prices can appear directionally accurate while failing to preserve the dispersion of the target series—a failure mode that conventional accuracy metrics cannot detect. Using daily RSS3 FOB price changes in the period 2018–2026, a VMD-Augmented BiLSTM forecasting design is employed as the empirical vehicle for testing this question. Forecasts are evaluated jointly through Pearson correlation, directional accuracy, class-conditional recall, and the Standard Deviation Ratio (StdR), with StdR serving as a diagnostic for variance collapse on differenced series. The deployed model appends all Variational Mode Decomposition (VMD) components directly to the economic feature matrix and feeds the augmented sequence into a bidirectional LSTM encoder with temporal attention; VMD is fitted using an expanding-window procedure to prevent information leakage. The design is compared to a conventional per-IMF decomposition–forecast pipeline, a Vanilla LSTM, ARIMA(2,0,2), and a dual-pathway BiLSTM–Transformer control. On a 175-observation deduplicated test set, the deployed model attains Pearson correlation of r=0.821±0.016, directional accuracy of 82.5%±1.8%, and StdR =1.091±0.060 across five random seeds. The Vanilla LSTM baseline attains directional accuracy of 82.29%±0.00—statistically indistinguishable from that of the deployed model—yet exhibits variance collapse (StdR =0.210±0.007), confirming that DA alone cannot distinguish predictive skill grounded in conditional dynamics from forecasts that merely reproduce the unconditional sign distribution. The principal contribution is methodological: A variance-sensitive evaluation protocol that distinguishes forecast skill grounded in conditional dynamics from directional but underdispersed predictions, demonstrated across three empirically distinct mechanisms by which variance collapse arises in this setting.

Forecasting doi: 10.3390/forecast8030042

Authors: Avinash N. Parde Kartik Koundal Utkarsh Bhautmage Michael Mau Fung Wong Christina Oikonomou Haris Haralambous

The Eastern Mediterranean, notably Cyprus, is a climate change hotspot facing severe heatwaves. Accurate numerical weather prediction of these extremes requires precise land–atmosphere modeling and initial and boundary conditions. This study assesses replacing the default USGS Land-Use and Land-Cover (LULC) dataset with the 10 m ESA WorldCover 2021 dataset in the Weather Research and Forecasting (WRF) model to simulate the 15–29 July 2023 Cyprus heatwave. The updated LULC increased urban representation six-fold. Statistical validations showed significant improvements in 2 m temperature, relative humidity, and 10 m wind speed predictions across 85% of observational sites. Dynamically, it restored urban thermal memory, effectively capturing the daytime Urban Cool Island effect and nocturnal heat release. Furthermore, radiosonde validations showed that the update corrected nocturnal Planetary Boundary Layer Height (PBLH) underestimations and dampened exaggerated daytime convective mixing. However, crucial limitations remain. High-frequency diagnostics indicated the model still suffers from damped thermal inertia, missing the abrupt temperature spikes and rapid nocturnal cooling typical of semi-arid microclimates. Additionally, the updated configuration failed to capture severe atmospheric stagnation during peak heatwave conditions, highlighting that deep-rooted kinetic errors persist within default boundary layer parameterizations despite static surface improvements.

Forecasting doi: 10.3390/forecast8030041

Authors: Marcel Figura Martin Bugaj Elvira Nica Gheorghe H. Popescu

The study develops and empirically evaluates a forecasting-orientated structural model in which future Bitcoin historical volatility is modelled as being associated with market sentiment and blockchain fundamentals through market uncertainty. Market Sentiment (MS) is specified as a behavioural construct, Blockchain Fundamentals (BF) as network conditions, and Market Uncertainty (MU) as an endogenous regime construct that consolidates signals shaping historical volatility at t+1. Using 262 weekly observations from January 2021 to January 2026, the analysis applies partial least squares structural equation modelling (PLS-SEM) with formative constructs and a forward-dated volatility target to preserve temporal ordering. Paths are evaluated with bootstrapping, effect sizes, and mediation analysis, while predictive performance is assessed using PLSpredict, the cross-validated predictive ability test (CVPAT), benchmark-based comparison, and Diebold-Mariano (DM) tests. MU emerges as the dominant predictor of Future Historical Volatility, denoted as HV(t+1) in the structural model (β = 0.864, p-value < 0.001; f2 = 2.036). The effect of BF is largely indirect, with 91.02% of the total effect transmitted via uncertainty, indicating indirect-only mediation. The model explains substantial variation in HV(t+1) (R2 = 0.791) and shows predictive relevance (Q2 predict = 0.287), while the benchmark-based results indicate mixed but competitive forecasting performance relative to persistence-based and econometric alternatives. These findings are consistent with a regime-based interpretation of Bitcoin volatility and highlight the explanatory and predictive relevance of an integrated behavioural-network-uncertainty architecture.

Forecasting doi: 10.3390/forecast8030040

Authors: Pedro Sobreiro Domingos Martinho Rui Martins Ricardo Vardasca

Predicting profitable entry signals in Bitcoin markets remains challenging due to price volatility, the absence of fundamental valuation frameworks, and methodological pitfalls that are common in the literature. In this study, we evaluate five machine learning classifiers using a 37-feature hierarchical multi-timeframe pipeline with price-level-agnostic normalization across four temporal resolutions (15-min, 4-h, daily, and 3-day), spanning January 2020 to November 2025. Binary training labels were generated via majority-vote aggregation across 54 stop-loss/take-profit combinations, producing 6951 balanced samples (48.5% positive class). Five algorithms—Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM—are compared using expanding-window TimeSeriesSplit validation (5 folds). Random Forest achieved the highest cross-validated ROC-AUC (0.6086), with all models showing modest but consistent discriminative ability (range 0.57–0.61). Feature importance analysis identifies 4-hour Bollinger Band position and RSI as dominant predictors, with all timeframes contributing meaningfully. A true out-of-sample holdout on 1136 independently generated 2025 samples confirms generalization, with Logistic Regression achieving 0.6087 ROC-AUC. A subtle multi-timeframe look-ahead bias in higher-timeframe data alignment is identified and corrected, which inflated performance by approximately 0.20 ROC-AUC points before correction. Event-driven backtesting on 2025 out-of-sample data yields a gross upper-bound return of +35.97% (185 trades, SL = 1%, TP = 2%, threshold = 0.7, Sharpe = 0.14) before transaction costs, after realistic round-trip fees, net returns are likely negligible. The central finding is that models with ROC-AUC ≈ 0.60 cannot reliably generate economically significant returns once transaction costs are accounted for. The methodology provides a reproducible framework for ML-based binary classification studies requiring transparent, bias-corrected validation across diverse market regimes.

Forecasting doi: 10.3390/forecast8030039

Authors: Jan Dvorsky Matus Senci Abdul Bashiru Jibril Zora Petrakova

AI/IoT initiatives are increasingly adopted in business, yet reported success varies substantially across firms. This study develops and evaluates a firm-level predictive framework for the reported AI/IoT success rate, measured on a bounded 0–100 scale. Using enterprise survey data from Slovakia and the Czech Republic (n = 1250), we compare a regularized linear baseline (Elastic Net) with nonlinear approaches (Decision Tree and Random Forest) under a consistent out-of-sample evaluation framework, and we examine the best-performing model using permutation importance and PDP/ICE tools. Random Forest achieves the strongest out-of-sample predictive performance and reduces absolute errors relative to Elastic Net for most test observations, although diagnostics also reveal a small tail of extreme errors. Across model families, ai_iot_advantage_share emerges as the most stable predictor of reported AI/IoT success. Nonlinear diagnostics indicate a threshold-like transition in predicted success around the mid-range of advantage attribution and a saturation pattern at higher values. Readiness and performance-related variables are associated with higher predicted success, whereas higher barrier levels are associated with lower predicted success. The results position value realization as the most informative predictive signal in the dataset and provide an interpretable basis for enterprise-level screening and managerial reflection rather than causal inference.

Forecasting doi: 10.3390/forecast8030038

Authors: Pablo Lara-Navarra Antonia Ferrer-Sapena Enrique A. Sánchez-Pérez

Singular Design Foresight (SDF) is proposed as a foundational methodological framework for advancing Design Foresight (DF) toward a more explicit, traceable, and evaluable scientific discipline. The framework formalizes DF as a structured cycle in which qualitative foresight inputs—such as signals, trends, and expert interpretations—are progressively transformed into analyzable representations that support decision closure under conditions of structural uncertainty. SDF combines an expert-defined conceptual universe with semantic projections to relate textual and contextual evidence to anticipatory constructs, enabling the generation of traceable indicators and structured configurations of viable futures. Within this architecture, the Stakeholder Viability Principle (SVP) functions as a filtering mechanism that delimits relevant futures according to continuity, agency, and axiological coherence, while Social Singularity captures context-specific critical transitions that shape when and why decision closure becomes necessary. The framework is organized in alignment with Design Science Research (DSR), adopting an evaluation logic centered on validity, utility, and attribution. Rather than presenting conclusive system-level validation, the article synthesizes summative evidence from previously published studies on semantic projections, singularity detection, and mixed expert–corpus foresight applications to support the plausibility, internal coherence, and operational feasibility of the proposed framework, while delimiting full integrated validation as a future research objective. SDF does not aim to provide deterministic prediction; instead, it enables auditable anticipatory representations and justified closure under uncertainty. In this sense, the framework is compatible with forecasting understood as the production of evaluable anticipations under explicit assumptions, while preserving the interpretive and situated character of strategic decision-making.

Forecasting doi: 10.3390/forecast8030037

Authors: Jung-Hyuk Kim Nam-Wook Cho

Retail demand forecasting for fast-moving consumer goods (FMCGs) presents significant challenges due to high product variety, demand intermittency, and uncertainty, which prevent any single model from capturing the diverse demand patterns. To address these challenges, this study proposes a hybrid clustering framework that integrates rule-based (Syntetos–Boylan Classification) and machine learning (ML) approaches, combining time-series embeddings with unsupervised learning to segment products by demand structure. Building on this framework, forecasting is conducted through a two-phase methodology: selecting optimal baseline algorithms per cluster (Phase 1), then enhancing them with embedding-based hybrid models (Phase 2). The effectiveness of this approach is demonstrated using a large-scale real-world dataset comprising over 3.8 million weekly sales records from 12,661 products across 691 stores. Results show that the proposed method improves forecasting accuracy by approximately 5–15% compared to conventional models. Furthermore, model performance varies with demand volatility, as different model–embedding combinations perform best under different conditions. Finally, the proposed diagnostic heuristic reduces experimental effort by 25–50%. Comparative analysis reveals that ML-based clustering outperforms rule-based methods under stable demand, whereas rule-based clustering is superior under high demand uncertainty, confirming that no single clustering paradigm is universally optimal. These findings demonstrate the practical value of adaptive hybrid frameworks for FMCGs demand forecasting.

Forecasting doi: 10.3390/forecast8030036

Authors: Thinawanga Hangwani Tshisikhawe Caston Sigauke Timotheous Brian Darikwa Saralees Nadarajah

Accurate wind speed forecasting is essential for the efficient integration of wind energy into power systems, as it directly influences generation scheduling, grid stability, and energy market operations. Forecast errors can lead to significant economic losses, including increased balancing costs, inefficient dispatch of backup generation, and penalties in electricity markets. However, wind behaviour is highly complex due to the influence of synoptic weather systems, terrain variability, and turbulence, which makes accurate prediction particularly challenging. This paper proposes a hybrid modelling framework that combines a linear regression mean model with Gaussian process (GP) residual modelling to improve forecast accuracy. Monitoring stations were grouped based on geographic coordinates and elevation, with cluster validation using the Hopkins statistic and silhouette analysis. The results show that for high-elevation inland stations (cluster 2), GP residual modelling improves forecast accuracy by up to 16.3%. In contrast, for low-elevation coastal stations (cluster 1), the GP approach does not yield improvements, indicating that its effectiveness depends strongly on the underlying wind regime.

Forecasting doi: 10.3390/forecast8030035

Authors: Veronika Labosova Lucia Duricova Katarina Kramarova Marek Durica

Financial distress prediction remains a central topic in corporate finance and risk management, with extensive research devoted to improving classification accuracy through increasingly sophisticated statistical and machine learning techniques. Nevertheless, the influence of data preparation on predictive performance has received comparatively less systematic attention. This study examines how an economically grounded data-preparation process affects the predictive performance of selected statistical and machine-learning models dedicated to predicting corporate financial distress. Using the chosen financial ratios, generally accepted indicators of corporate financial stability and economic performance, financial distress models are estimated on both raw, unprocessed input data and pre-processed data involving the exclusion of economically implausible accounting values, treatment of missing observations, and class balancing. In light of the above, the study adopts a structured methodological approach to assess the predictive performance of selected classification models, namely decision tree algorithms (CART, CHAID, and C5.0), artificial neural networks (ANNs), logistic regression (LR), and linear discriminant analysis (DA), using confusion-matrix–based evaluation and a comprehensive set of evaluation measures. The results suggest that the process of input data preparation is a critical factor, significantly improving the predictive performance of financial distress prediction models across most modelling techniques employed. The most pronounced gains are observed in decision tree models. ANNs also demonstrate marked improvement after input data preparation, whereas LR benefits more moderately, and linear DA remains limited despite preprocessing. The average gain in accuracy across all six modelling techniques, calculated as the difference between pre-processed and raw performance for each method and averaged across methods, was approximately 15.6 percentage points, with specificity improving by approximately 26.9 percentage points on average, amounting to roughly half the performance variation attributable to algorithm choice, which underscores that data preparation is a primary determinant of model reliability alongside algorithm selection. A step-level detailed analysis further shows that missing value imputation is the dominant driver of improvement for tree-based models, while class balancing contributes most for ANNs and logistic regression. The findings highlight that reliable financial distress prediction depends not only on technique selection but also on the consistency and economic plausibility of the input data, underscoring the central role of structured data preparation in developing robust early-warning models.

Forecasting doi: 10.3390/forecast8020034

Authors: Francisco Augusto Nuñez Perez Francisco Javier Aguilar Mosqueda Adrian Ramos Cuevas Jaqueline Muñoz Beltran Jose Cruz Nuñez Perez

Short-horizon equity forecasting remains challenging because daily prices are noisy, heavy-tailed, and subject to structural breaks and regime shifts. We develop a fully automated, reproducible, and leakage-controlled multi-model pipeline for daily forecasting with horizon-specific configuration selection. The task is formulated as predicting cumulative H-day log-returns from OHLCV-derived information and converting them to implied price forecasts. All model families share a homologated design: causal feature construction, a strictly chronological split with an explicit purging rule to prevent label-window overlap for multi-day targets, training-only robustification (winsorization and adaptive clipping), and a unified metric suite computed consistently in return and price spaces. The framework benchmarks transparent baselines (zero- and mean-return), gradient-boosted trees (XGBoost), and deep temporal models (LSTM and CNN/TCN). Lookback length L∈{60,180,500} is selected via an internal walk-forward procedure on the pre-evaluation block, and final performance is reported on an external hold-out segment (last 15% of instances). Experiments on daily data for MT, DELL, and the S&P 500 index (through 3 February 2026) show that all families achieve similarly strong price-level fit at H=1, largely driven by persistence in the price process, while separation across families becomes more visible at H=5. However, predictive performance in return space remains weak, with R2 close to zero or negative, and Diebold–Mariano tests do not provide consistent evidence of statistical superiority over naive benchmarks. Under an operational rule that minimizes hold-out RMSE on the price scale, selected models are asset- and horizon-dependent, supporting horizon-wise selection rather than a single global architecture. Overall, the primary contribution lies in the proposed leakage-controlled evaluation and benchmarking framework rather than in demonstrating consistent predictive gains in financial time series forecasting.

Forecasting doi: 10.3390/forecast8020033

Authors: Thabang Molefi Tshegofatso Botlhoko Tlhalitshi Volition Montshiwa

This study aimed to compare RNN algorithms and select the best-performing one between the GRU and LSTM for forecasting South African adjusted closing gold prices. The study used weekly secondary data sourced from Yahoo Finance and partitioned into three regimes, pre-COVID-19, COVID-19, and post-COVID-19, as well as the overall sample. The results indicated that the GRU algorithm consistently outperformed the LSTM algorithm across all evaluation periods based on the selected metrics, except during the COVID-19 period, where LSTM exhibited slightly better performance. Consequently, the GRU algorithm was identified as the best-performing algorithm for the South African adjusted closing gold price series. The relative effectiveness of GRU and LSTM algorithms in financial time series forecasting was clarified by the results. By integrating GRU-based forecasts into development finance frameworks, stakeholders can strengthen resilience against global shocks, improve financial planning, and foster more stable pathways for economic development. The authors recommended that future studies explore the performance of the GRU and LSTM with other advanced algorithms like Transformer architectures, hybrid algorithms, or traditional statistical methods to further enhance the forecasting robustness.

Forecasting doi: 10.3390/forecast8020032

Authors: Monica Borunda Luis Conde-López Gerardo Ruiz-Chavarría Guadalupe Lopez Lopez Victor M. Alvarado Edgardo de Jesús Carrera Avendaño

Short-term load forecasting is essential for the reliable, secure, efficient, and economic operation of modern power systems and electricity markets. Among many forecasting strategies, the similar day (SD) approach for short-term load forecasting was among the earliest used to assess power demand and remains one of the most intuitive and widely adopted techniques worldwide. However, over time, increasing system complexity, richer datasets, and advances in computational intelligence have led to the evolution of SD methodologies beyond heuristic-based rule formulations. This work presents a study of the relevant literature on short-term load forecasting using SD methods reported between 2000 and 2025. This study analyzes how similarity is defined, how forecasts are generated, and how both stages interact within the complete forecasting process in the reviewed literature. Based on these criteria, a unified taxonomy is proposed to classify SD methods into conventional, intelligent, and hybrid formulations. This study provides insight into the methodologies, their performance, and the systems in which they have been tested. The results show that SD-based approaches remain competitive for short-term forecasting and that incorporating artificial intelligence techniques can further enhance their accuracy.

Forecasting doi: 10.3390/forecast8020031

Authors: Antonio Naimoli Giuseppe Storti

This paper examines whether climate, geopolitical and economic policy uncertainty indices improve Value-at-Risk (VaR) and Expected Shortfall (ES) forecasts for green and brown stocks. We extend the Realized-ES-CAViaR framework by incorporating physical and transition climate risk, geopolitical risk and economic policy uncertainty indices alongside a high-low range volatility estimator. Using daily data for the iShares Global Clean Energy ETF (ICLN) and the iShares Global Energy ETF (IXC) over the period January 2012–December 2024, we evaluate alternative model specifications at the 1% and 2.5% risk levels through backtesting procedures, strictly consistent scoring rules and the Model Confidence Set methodology. Results reveal a pronounced asymmetry in the predictive content of risk indices across asset classes and quantile levels. Transition climate risk dominates tail-risk forecasting at the 1% level for both asset classes, while geopolitical risk and economic policy uncertainty emerge as the leading factors at the 2.5% level for green and brown stocks, respectively. These findings highlight the heterogeneous channels through which uncertainty shocks propagate into financial tail-risk, with direct implications for risk management and regulatory oversight during the low-carbon transition.

Forecasting doi: 10.3390/forecast8020030

Authors: Achilleas Tampouris Chaido Dritsaki

This paper presents a hybrid econometric and machine-learning framework for forecasting GDP that bridges long-run structure with short-run regime dynamics. Using annual World Bank data spanning 1960 to 2024, the framework combines three complementary components: an ARIMA baseline that captures persistence, a three-state Hidden Markov Model (HMM) that provides probabilistic regime identification, and an LSTM-based extension that learns nonlinear patterns associated with regime transitions. Detailed out-of-sample forecasting evidence is reported for five representative countries (the United States, China, Germany, India, and Greece), chosen to illustrate performance across different volatility profiles and economic environments. Across these case studies, the integrated HMM–LSTM approach often delivers lower forecast errors than the benchmark alternatives, although the magnitude of the gains is not uniform across countries. Beyond point forecasting performance, the regime layer yields an interpretable probabilistic representation of business cycle conditions that can support real-time monitoring and early-warning assessment. By combining transparency with adaptability, the proposed framework contributes to the forecasting literature and provides a practical decision-support tool under heightened macroeconomic uncertainty.

Forecasting doi: 10.3390/forecast8020029

Authors: Suzana Lović Obradović Dunja Demirović Bajrami Marko Filipović

Non-communicable diseases (NCDs) are the leading causes of mortality in Serbia, with cardiovascular diseases (CVDs) accounting for a substantial share of premature mortality. In alignment with Sustainable Development Goal (SDG) Target 3.4, which aims to reduce premature mortality from NCD by one-third by 2030 relative to 2015, this study forecasts changes in CVD mortality counts at the municipal level in Serbia. Time-series data for the period 2005–2022 were analyzed within a spatio-temporal forecasting framework implemented in the Space Time Pattern Mining toolbox in ArcGIS Pro (Version 3.1). Three established forecasting models (Curve Fit Forecast, Exponential Smoothing, and Forest-based) were applied, and the most accurate model for each municipality was selected using location-specific municipality-level validation. The results reveal pronounced spatial variation: approximately half of the municipalities (51.2%) are forecasted to experience a decline in CVD mortality counts by 2030, while others are expected to show increases or no statistically significant change. Forecasted differences range from a 15.1% decrease to a 13.9% increase across municipalities, indicating heterogeneous spatial trajectories and suggesting that achieving SDG Target 3.4 may remain challenging without targeted interventions across municipalities where mortality reductions are not forecasted. Although the study does not introduce new forecasting methods, it provides a novel spatially disaggregated application of multi-model forecasting to support municipality-level monitoring of SDG 3.4. The results underscore the need for geographically differentiated public health policies and demonstrate the value of spatial forecasting approaches for supporting equitable and targeted health planning.

Forecasting doi: 10.3390/forecast8020028

Authors: Nikolaos Kanellos Dimitrios Katsianis Dimitris Varoutas

Medium-term 5G/NB-IoT planning is made difficult by simultaneous uncertainty in device adoption and per-device traffic behavior because deterministic point forecasts do not quantify overload risk or support reliability-based capacity decisions. A diffusion-to-dimensioning workflow is proposed in which S-curve adoption modeling, bounded usage priors, scenario stress testing, and Monte Carlo uncertainty propagation are combined to generate predictive demand distributions, exceedance curves, and quantile-based capacity rules. The framework is applied to a Great Britain case study for 2025–2029 using smart meter deployment data and an M2M-based proxy for asset-tracking adoption. Analysis shows that planning-year upper-tail outcomes are driven primarily by asset-tracking usage uncertainty rather than by proxy scale alone. A ±30% perturbation of the AT adoption anchor changes Q0.95 by approximately ±29.8%, whereas stressed AT usage increases Q0.95 by 74.4%. Plausible positive dependence among key AT operational inputs further raises Q0.95 by 18.3–22.5%. Limited hold-out evaluation provides strong out-of-sample support for the smart meter adoption stage and plausibility-only support for the shorter AT proxy. The framework is intended for medium-term, data-lean planning settings and is designed to support transparent risk-based capacity decisions rather than deterministic point sizing.

Forecasting doi: 10.3390/forecast8020027

Authors: Ali Louati Hassen Louati

Timely labor market monitoring is essential for policy design and operational planning, yet annual reports can mask turning points and subgroup heterogeneity. This paper develops a reproducible monitoring and prediction framework using administrative statistics from the General Organization for Social Insurance (GOSI) in the Saudi Open Data Portal. We document descriptive patterns in formal participation and insurable wages, including age-group dispersion, stable correlation structure, and explicit handling of an anomalous wage release and limited missing wage entries. We then formulate from non-salary administrative descriptors. Under leakage control, Random Forest models achieve accuracy around 0.71 across releases. Most errors are concentrated between adjacent wage bands, which is consistent with threshold discretization of a continuous wage distribution. To support operational deployment, we add out-of-time validation across releases and probabilistic assessment, showing that predictive skill transfers across updates and that calibration improves the reliability of probability scores for monitoring thresholds. Overall, the results indicate that administrative releases contain persistent actionable signals for wage segmentation without salary-derived inputs, supporting forecasting-oriented surveillance and early-warning dashboards.

Forecasting doi: 10.3390/forecast8020026

Authors: Peio Garcia-Pinilla Aranzazu Jurio Maria Figols Daniel Paternain

Indoor CO2 forecasting supports proactive ventilation control that balances air quality with energy efficiency. While Machine Learning (ML) models have shown strong performance in controlled settings such as schools, their generalization across indoor spaces with diverse occupancy dynamics remains poorly characterized. We present a systematic benchmark of 11 forecasting models spanning simple baselines, statistical methods, classical ML, deep learning, ensembles, and foundation models using 18 weeks of IoT sensor data spanning six real-world use cases: conference rooms, dining halls, hospitals, food markets, offices and student residences. Performance depends strongly on the prediction horizon and on the regularity of occupancy-driven CO2 patterns. Simple baselines tend to perform best at short horizons (10 min ahead), while ensembles and fine-tuned foundation models provide more robust accuracy at longer horizons (4 h ahead). Remarkably, zero-shot foundation models demonstrate the ability to outperform trained classical models in data-scarce scenarios, challenging the traditional paradigm of localized training. These findings indicate that optimal forecasting strategies are context-dependent and challenge the assumption of universal model superiority.

Forecasting doi: 10.3390/forecast8020025

Authors: Roshan Sewa Bibas Pokhrel Bikash Subedi Roshan Raj Karki Bishal Poudel Ajay Kalra

Rapid urbanization has transformed the face of Texas by converting agricultural and natural lands into expanding built-up areas. This study analyzes and simulates land-use and land-cover (LULC) changes in Kaufman County, Texas, one of the fastest-growing counties in the United States, using a hybrid Cellular Automata–Artificial Neural Network (CA–ANN) model within the Quantum Geographic Information System (QGIS) Modules for Land-Use Change Evaluation (MOLUSCE) framework. Multitemporal NLCD datasets (2001, 2011, and 2021) and six spatial drivers: Elevation, Slope, Aspect, Distance from Roads and Rivers, and Built-up Density were used in the modeling framework. Transition relationships were calibrated using the 2001–2011 LULC data, and the model was validated by simulating the 2021 LULC map from the 2011 baseline. The calibrated model was then used to simulate future LULC scenarios for 2031, 2041, and 2051. Model validation yielded an overall Kappa value of 0.84 and a correctness of 90.9%, indicating high similarity between the observed and simulated maps. The results indicate simulated urban expansion, with built-up areas increasing by nearly 30% by 2051 at the expense of cropland and open areas, with forest and water bodies slightly increasing, and wetlands remaining stagnant. The CA–ANN model effectively captured the nonlinear, spatially dependent land-transition patterns using open-source tools. These findings provided useful information for sustainable land-use planning and environmental management, with the potential to incorporate spatial modeling into regional development strategies in rapidly urbanizing areas of Texas.

Forecasting doi: 10.3390/forecast8020024

Authors: Stephan Kolassa

The symmetric Mean Absolute Percentage Error (sMAPE) is a forecast error metric that has been proposed as an alternative to the more common Mean Absolute Percentage Error (MAPE), which is undefined whenever an actual is zero; the sMAPE does not have this problem. Thus, the sMAPE at first glance appears to be more suitable for evaluating forecasts of low volume or intermittent count demand time series. However, the sMAPE suffers from a number of other shortcomings; e.g., it is 2 for a zero actual regardless of the forecast, it always rewards (elicits) integer forecasts 0, 1, 2, …, if actuals are counts, and it elicits a (typically useless) zero forecast for sufficiently intermittent actuals. This paper collects such properties and discusses their real-world implications so the forecaster can make an informed decision as to whether to use the sMAPE or an alternative. In our opinion, the sMAPE is either unnecessary or dangerous; it should not be used.

Forecasting doi: 10.3390/forecast8020023

Authors: Igor Barbosa de Andrade Duarte Márcio Poletti Laurini

This paper examines when textual information from central bank communication improves forecasts of policy rate changes. Using the minutes of the Brazilian Central Bank’s Monetary Policy Committee (Copom), we study whether textual content helps predict changes in the Selic target rate between consecutive meetings. The minutes are encoded using dense sentence-level embeddings, and low-dimensional textual factors are extracted via principal component analysis estimated exclusively on the training sample to prevent look-ahead bias. Predictive performance is assessed out of sample using an expanding-window backtesting framework and compared against standard forecasting benchmarks, including persistence and random-walk specifications, linear autoregressive models, regularized regressions, and state-space models estimated via the Kalman filter. We find that text-based predictors perform poorly when used in isolation but deliver meaningful forecast improvements when combined with short-run dynamics and regularization. These gains are economically relevant and arise primarily in episodes associated with policy rate adjustments, whereas simple persistence-based forecasts remain difficult to outperform during rate-hold periods. Overall, the results indicate that central bank communication contains forward-looking information that is valuable for forecasting policy changes, but that this information is sparse, episodic, and best extracted through disciplined regularization and dynamic modeling rather than purely cross-sectional textual signals.

Forecasting doi: 10.3390/forecast8020022

Authors: Yijia Zhao Chaohui Chen Yongqiang Jiang Jiajun Li Xiong Chen Jiwen Zhang

In the context of global warming, the prediction of extreme precipitation events faces great challenges, especially the ensemble forecast of convective-scale heavy precipitation. Taking the heavy rainstorm in Zhengzhou on 20 July 2021 as an example, this paper aims to explore the performance of the convective-scale ensemble forecasting system based on the local breeding model cultivation method (LBGM) in extreme precipitation forecasting, and reveal the key physical mechanisms affecting the quality of forecasting. The traditional scoring (TS, Bias), neighborhood FSS and Contiguous Rain Area (CRA) methods were used to systematically evaluate the precipitation forecast, and the superior and inferior forecast members were diagnosed and analyzed by combining physical quantities such as isentropy vortex, relative vorticity, and water vapor flux divergence. The results show that: (1) the LBGM-EPS system can better capture the spatial distribution and intensity of heavy precipitation, which is better than the single deterministic forecast; (2) The CRA method is better than the traditional score in describing the spatial structure and intensity of precipitation, and can effectively identify the good and bad members of the forecast. (3) The reason why the dominant forecast members perform better is that the simulation of the dynamic-thermal structure of the mesoscale convective vortex is more reasonable, especially the coupling mechanism of the downward transmission of the high-level vortex and the convergence of water vapor at the lower level is better. The preliminary application of convective-scale ensemble forecasting based on the LBGM in this study has reference value for improving the prediction ability of extreme precipitation.

Forecasting doi: 10.3390/forecast8020021

Authors: Katleho Makatjane Diteboho Xaba

This paper offers a hybrid forecasting approach that merges a local-level state space Kalman filter with a Long-Short-Term Memory (LSTM) neural network to assess the downside risk of the Botswana Pula versus the US Dollar (BWP/USD). Inspired by the inability of conventional econometric models to capture complex latent structural shifts and nonlinear patterns, our architecure uses a bottom-up hierarchical methodology in which the smoothed level component of the exchange rate is isolated by the Kalman filter and subsequently fed into the LSTM architecture. Three key indicators for assessing downside risk—Maximum Drawdown (MDD), Conditional Drawdown-at-Risk (CDaR), and Downside Deviation—are used to assess model performance across various time-frames (7, 30, 90, 180, and 240 days). As confirmed by Kupiec and Christoffersen’s backtesting processes, the findings show a high degree of alignment between projected and actual values, with negligible downside deviation bias and robust calibration. Moreover, global economic and geopolitical shocks, such as the COVID-19 pandemic, the Russia–Ukraine conflict, and the 2015–2016 Shanghai Stock Exchange crash, are important factors that influence exchange rate volatility, according to explainable artificial intelligence techniques, particularly SHAP (SHapley Additive exPlanations) analysis. Downside risk is also greatly increased by regional currency links, especially the impact of the ZAR/BWP exchange rate. On the other hand, domestic temporal variables, such as week, quarter, and month, have very little impact. These results emphasise how Botswana’s currency rate is structurally vulnerable to external shocks and how crucial it is to include both global and regional considerations in risk analysis. The research concludes that the accuracy and transparency of projections for exchange rate risk significantly improve when practical filtering is combined with deep learning and explainable AI. To improve macroeconomic resilience and guide successful financial risk management plans in emerging market environments, policymakers are advised to employ AI-driven forecasting techniques, enhance regional monetary coordination, and set up real-set learning systems.

Forecasting doi: 10.3390/forecast8020020

Authors: Muhammad Hanif

The study documents the impact of the external sector on movements of the Pakistan Stock Exchange (PSX), covering conventional and Islamic indices. Selected variables include international trade, foreign investment, remittances, oil, gold, and currency markets, as well as the KSE-100 and KMI-30 indices. The sample period covers the latest 130 months, from 2015/01 to 2025/10. Results are documented through descriptive statistics, pairwise correlations, and OLS regression. Stability of coefficients during the review period is checked by calculating BTC-Var and switching Var. Outstanding momentum is evident in market indices (in the final phase), accompanied by growth in remittances, while the national currency has experienced an alarming depreciation. The combined impact of the external sector is not in the higher range for either index (adjusted R-square values are low). A group of four variables (remittances, oil, gold, and currency markets) was significant for the conventional index, while a group of three variables (oil, gold, and currency markets) was significant for the Islamic index. All significant variables contribute positively to stock index movements, except the exchange rate. BTC-Var and switching var suggest instability of relationships and regime-dependent var dynamics. The findings are beneficial for managers and investors in predicting index movements and portfolio diversification, as well as for relevant authorities in making policy decisions that promote prudent exchange-rate management and facilitate remittances. To the best of the author’s knowledge, this study is among the few that jointly examine the impact of external-sector variables on stock market movements.

Forecasting doi: 10.3390/forecast8020019

Authors: Priyanka Aggarwal Nevi Danila Eddy Suprihadi Manoj Kumar Manish

This study investigates regime-dependent forecasting of the Saudi stock market by combining macro-controlled dependence analysis with nonlinear predictive modeling. Using daily data from September 2010 to August 2025, we analyze the interaction between the Tadawul All Share Index (TASI) returns and crude oil returns while controlling for inflation and interest-rate dynamics. A four-variable VAR with macro controls is estimated separately in pre- and post-COVID regimes to characterize directional predictability and changes in transmission lags. We then evaluate out-of-sample return forecasting performance across econometric benchmarks (ARIMA, ARIMAX, and VAR) and machine learning models (LSTM and XGBoost) under a strictly time-ordered expanding-window design with sequential train/validation/test partitioning. The results indicate that traditional linear benchmarks exhibit limited predictive ability in both regimes, with negative out-of-sample explanatory power. By contrast, XGBoost delivers the strongest overall performance, achieving positive out-of-sample R2 in both regimes (0.046 in pre-COVID and 0.010 in post-COVID), together with the lowest forecast errors (RMSE = 0.0081 pre-COVID; 0.0078 post-COVID). Interpretability analysis further reveals a regime-sensitive shift in drivers: short-horizon equity lag dynamics dominate during stable periods, whereas oil-related and macro-financial variables gain importance under turbulent conditions. Economic-value evaluation supports the practical relevance of these gains, showing that XGBoost-based signals yield superior risk-adjusted trading outcomes and remain favorable under downside-risk and drawdown-based assessment. Overall, these findings highlight that forecasting in oil-linked emerging markets is inherently regime-dependent and that nonlinear ensemble learners, particularly XGBoost, provide a more robust and economically meaningful approach under structural change.

Forecasting doi: 10.3390/forecast8010018

Authors: Nkosinathi Emmanuel Radebe Bomi Cyril Nomlala Frank Ranganai Matenda

Persistent fiscal stress in South African municipalities undermines service delivery, yet practical tools for early detection remain limited. This study predicts one-year-ahead municipal financial distress to support risk-based prioritisation. We develop machine learning models using a 2018/19–2022/23 municipality panel, combining 13 financial health indicators from State of Local Government (SoLG) reports with selected socio-economic variables. Penalised logistic regression is benchmarked against random forest and XGBoost under a leakage-aware, time-ordered split into training, validation, and an out-of-time test year; class imbalance is handled through class weighting. Performance is evaluated using PR-AUC, ROC-AUC, calibration, and a capacity-constrained Top-30 rule. All models outperform a naïve last-year baseline on the out-of-time test (PR-AUC 0.934–0.954; ROC-AUC 0.886–0.923), with bootstrap intervals supporting robustness. Random forest performs best overall, while penalised logistic regression remains competitive. Under the Top-30 rule (12.3% workload), precision is high (precision@30 0.967–1.000) while recall is modest (recall@30 0.186–0.192). SHAP values and logistic odds ratios identify liquidity, solvency, cash coverage, and employment deprivation as key drivers. The Top-30 rule corresponds to an annual intensive monitoring portfolio that is reasonable under constrained staffing and budget capacity in national and provincial oversight units, while probability thresholds are reported as conventional benchmarks rather than as policy triggers.

Forecasting doi: 10.3390/forecast8010017

Authors: Alberto Garinei Massimiliano Proietti Alessandro Vispa Stefano Speziali Giovanni Bartolini Marcello Marconi Emanuele Piccioni Matteo Martini Francesca Fallucchi Romeo Giuliano Ernesto William De Luca Umberto Di Matteo Valerio Lemma

The SAIFIN project (Satellite data and Artificial Intelligence for FINtech) develops a novel algorithmic trading system that integrates satellite imagery, financial data, and advanced artificial intelligence to enhance decision-making, particularly in commodity and agricultural markets. This paper presents the motivation, design, implementation, and validation of the SAIFIN framework. Leveraging alternative data and modular multi-agent architectures, SAIFIN aims to deliver robust, context-aware trading signals in diverse market conditions.

Forecasting doi: 10.3390/forecast8010016

Authors: Letizia Caravella Stefania Gentili

New Zealand, located along the boundary between the Pacific and Australian plates, is among the most seismically active regions in the world. In such an area, reliable short-term forecasting of strong aftershocks is essential for seismic risk mitigation. In this study, we apply NESTORE (NExt STrOng Related Earthquake), a machine learning probabilistic forecasting algorithm, to the New Zealand earthquake catalogue to evaluate the probability that a mainshock of magnitude Mm will be followed by an event of magnitude ≥ Mm − 1 within a defined space–time window. NESTORE uses nine features describing early post-mainshock seismicity and outputs the probability that a cluster is Type A (i.e., containing a strong aftershock) or not (Type B). We assess performance using two testing strategies: chronological training–testing splits and k-fold cross-validation and refine the training set using the REPENESE outlier-detection procedure. The k-fold approach proves more robust than the chronological one, despite changes in catalogue characteristics over time. Eighteen hours after the mainshock, NESTORE correctly classified 88% of clusters (75% for Type A and 92% for Type B; Precision = 0.75). Notably, the highly destructive 2010–2011 Canterbury–Christchurch sequence was correctly identified as Type A. These findings support the applicability of NESTORE for short-term aftershock forecasting in New Zealand.

Forecasting doi: 10.3390/forecast8010015

Authors: Wentian Lu Zhenming Lu Wenjie Liu Yifeng Cao

Accurate wind power forecasting is critical for enhancing the operational efficiency and stability of electrical power grids. Conventional single-variable signal decomposition forecasting methods ignore the coupling relationship between wind power and multiple meteorological data, thus limiting prediction accuracy. This study proposes an accurate and fast short-term wind power prediction approach based on series-core fusion technology considering multiple meteorological data. In the data preprocessing stage, the multivariate variational mode decomposition (MVMD) algorithm decomposes wind power and meteorological variables into the same predefined number of frequency-aligned intrinsic mode functions (IMFs), thereby enhancing feature representation and improving forecasting accuracy via a more comprehensive and detailed dataset representation. During the training stage, the series-core fused time series (SOFTS) model establishes the connection among wind power channel and other meteorological variable channels for each IMF, achieving fast convergence through its streamlined and parallel structure. In the forecasting stage, the final wind power prediction is generated by the reconstruction of all IMFs. Furthermore, we conducted a comprehensive performance evaluation by comparing the proposed MVMD-SOFTS model with eight alternative models, including the CNN model, the TCN model, the LSTM model, the GRU model, the Transformer model, the SOFTS model, the CEEMDAN-SOFTS model, and the VMD-SOFTS model. The results indicate that MVMD-SOFTS outperformed all other models, demonstrating its effectiveness in capturing the multifaceted relationships in wind power forecasting.

Forecasting doi: 10.3390/forecast8010014

Authors: Umawadee Detthamrong Rapeepat Klangbunrueang Wirapong Chansanam Rasita Dasri

Sustainable corporate governance plays an essential role in promoting responsible economic growth and enhancing social and environmental well-being in emerging economies. In this context, Environmental, Social, and Governance (ESG) performance has become an important indicator of a firm’s commitment to sustainable development and its alignment with the United Nations Sustainable Development Goals, particularly SDG 8 and SDG 12. This study investigates the impact of Environmental, Social, and Governance (ESG) performance on the financial sustainability of publicly listed companies in Thailand, a rapidly developing Southeast Asian economy where empirical evidence remains limited. Using an unbalanced panel dataset of 965 firm-year observations across multiple industries, multiple regression models were employed to assess the influence of ESG performance on two financial indicators: return on capital employed and return on assets. Granger causality tests were also conducted to explore directional relationships between sustainability performance and financial outcomes. The empirical results reveal a significant negative short-term association between ESG performance and return on assets (ROA), whereas the relationship with return on capital employed (ROCE) is statistically insignificant. The causality analysis indicates that ESG performance Granger-causes ROA, implying that sustainability-driven strategic decisions may precede and influence financial outcomes over time. Additionally, leverage emerges as a key constraint to financial sustainability, negatively affecting both ROCE and ROA. These findings underscore the challenge of striking a balance between sustainability investments and immediate profitability in emerging markets. Policymakers and business leaders are encouraged to promote supportive governance frameworks, reduce financial barriers, and foster ESG-driven practices that contribute to long-term sustainable competitiveness and inclusive development.

Forecasting doi: 10.3390/forecast8010013

Authors: Simla Durmus Deniz Demirhan Ismail Gultepe Onur Durmus

The main objective of this work is to characterize Sudden Stratospheric Warming (SSW) conditions and their impact on local weather forecasting and climate change, using SSW definition criteria. The SSWs strongly affect Arctic vortex structure and midlatitude weather conditions. This work evaluates the frequency, amplitude, and dynamical–thermal characteristics of SSWs under historical and Representative Concentration Pathway (RCP) 4.5 scenarios, focusing on stratospheric air temperature (Ts) and zonal wind speed (Uh) at the 10° N and 60° N latitudes. The fifth-generation ECMWF atmospheric reanalysis (ERA5) is employed as the reference dataset. Simulations of five Coupled Model Intercomparison Project Phase 5 (CMIP5) models, represented by M1 to M5, are analyzed. The primary group of models included 1) the Australian Community Climate and Earth-System Simulator, version 1.3 (ACCESS1-3, M1), 2) the Hadley Center Global Environmental Model, version 2—Carbon Cycle (HadGEM2-CC, M2), and 3) the Max Planck Institute Earth System Model—Medium Resolution (MPI-ESM-MR, M3). The analysis period covers SSW events related to the Quasi-Biennial Oscillation (QBO) in the Northern Hemisphere (NH) from 1980 to 2100. The key findings indicate that while M1, M2, and M3 simulate SSW occurrence correctly for the 21st century, they exhibit significant systematic deficiencies in capturing the structural dynamics of SSW events. Specifically, the M1, M2, and M3 models underestimate the polar stratospheric temperature amplitude (Tamp) by approximately 75–80% and zonal wind amplitude (Uamp) by more than 60% compared to the ERA5 analysis. Furthermore, ERA5 exhibits a strong negative correlation (R ≈ −0.8) between Uh and Ts that is not estimated accurately using the present models. The importance of the horizontal resolution of the models and wave–mean flow interactions in determining SSW intensity and occurrence is also found to be a critical metric. Results suggest that SSW definition criteria affect Arctic and midlatitude weather system prediction at a rate of 61–82%. It is concluded that the primary configurations of CMIP5 models for accurately capturing the dynamical structure and evolution of QBO–SSW interactions are needed, and that they affect future projections of SSW events.

Forecasting doi: 10.3390/forecast8010012

Authors: Mariusz Ptak Mariusz Sojka Soufiane Haddout Teerachai Amnuaylojaroen

Temperature is a fundamental property of water that determines its quality and the course of both biotic and physical processes. Therefore, the distribution and future changes in thermal conditions are crucial for the functioning of the hydrosphere. In this study, a hybrid air2water model was used to determine the course of the sea surface temperature, which allows for its prediction using a minimal set of input data based on the air temperature. The widespread availability of air temperature measurements worldwide offers broad potential for the model’s application, which is especially important in coastal zones—the most dynamic and diverse areas of marine ecosystems, and simultaneously the most exposed to anthropogenic pressure. The study analyzes four hydrological stations in the southern part of the Baltic Sea, where the results confirm the high predictive capabilities of the air2water model for sea surface temperature. Depending on the adopted climate change scenarios, the average rate of sea surface temperature increase by the end of the 21st century is projected to be 0.15 °C per decade (SSP2-4.5) and 0.33 °C per decade (in the case of the SSP5-8.5 scenario). If these projections come true, they should be considered unfavorable, and such a situation will require taking into account changes in the thermal regime in the functioning of the Baltic Sea. More broadly, this simple yet effective method for predicting thermal conditions may be applied in interdisciplinary research as well as in the management of coastal marine zones.

Forecasting doi: 10.3390/forecast8010011

Authors: Juan Huang Ching-Wu Chu Hsiu-Li Hsu

The Baltic Dry Index (BDI) measures the cost of transporting dry bulk commodities such as coal, iron ore, and grain. As a key indicator of global trade, supply chain dynamics, and overall economic activity, accurate short-term forecasting of the BDI is crucial. This paper compares six univariate methods to obtain a more precise short-term BDI prediction model, providing valuable insights for decision-makers. The six forecasting techniques include Grey Forecast, ARIMA, Support Vector Regression, LSTM, GRU and EMD-SVR-GWO. Model performance is evaluated using three common metrics: Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). Our findings reveal that the novel EMD-SVR-GWO model outperforms the other univariate methods, demonstrating superior accuracy in forecasting monthly BDI trends. This study contributes to improved BDI prediction, aiding managers in strategic planning and decision-making.

Forecasting doi: 10.3390/forecast8010010

Authors: Lersak Phothong Anupong Sukprasert Sutana Boonlua Prapaporn Chubsuwan Nattakron Seetha Rotcharin Kunsrison

Accurate early-warning forecasting of corporate financial distress remains a critical challenge due to nonlinear financial relationships, severe data imbalance, and the high operational costs of false alarms in risk-monitoring systems. This study proposes an explainable voting ensemble framework for early-warning forecasting of corporate financial distress using lagged accounting-based financial information. The proposed framework integrates heterogeneous base learners, including Decision Tree, Neural Network, and k-Nearest Neighbors models, and is evaluated using financial statement data from 752 publicly listed firms in Thailand, comprising sixteen financial ratios across six dimensions: liquidity, operating efficiency, debt management, profitability, earnings quality, and solvency. To ensure robustness under imbalanced and rare-event conditions, the study employs feature selection, data normalization, stratified cross-validation, resampling techniques, and repeated validation procedures. Empirical results demonstrate that the proposed Voting Ensemble delivers a precision-oriented and decision-relevant forecasting profile, outperforming classical classifiers and maintaining greater early-warning reliability when benchmarked against advanced tree-based ensemble models. Probability-based evaluation further confirms the robustness and calibration stability of the proposed framework under repeated cross-validation. By adopting a forward-looking, early-warning perspective and integrating ensemble learning with explainable machine learning principles, this study offers a transparent and scalable approach to financial distress forecasting. The findings offer practical implications for auditors, investors, and regulators seeking reliable early-warning tools for corporate risk assessment, particularly in emerging market environments characterized by data imbalance and heightened uncertainty.

Forecasting doi: 10.3390/forecast8010009

Authors: Sergei Soldatenko Genrikh Alekseev Vladimir Loginov Yaromir Angudovich Irina Danilovich

Improving the accuracy of climate and long-range meteorological forecasts is an important objective for many economic sectors: agriculture, energy and utilities, transportation and logistics, construction, disaster risk management, insurance and finance, retail, tourism and leisure. Traditional physical models face limitations at ultra-long lead times, which motivates the development of empirical–statistical approaches, including those leveraging deep learning techniques. In this study, using ERA5 reanalysis data and archives of major climate indices for the period 1950–2024, we examine statistical relationships between climate indices associated with large-scale atmospheric and oceanic patterns in the Northern Hemisphere and surface air temperature anomalies in selected mid- and high-latitude regions. The aim is to assess the predictive skill of these indices for seasonal temperature anomalies within empirical forecasting frameworks. To this end, we employ cross-correlation and cross-spectral analyses, as well as regression modeling. Our findings indicate that the choice of the most informative predictors strongly depends on the target region and season. Among the major indices, AMO and EA/WR emerge as the most informative for forecasting purposes. The Niño 4 and IOD indices can be considered useful predictors for the Eastern Arctic. Notably, the strongest correlations between the AMO, EA/WR, Niño 4, and IOD indices and surface air temperature occur at one- to two-year lags. To illustrate the predictive potential of the four selected indices, several multiple regression models were developed. The results obtained from these models confirm that the chosen set of indices effectively captures the main sources of variability relevant to seasonal and interannual temperature prediction across the analyzed regions. In particular, approximately 64% of the forecasts have errors less than 0.674 times the standard deviation.

Forecasting doi: 10.3390/forecast8010008

Authors: Anne-Marie Sassenberg Nirmal Acharya Padmaja Kar Mohammad Sadegh Eshaghi

AI Recommender Systems (RecSys) function as personalised forecasting engines, predicting user preferences to reduce information overload. However, the efficacy of these systems is often bottlenecked by the “Last Mile” of forecasting: the end-user’s willingness to adopt and rely on the prediction. While the existing literature often assumes that algorithmic accuracy (e.g., low RMSE) automatically drives utilisation, empirical evidence suggests that users frequently reject accurate forecasts due to a lack of trust or cognitive friction. This study challenges the utilitarian view that users adopt systems simply because they are useful, instead proposing that sustainable adoption requires a state of Cognitive Absorption—a psychological flow state enabled by the Cognitive Economy of trust. Grounded in the Motivation–Opportunity–Ability (MOA) framework, we developed the Trust–Absorption–Intention (TAI) model. We analysed data from 366 users of a major predictive platform using Partial Least Squares Structural Equation Modelling (PLS-SEM). The Disjoint Two-Stage Approach was employed to model the reflective–formative Higher-Order Constructs. The results demonstrate that Cognitive Trust (specifically the relational dimensions of Benevolence and Integrity) operates via a dual pathway. It drives adoption directly, serving as a mechanism of Cognitive Economy where users suspend vigilance to rely on the AI as a heuristic, while simultaneously freeing mental resources to enter a state of Cognitive Absorption. Affective Trust further drives this immersion by fostering curiosity. Crucially, Cognitive Absorption partially mediates the relationship between Cognitive Trust and adoption intention, whereas it fully mediates the impact of Affective Trust. This indicates that while Cognitive Trust can drive reliance directly as a rational shortcut, Affective Trust translates to adoption only when it successfully triggers a flow state. This study bridges the gap between algorithmic forecasting and behavioural adoption. It introduces the Cognitive Economy perspective: Trust reduces the cognitive cost of verifying predictions, allowing users to outsource decision-making to the AI and enter a state of effortless immersion. For designers of AI forecasting agents, the findings suggest that maximising accuracy may be less effective than minimising cognitive friction for sustaining long-term adoption. To solve the cold start problem, platforms should be designed for flow by building emotional rapport and explainability, thereby converting sporadic users into continuous data contributors.

Forecasting doi: 10.3390/forecast8010007

Authors: Jie Ji Shouyang Wang Yunjie Wei

To address the nonlinear nature of exchange rates where drivers vary by time horizon, this paper proposes a CEEMDAN-PE-CatBoost-SHAP framework. Analyzing USD/CNY data (2012–2024), we decomposed rates into high, medium, and low frequencies to bridge machine learning with economic interpretability. Empirical results revealed distinct frequency-dependent drivers: high-frequency fluctuations depend on market sentiment; medium-frequency variations follow Fed policies; and low-frequency trends reflect fundamentals like gold prices. SHAP analysis provides transparent attribution of these factors. This multi-scale approach isolates heterogeneous drivers, offering policymakers and investors a nuanced paradigm for managing currency risks. The study significantly clarifies how different economic factors shape exchange rate dynamics across varying time scales.

Forecasting doi: 10.3390/forecast8010006

Authors: Andrea Tri Rian Dani Nur Chamidah I. Nyoman Budiantara Budi Lestari Dursun Aydin

We introduce the Semiparametric Time Series Regression with Mixed Additive Spline Fourier (STSR–MASF) model as an innovative approach for analyzing time series data with complex patterns. The model combines the flexibility of the spline estimator in capturing nonlinear variations across specific sub-intervals and the strength of the Fourier series in representing periodically recurring patterns. Within the semiparametric regression framework, STSR–MASF integrates both linear parametric and nonparametric components, with the optimal number of knots and oscillations determined using the Generalized Cross-Validation (GCV) criterion. The model was trained and tested using Earth’s skin temperature data from the National Aeronautics and Space Administration (NASA) MERRA-2 for East Kalimantan, Indonesia, a tropical rainforest region. The results demonstrate that the STSR–MASF model provides more accurate estimations and forecasts compared to six previous methods proposed in earlier studies with highly accurate predictions. This innovation not only offers methodological advancements in nonlinear time series modeling, but also contributes practical insights into understanding variations in Earth’s skin temperature in tropical regions, supporting broader efforts toward global climate change mitigation.

Forecasting doi: 10.3390/forecast8010005

Authors: Ioannis Stergiou Nektaria Traka Dimitrios Melas Efthimios Tagaris Rafaella-Eleni P. Sotiropoulou

Accurate air quality forecasting remains challenging due to persistent biases in chemical transport models. Addressing this challenge, the current study develops pollutant-specific deep learning frameworks that correct systematic errors in the Community Multiscale Air Quality (CMAQ) simulations of nitrogen dioxide (NO2) and coarse particulate matter (PM10) over California. Building upon a previous study on ozone bias correction, a hybrid CNN–Attention–LSTM architecture is adapted, and a weighted Huber loss function is introduced for PM10 to enhance the detection of extreme pollution events through a gated tail-weighting mechanism. Using data from twenty EPA monitoring stations (ten per pollutant) for 2010–2014, the proposed approach achieves substantial performance gains over the CMAQ baseline. For NO2, RMSE decreases by ~51% with an average systematic bias reduction of ~80% and a random error reduction of ~42%. For PM10, RMSE improves by ~49% while the systematic and random errors decrease by ~94% and ~33%, respectively. The PM10 model also shows high consistency with observations (Index of Agreement improvement of ~105%) and a strong ability to capture peak events (F1 score improvement of ~270%), while the NO2 model achieves large gains in explanatory power (R2 improvement averaging ~816%). Both pollutants also demonstrate enhanced temporal agreement between predictions and observations, as confirmed by the Dynamic Time Warping analysis (NO2: ~55%, PM10: ~58%). These results indicate that pollutant-specific loss functions and architectural tuning can significantly improve both accuracy and event sensitivity, offering a transferable framework for bias correction across multiple pollutants and regions.

Forecasting doi: 10.3390/forecast8010004

Authors: Jae-Hyeok Seok Hee-Wook Choi Sang-Sam Lee

This study employed tree-based machine learning (ML) algorithms to predict low-level wind shear (LLWS) at Jeju International Airport (ICAO: RKPC). Hourly meteorological data from 47 observation stations across Jeju Island, collected between 2019 and 2023, were split into training (60%), validation (20%), and test (20%) sets to develop individual prediction models for lead times ranging from 1 to 6 h. A probabilistic prediction model was developed by assigning weights to individual models according to their true skill statistic performance. Validation using an independent 2024 dataset showed that the light gradient boosting machine-based probabilistic model exhibited the highest predictive performance, achieving an area under the receiver operating characteristic curve of 0.883. The Shapley additive explanation analysis identified wind components (U, V) as key variables, contributing over 50%, with the significance of pressure and temperature slightly increasing over long-term prediction times (4–6 h). In addition, spatial analysis revealed that nearby airport stations were more influential for short-term prediction times (1–2 h), whereas Mount Halla and offshore stations north of the airport gained greater influence for medium-to long-term prediction times (3–6 h). The ML-based LLWS prediction model offers high accuracy and interpretability, supporting stepwise warning systems and aiding aviation decision-making.

Forecasting doi: 10.3390/forecast8010003

Authors: Luis Conde-López Monica Borunda Gerardo Ruiz-Chavarría Tomás Aparicio-Cárdenas

Short-term load forecasting is fundamental for the effective and reliable operation of power systems. Very accurate forecasting methods often involve complex hybrid approaches that combine statistical, physical, and/or intelligent techniques. In this work, we present an innovative, clear, and effective methodology for short-term hourly peak load forecasting that is both simple and highly accurate. The methodology is based on the load forecast used for electricity market purposes, together with fine-tuning dynamic estimation. As a case study, the methodology was applied and tested in Mexico’s interconnected power system. It was implemented across various regions and at both regional and load-\ zone levels of this interconnected power system and, even under a variety of standard and extreme load conditions, achieved outstanding results.

Forecasting doi: 10.3390/forecast8010002

Authors: Lee-Wen Yang Nguyen Thi Thanh Binh Jiang Meng Yi

This study compares Logit, Probit, Extreme Value, and Artificial Neural Network (ANN) models using data from 2012 to 2024 in the Taiwan electronics industry. ANN outperforms traditional models, achieving 98% accuracy in predicting financial distress. Two robust distress signals are identified: Return on Assets (threshold: 7.03%) and Total Asset Growth (threshold: −9.05%). The nonlinear impacts of financial distress on variables are analyzed, with a focus on contextual considerations in decision-making. These findings bring attention to the importance of utilizing advanced techniques like ANN for improved predictive accuracy, offering profound clarification for risk assessment and management.

Forecasting doi: 10.3390/forecast8010001

Authors: Zhihui Li Zhiming Han Huanwei Zhang Qixiang Liao

The high-precision prediction of near-space atmospheric temperature holds significant importance for aerospace, national defense security, and climate change research. To address the deficiencies of extracting features in conventional convolutional neural networks, this paper designs a ConvLSTM hybrid model that combines the spatiotemporal feature extraction capability of 3D convolution with a residual attention mechanism, effectively capturing the dynamic evolution patterns of the near-space temperature field. The comparative analysis with various models, including GRU, shows that the proposed model demonstrates superior performance, achieving an RMSE of 2.433 K, a correlation coefficient R of 0.993, and an MRE of 0.76% on the test set. Seasonal error analysis reveals that the prediction stability is better in winter than in summer, with errors in the mesosphere primarily stemming from the complexity of atmospheric processes and limitations in data resolution. Compared to traditional CNNs and single time-series models, the proposed method significantly enhances prediction accuracy, providing a new technical approach for near-space environmental modeling.

Forecasting doi: 10.3390/forecast7040080

Authors: Romulo Oliveira Deivid Campos Katia Bicalho Bruno Macêdo Matteo Bodini Camila Saporetti Leonardo Goliatt

Unconfined Compressive Strength (UCS) of stabilized soils is commonly used for evaluating the effectiveness of soil improvement techniques. Achieving target UCS values through conventional trial-and-error approaches requires extensive laboratory experiments, which are time-consuming and resource-intensive. Automated Machine Learning (AutoML) frameworks offer a promising alternative by enabling automated, reproducible, and accessible predictive modeling of UCS values from more readily obtainable index and physical soil and stabilizer properties, reducing the reliance on experimental testing and empirical relationships, and allowing systematic exploration of multiple models and configurations. This study evaluates the predictive performance of five state-of-the-art AutoML frameworks (i.e., AutoGluon, AutoKeras, FLAML, H2O, and TPOT) using analyses of results from 10 experimental datasets comprising 2083 samples from laboratory experiments spanning diverse soil types, stabilizers, and experimental conditions across many countries worldwide. Comparative analyses revealed that FLAML achieved the highest overall performance (average PI score of 0.7848), whereas AutoKeras exhibited lower accuracy on complex datasets; AutoGluon , H2O and TPOT also demonstrated strong predictive capabilities, with performance varying with dataset characteristics. Despite the promising potential of AutoML, prior research has shown that fully automated frameworks have limited applicability to UCS prediction, highlighting a gap in end-to-end pipeline automation. The findings provide practical guidance for selecting AutoML tools based on dataset characteristics and research objectives, and suggest avenues for future studies, including expanding the range of AutoML frameworks and integrating interpretability techniques, such as feature importance analysis, to deepen understanding of soil–stabilizer interactions. Overall, the results indicate that AutoML frameworks can effectively accelerate UCS prediction, reduce laboratory workload, and support data-driven decision-making in geotechnical engineering.

Forecasting doi: 10.3390/forecast7040079

Authors: Syarifah Inayati Nur Iriawan Irhamah Uha Isnaini

This research tackles the challenge of forecasting nonlinear time series data with stochastic structural variations by proposing the Markov switching autoregressive model with time-varying parameters (MSAR-TVP). Although effective in modeling dynamic regime transitions, the Classical MSAR-TVP faces challenges with complex datasets. To address these issues, a Bayesian MSAR-TVP framework was developed, incorporating flexible parameters that adapt dynamically across regimes. The model was tested on two periods of U.S. real GNP data: a historically stable segment (1952–1986) and a more complex, modern segment that includes more economic volatility (1947–2024). The Bayesian MSAR-TVP demonstrated superior performance in handling complex datasets, particularly in out-of-sample forecasting, outperforming the Bayesian AR-TVP, Classical MSAR-TVP, and Classical MSAR models, as evaluated by mean absolute percentage error (MAPE) and mean absolute error (MAE). For in-sample data, the Classical MSAR-TVP retained its stability advantage. These findings highlight the Bayesian MSAR-TVP’s ability to address parameter uncertainty and adapt to data fluctuations, making it highly effective for forecasting dynamic economic cycles. Additionally, the two-year forecast underscores its practical utility in predicting economic cycles, suggesting continued expansion. This reinforces the model’s significance for economic forecasting and strategic policy formulation.

Forecasting doi: 10.3390/forecast7040078

Authors: Kathleen Marshall Park Sarthak Pattnaik Natasya Liew Triparna Kundu Ali Ozcan Kures Eugene Pinsky

Pharmaceutical manufacturing and logistics rely on accurate prediction and decision making to safeguard product quality, delivery reliability, and patient outcomes. Despite rapid advances in artificial intelligence (AI) and machine learning (ML), few studies benchmark model performance across the diverse operational demands of global pharmaceutical supply chains. Predictive setbacks contribute to financial losses, reduced supply chain efficacy, and potential adverse health consequences, yet understanding these failures offers firms opportunities to refine strategy and strengthen resilience. Drawing on 1.2 million shipments spanning 39 countries, we compare traditional statistical models (ARIMA), ensemble methods (random forests, gradient boosting), and deep neural networks (LSTM, GRU, CNN, ANN) across pricing, demand forecasting, vendor management, and shipment planning. Gradient boosting produced the strongest pricing performance, while ARIMA delivered the lowest demand-forecasting errors but with limited explanatory power; neural networks captured nonlinear demand shocks and achieved superior maintenance-risk classification. We also identified three vendor performance clusters—high-performing, cost-efficient, and mixed-reliability vendors—enabling firms to better align shipment criticality with vendor capabilities by prioritizing high performers for urgent deliveries, leveraging cost-efficient vendors for non-urgent volumes, and managing mixed performers through targeted oversight. These insights highlight the value of our evidence-based roadmap for selecting algorithms in high-stakes healthcare logistics, in rapidly evolving, technologically complex global contexts where increasing algorithmic sophistication elevates the standards for safer, smarter pharmaceutical supply chains.

Forecasting doi: 10.3390/forecast7040077

Authors: Marko Corn Anže Murko Primož Podržaj

This study explores the role of decentralized physical infrastructure networks (DePINs) in enhancing solar energy forecasting, focusing on how network density influences prediction accuracy and economic viability. Using machine learning models applied to production data from 47 residential PV systems in Utrecht, Netherlands, we developed a hierarchical forecasting framework: Level 1 (clear-sky baseline without historical data), Level 2 (solo forecasting using only local historical data), and Level 3 (networked forecasting incorporating data from neighboring installations). The results show that networked forecasting substantially improves accuracy: under solo forecasting conditions (Level 2), the Random Forests model reduces Mean Absolute Error (MAE) by 17% relative to the Level 1 baseline, and incorporating all available neighbors (Level 3) further reduces the MAE by an additional 34% relative to Level 2, corresponding to a total improvement of 45% compared with Level 1. The largest accuracy gains arise from the first 10–15 neighbors, highlighting the dominant influence of local spatial correlations. These forecasting improvements translate into significant economic benefits. Imbalance costs decrease from EUR 1618 at Level 1 to EUR 1339 at Level 2 and further to EUR 884 at Level 3, illustrating the financial impact of both solo and networked data sharing. A marginal benefit analysis reveals diminishing returns beyond approximately 10–15 neighbors, consistent with spatial saturation effects within 5–10 km radii. These findings provide a quantitative foundation for incentive mechanisms in DePIN ecosystems and demonstrate that privacy-preserving data sharing mitigates data fragmentation, reduces imbalance costs for energy traders, and creates new revenue opportunities for participants, thereby supporting the development of decentralized energy markets.

Forecasting doi: 10.3390/forecast7040076

Authors: Sean Guidry Stanteen Jianzhong Su Paul Flanagan Xunchang John Zhang

This study introduces a novel k-nearest neighbors (kNN) method of forecasting precipitation at weather-observing stations. The method identifies numerous monthly temporal patterns to produce precipitation forecasts for a specific month. Compared to climatological forecasts, which average the observed precipitation over the prior thirty years, and other existing contemporary iterations of kNN, the proposed novel kNN method produces more accurate forecasts on a consistent basis. Specifically, the novel kNN method produces improved root mean square errors (RMSE), mean relative errors, and Nash–Sutcliffe coefficients when compared to climatological and other kNN forecasts at five weather stations in Oklahoma. Rather than looking at the daily data for feature vectors, this novel kNN method takes so many days and evenly groups them, using the resulting average as one feature each. All methods tested were lacking in the ability to forecast wet extremes; however, the novel kNN method produced more frequent higher precipitation forecasts compared to climatology and the two other kNN methods tested.

Forecasting doi: 10.3390/forecast7040075

Authors: Mahan Hajiabbasi Somehsaraie Soheyla Tofighi Zhaoan Wang Jun Wang Shaoping Xiao

Time series models are considered among the most intricate models in machine learning. Due to sharp temporal variations, time series models normally fall short in predicting the peaks or local minima accurately. To overcome this challenge, we proposed a novel custom loss function, Enhanced Peak (EP) loss, specifically designed to pinpoint peaks and troughs in time series models, to address underestimations and overestimations in the forecasting process. EP loss applies an adaptive penalty when prediction errors exceed a specified threshold, encouraging the model to focus more effectively on these regions. To evaluate the effectiveness and versatility of EP loss, the loss function was tested on three highly variable datasets: NOx emissions, streamflow measurements, and gold price, implementing Gated Recurrent Unit and Transformer-based models. The results consistently demonstrated that EP loss significantly mitigates peak prediction errors compared to conventional loss functions, highlighting its potential for highly variable time series applications.

Forecasting doi: 10.3390/forecast7040074

Authors: Nikolaos Kanellos Dimitrios Katsianis Dimitris Varoutas

This paper presents a novel system dynamics-based framework for forecasting market share evolution in the telecommunications sector. The framework conceptualizes market share as flows of subscribers—driven by churn, attraction, and market growth—between interconnected compartments representing providers. It is designed to operate with limited available market data and incorporates stochastic processes to capture market uncertainty, enabling risk-informed forecasts. The framework is applied to the Greek mobile telecommunications market using historical data (2006–2022), with a 5-year hold-back period for validation. Results highlight the dominant role of churn management in market share variability, particularly for the incumbent provider Cosmote, while subscriber attraction parameters show moderate influence for alternative providers Vodafone and Wind Hellas. Sensitivity analysis confirms the model’s robustness and identifies key drivers of forecast variability. The proposed framework provides actionable insights for strategic decision-making, making it a valuable tool for providers and policymakers to address churn, optimize attraction strategies, and ensure long-term competitiveness in dynamic markets.

Forecasting doi: 10.3390/forecast7040073

Authors: Nehalben Ranabhatt Sérgio Barreto Marco Pimpão Pedro Prates

The automobile industry is one of the world’s largest manufacturing sectors and a key contributor to economic growth. Demand forecasting plays a critical role in supply chain management within the automotive sector. Reliable forecasts are essential for production planning, inventory control, and meeting market demands efficiently. However, accurately predicting demand remains a challenge due to the influence of external factors such as socioeconomic trends and weather conditions. This study presents a systematic literature review of the forecasting methods employed within the automotive industry, encompassing both vehicle and spare parts demand. Following PRISMA guidelines, 63 publications were identified and analyzed, covering traditional statistical models such as ARIMA and SARIMA, as well as state-of-the-art artificial intelligence approaches, including artificial neural networks. The review finds that classical statistical models remain prevalent for vehicle demand forecasting, Croston’s method dominates spare parts forecasting, and AI-based techniques increasingly outperform conventional models in recent studies. Furthermore, the review compiles a broad set of external variables influencing demand and highlights the common challenges associated with demand forecasting. It concludes by outlining potential directions for future research.

Forecasting doi: 10.3390/forecast7040072

Authors: Xu Wang Yingjie Liu Zhenao Guo Tengfei Yang Xu Gong Zhichong Lyu

Accurately predicting carbon trading price is challenging due to pronounced nonlinearity, non-stationarity, and sensitivity to diverse factors, including macroeconomic conditions, market sentiment, and climate policy. This study proposes a novel hybrid forecasting framework that integrates multidimensional news text analysis, ICEEMDAN (Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) decomposition, and machine learning to predict carbon prices in China’s pilot trading prices. We first extract a market sentiment index from news texts in the WiseSearch News Database using a customized Chinese carbon-market dictionary. In addition, a price trend index and topic intensity index are derived using Latent Dirichlet Allocation (LDA) and Convolutional Neural Networks (CNN), respectively. All feature sequences are subsequently decomposed and reconstructed using sample-entropy-based ICEEMDAN approach. The resulting multi-frequency components were then used as inputs for a range of machine-learning models to evaluate predictive performance. The empirical results demonstrate that the incorporation of multidimensional text information on China’s carbon market, combined with financial features, yields a substantial gain in prediction accuracy. Our integrated decomposition-ensemble framework achieves optimal performance by employing dedicated models—BiGRU, XGBoost, and BiLSTM for the high-frequency, low-frequency, and trend components, respectively. This approach provides policymakers, regulators, and investors with a more reliable tool for forecasting carbon prices and supports more informed decision-making, offering a promising pathway for effective carbon-price prediction.

Forecasting doi: 10.3390/forecast7040071

Authors: Turan Cansu Eren Bas Tamer Akkan Erol Egrioglu

Classical time series methods are widely employed to analyze linear time series with a limited number of observations; however, their effectiveness relies on several strict assumptions. In contrast, artificial neural networks are particularly suitable for forecasting problems due to their data-driven nature and ability to address both linear and nonlinear challenges. Furthermore, recurrent neural networks feed the output back into the network as input, utilizing this feedback mechanism to enrich the information provided to the model. This study proposes a novel recurrent hybrid intuitionistic forecasting method utilizing a modified pi–sigma neural network, principal component analysis (PCA), and simple exponential smoothing (SES). In the proposed framework, lagged time series variables and principal components derived from the membership and non-membership values of an intuitionistic fuzzy clustering method are used as inputs. A modified particle swarm optimization (PSO) algorithm is employed to train this new hybrid network. By integrating PCA, modified pi–sigma neural networks (MPS-ANNs), and SES within a recurrent hybrid structure, the model simultaneously captures linear and nonlinear dynamics, thereby enhancing forecasting accuracy and stability. The performance of the proposed model is evaluated using diverse financial and environmental datasets, including CMC-Open (I–IV), NYC water consumption, OECD freshwater use, and ROW series. Comparative results indicate that the proposed method achieves superior accuracy and stability compared to other fuzzy-based approaches.

Forecasting doi: 10.3390/forecast7040070

Authors: Laurențiu-Gabriel Frâncu Alexandra Constantin Maxim Cetulean Diana Andreia Hristache Monica Maria Dobrescu Raluca Andreea Popa Alexandra-Ioana Murariu Roxana Lucia Ungureanu

Policymakers in small open economies need reliable signals of incipient private consumption downturns, yet traditional indicators are revised, noisy, and often arrive too late. This study develops a Romanian-specific early warning system that combines a time-varying parameter VAR with stochastic volatility and exogenous drivers (TVP-SV-VARX) with modern machine learning classifiers. The structural layer extracts regime-dependent anomalies in the macro-financial transmission to household demand, while the learning layer transforms these anomalies into calibrated probabilities of short-term consumption declines. A strictly time-based evaluation design with rolling blocks, purge and embargo periods, and rare-event metrics (precision–recall area under the curve, PR-AUC, and Brier score) underpins the assessment. The best-performing specification, a TVP-filtered random forest, attains a PR-AUC of 0.87, a ROC-AUC of 0.89, a median warning lead of one quarter, and no false positives at the chosen operating point. A sparse logistic calibration model improves probability reliability and supports transparent communication of risk bands. The time-varying anomaly layer is critical: ablation experiments that remove it lead to marked losses in discrimination and recall. For implementation, the paper proposes a three-tier WATCH–AMBER–RED scheme with conservative multi-signal confirmation and coverage gates, designed to balance lead time against the political cost of false alarms. The framework is explicitly predictive rather than causal and is tailored to data-poor environments, offering a practical blueprint for demand-side macroeconomic early warning in Romania and, by extension, other small open economies.

Forecasting doi: 10.3390/forecast7040069

Authors: Xi Lu Jieni Li Rajender R. Aparasu Nebil Yusuf Cen Wu

There is a growing interest in applying statistical machine learning methods, such as LASSO regression and its extensions, to analyze healthcare datasets. The existing study has examined LASSO and group LASSO regression with categorical predictors that are widely used in healthcare studies to represent variables with nominal or ordinal categories. Despite the success of these studies, statistical inference procedures and quantifying uncertainty for regression with categorical predictors have largely been overlooked, partly due to the theoretical challenges practitioners face when applying these methods in behavioral research. In this article, we aim to fill this gap by investigating from a Bayesian perspective. Specifically, we conduct Bayesian LASSO analysis with categorical predictors under different coding strategies, and thoroughly investigate the impact of four representative coding strategies on variable selection and prediction. In particular, we have conducted uncertainty quantification in terms of marginal Bayesian credible intervals by leveraging the advantage that fully Bayesian analysis can enable exact statistical inference even on finite samples. In this study, we demonstrate that the variable selection, estimation and prediction of Bayesian LASSO are influenced by the coding strategies with the real-world Medical Expenditure Panel Survey (MEPS) data. The performance of Bayesian LASSO has also been compared with LASSO and linear regression.

Forecasting doi: 10.3390/forecast7040068

Authors: Dean Fantazzini

In this study, we extend research on stablecoin credit risk by introducing a novel rule-of-thumb approach to determine whether a stablecoin is “dead” or “alive” based on a simple price threshold. Using a comprehensive dataset of 98 stablecoins, we classify a coin as failed if its price falls below a predefined threshold (e.g., $0.80), validated through sensitivity analysis against established benchmarks such as CoinMarketCap delistings and Feder et al. (2018) methodology. We employ a wide range of panel binary models to forecast stablecoins’ probabilities of default (PDs), incorporating stablecoin-specific regressors. Our findings indicate that panel Cauchit models with fixed effects outperform other models across different definitions of stablecoin failure, while lagged average monthly market capitalization and lagged stablecoin volatility emerge as the most significant predictors—outweighing macroeconomic and policy-related variables. Random forest models complement our analysis, confirming the robustness of these key drivers. This approach not only enhances the predictive accuracy of stablecoin PDs but also provides a practical, interpretable framework for regulators and investors to assess stablecoin stability based on credit risk dynamics.

Forecasting doi: 10.3390/forecast7040067

Authors: Song Nie Zhi-Qiang Jiang

Amid escalating global climate change and geopolitical tensions threatening food supply chains, the three provinces of Northeast China, which serve as a major grain production base, play a crucial role in ensuring national food security. However, the region is experiencing more frequent extreme climatic events and increasing limitations on arable land. This necessitates an evaluation of the combined effects of climate conditions and sown area on rice (Oryza sativa L.) yields. Utilizing provincial panel data from 1990 to 2022, this study conducts baseline panel regression analyses at both the national and Northeast China levels. The results consistently identify the value of the standardized precipitation evapotranspiration index (SPEI) on September as a key climatic factor exerting a significant negative effect on rice total yield, whereas the rice sown area is a robust positive determinant. Based on these findings, we develop a dual-factor analytical framework that incorporates both climatic conditions and rice sown area, utilizing SPEI1-Sep. to identify critical growth stages of rice, with the aim of providing a more comprehensive understanding of their combined effects on yield. To further support predictive accuracy, the comparative performance assessments of the Extreme Gradient Boosting (XGBoost), random forest (RF), and Autoregressive Integrated Moving Average (ARIMA) models are conducted. The results show that the ARIMA model outperforms others in forecasting. Forecasts for 2023–2027 indicate slow yield growth in Jilin Province, with a 1.5% annual increase. Heilongjiang shows minor fluctuations, stabilizing between 24.97 and 25.56 million tons. Liaoning’s yield remains stable, projected between 5.13 and 5.20 million tons. These trends suggest limited overall yield expansion, highlighting the need for region-specific policies and resource management to ensure China’s grain security. This study clarifies the interplay between climate and sown area, demonstrates the relative forecasting advantage of ARIMA in this setting, and provides evidence to support managing yield variability and optimizing agricultural policy in Northeast China, with implications for long-term national food security.

Forecasting doi: 10.3390/forecast7040066

Authors: Adam Booth Philip James Stephen McGough Ellis Solaiman

Accurately forecasting air quality could lead to the development of dynamic, data-driven policy-making and improved early warning detection systems. Deep learning has demonstrated the potential to produce highly accurate forecasting models, but it is noted that much literature focuses on narrow datasets and typically considers one geographic area. In this research, three diverse air quality datasets are utilised to evaluate four deep learning algorithms, which are feedforward neural networks, Long Short-Term Memory (LSTM) recurrent neural networks, DeepAR and Temporal Fusion Transformers (TFTs). The study uses these modules to forecast CO, NO2, O3, and particulate matter 2.5 and 10 (PM2.5, PM10) individually, producing a 24 h forecast for a given sensor and pollutant. Each model is optimised using a hyperparameter and a feature selection process, evaluating the utility of exogenous data such as meteorological data, including wind speed and temperature, along with the inclusion of other pollutants. The findings show that the TFT and DeepAR algorithms achieve superior performance over their simpler counterparts, though they may prove challenging in practical applications. It is noted that while some covariates such as CO are important covariates for predicting NO2 across all three datasets, other parameters such as context length proved inconsistent across the three areas, suggesting that parameters such as context length are location and pollutant specific.

Forecasting doi: 10.3390/forecast7040065

Authors: Efstratios Bilis Theophilos Papadimitriou Konstantinos Diamantaras Konstantinos Goulianas

This study introduces a Transformer-based forecasting tool termed EXPERT (EXchange rate Prediction using Encoder Representation from Transformers) and applies it to exchange rate forecasting. We developed and trained a Transformer-based forecasting model, then evaluated its performance on nine currency pairs with various characteristics. Finally, we benchmarked its effectiveness against six established forecasting models: Linear Regression, Random Forest, Stochastic Gradient Descent, XGBoost, Bagging Regression, and Long Short-Term Memory. Our dataset covers the period from 1999 to 2022. The models were evaluated for their ability to predict the next day’s closing price using three performance metrics. In addition, the EXPERT system was evaluated on its ability to extend forecast horizons and as the core of a trading strategy. The model’s robustness was further evaluated using the Multiple Comparisons with the Best (MCB) metric on five dataset samples.

Forecasting doi: 10.3390/forecast7040064

Authors: Daniele Girolimetto

In many different applications such as retail, energy, and tourism, forecasts for a set of related time series must satisfy both linear and non-negativity constraints, as negative values are meaningless in practice. Standard regression-based reconciliation approaches achieve coherence with linear constraints, but may generate negative forecasts, reducing interpretability and usability. This paper develops and evaluates three alternative strategies for non-negative forecast reconciliation. First, reconciliation is formulated as a non-negative least squares problem and solved with the operator splitting quadratic program, allowing flexible inclusion of additional constraints. Second, we propose an iterative non-negative reconciliation with immutable forecasts, offering a practical optimization-based alternative. Third, we investigate a family of set-negative-to-zero heuristics that achieve efficiency and interpretability at minimal computational cost. Using the Australian Tourism Demand dataset, we compare these approaches in terms of forecast accuracy and computation time. The results show that non-negativity constraints consistently improve accuracy compared to base forecasts. Overall, set-negative-to-zero achieve near-optimal performance with negligible computation time, the block principal pivoting algorithm provides a good accuracy–efficiency compromise, and the operator splitting quadratic program offers flexibility for incorporating additional constraints in large-scale applications.

Forecasting doi: 10.3390/forecast7040063

Authors: László Vancsura Tibor Tatay Tibor Bareith

Accurate inflation forecasting is of central importance for monetary authorities, governments, and businesses, as it shapes economic decisions and policy responses. While most studies focus on headline inflation, this paper analyses the Harmonised Index of Consumer Prices (HICP) and its 12 subcomponents in the euro area over the period 2000–2023, covering episodes of financial crisis, economic stability, and recent inflationary shocks. We apply a broad set of machine learning and deep learning models, systematically optimized through grid search, and evaluate their performance using the Normalized Mean Absolute Error (NMAE). To complement traditional accuracy measures, we introduce the Forecastability Index (FI) and the Interquartile Range (IQR), which jointly capture both the difficulty and robustness of forecasts. Our results show that RNN and LSTM architectures consistently outperform traditional approaches such as SVR and RFR, particularly in volatile environments. Subcomponents such as Health and Education proved easier to forecast, while Recreation and culture and Restaurants and hotels were among the most challenging. The findings demonstrate that macroeconomic stability enhances forecasting accuracy, whereas crises amplify errors and inter-model dispersion. By highlighting the heterogeneous predictability of inflation subcomponents, this study provides novel insights with strong policy relevance, showing which categories can be forecast with greater confidence and where uncertainty requires more cautious intervention.

Forecasting doi: 10.3390/forecast7040062

Authors: Harrison Katz Thomas Maierhofer

Accurate forecasts of the U.S. renewable energy consumption mix are essential for planning transmission upgrades, sizing storage, and setting balancing market rules. We introduce a Bayesian Dirichlet ARMA model (BDARMA) tailored to monthly shares of hydro, geothermal, solar, wind, wood, municipal waste, and biofuels from January 2010 through January 2025. The mean vector is modeled with a parsimonious VAR(2) in additive log ratio space, while the Dirichlet concentration parameter follows an intercept plus five Fourier harmonics, allowing for seasonal widening and narrowing of predictive dispersion. Forecast performance is assessed with a 61-split rolling origin experiment that issues twelve month density forecasts from January 2019 to January 2024. Compared with three alternatives (a Gaussian VAR(2) fitted in transform space, a seasonal naive approach that repeats last year’s proportions, and a drift-free ALR random walk), BDARMA lowers the mean continuous ranked probability score by 15 to 60 percent, achieves componentwise 90 percent interval coverage near nominal, and maintains point accuracy (Aitchison RMSE) on par with the Gaussian VAR through eight months and within 0.02 units afterward. These results highlight BDARMA’s ability to deliver sharp and well-calibrated probabilistic forecasts for multivariate renewable energy shares without sacrificing point precision.

Forecasting doi: 10.3390/forecast7040061

Authors: Yu Chao Nur Fazidah Elias Yazrina Yahya Ruzzakiah Jenal

Financial sustainability in higher education is increasingly fragile due to policy shifts, rising costs, and funding volatility. Legacy early-warning systems based on static thresholds or rules struggle to adapt to these dynamics and often overlook fairness and interpretability—two essentials in public-sector governance. We propose a university financial risk early-warning framework that couples a causal-attention Transformer with Multi-Objective Bayesian Optimization (MBO). The optimizer searches a constrained Pareto frontier to jointly improve predictive accuracy (AUC↑), fairness (demographic parity gap, DP_Gap↓), and computational efficiency (time↓). A sparse kernel surrogate (SKO) accelerates convergence in high-dimensional tuning; a dual-head output (risk probability and health score) and SHAP-based attribution enhance transparency and regulatory alignment. On multi-year, multi-institution data, the approach surpasses mainstream baselines in AUC, reduces DP_Gap, and yields expert-consistent explanations. Methodologically, the design aligns with LLM-style time-series forecasting by exploiting causal masking and long-range dependencies while providing governance-oriented explainability. The framework delivers earlier, data-driven signals of financial stress, supporting proactive resource allocation, funding restructuring, and long-term planning in higher education finance.

Forecasting doi: 10.3390/forecast7040060

Authors: Vahdettin Demir

Accurate precipitation forecasting plays a crucial role in sustainable water resource management, especially in arid regions like Konya, one of Turkey’s driest areas. Reliable forecasts support effective water budgeting, agricultural planning, and climate adaptation efforts in the region. This study investigates the performance of different deep learning training algorithms in forecasting monthly precipitation using Long Short-Term Memory (LSTM) networks, a method tailored for time-series prediction. A comprehensive dataset comprising 39 years (1984–2022) of precipitation records was utilized, obtained from the Turkish State Meteorological Service (MGM) as ground-based observations and from NASA’s POWER database as remote sensing data, and was split into 80% for training and 20% for testing. A comparative analysis of three widely used optimization algorithms, Adaptive Moment Estimation (ADAM), Root Mean Square Propagation (RMSProp), and Stochastic Gradient Descent with Momentum (SGDM), revealed that ADAM consistently outperformed the others in forecasting accuracy. Model performance was evaluated with statistical metrics, and the LSTM-ADAM combination achieved the best results. In the final phase, cross-validation was applied using MGM and NASA data sources in a crosswise manner to test model generalizability and data source independence. The best performance was observed when the model was trained with MGM data and tested with NASA data, achieving a remarkably low RMSE of 3.62 mm, MAE of 2.93 mm, R2 of 0.9966, and NSE of 0.9686. When trained with NASA data and tested with MGM data, the model still demonstrated strong performance, with an RMSE of 4.48 mm, MAE of 3.22 mm, R2 of 0.9921, and NSE of 0.9678. These results demonstrate that satellite and ground-based data can be used interchangeably under suitable conditions, while also confirming the superiority of the ADAM optimizer in LSTM-based precipitation forecasting.

Forecasting doi: 10.3390/forecast7040059

Authors: Guilherme Armando de Almeida Pereira Kiara de Deus Demura

This study evaluates the effect of simple data-level balancing techniques on predicting school dropout across all state public high schools in Espírito Santo, Brazil. We trained Logistic Regression with LASSO (LR), Random Forest (RF), and Naive Bayes (NB) models on first-quarter data from 2018–2019 and forecasted dropouts for 2020, with additional validation in 2022. Facing strong class imbalance, we compared three balancing methods—RUS, SMOTE, and ROSE—against models trained on the original data. Performance was assessed using accuracy, sensitivity, specificity, precision, F1, AUC, and G-mean. Results show that the imbalance severely harmed RF and NB trained without balancing, while Logistic Regression remained more stable. Overall, balancing techniques improved most metrics: RUS and ROSE were often superior, while SMOTE produced mixed results. Optimal configurations varied by year and metric, and RUS and ROSE made up most of the best combinations. Although most configurations benefited from balancing, some decreased performance; therefore, we recommend systematic testing of multiple balancing strategies and further research into SMOTE variants and algorithm-level approaches.

Forecasting doi: 10.3390/forecast7040058

Authors: Latif Bukari Rashid Shahzada Zaman Shuja Shafiqur Rehman

As the world is shifting toward cleaner energy sources, accurate forecasting of solar radiation is critical for optimizing the performance and integration of solar energy systems. In this study, we explore eight machine learning models, namely, Random Forest Regressor, Linear Regression Model, Artificial Neural Network, k-Nearest Neighbors, Support Vector Regression, Gradient Boosting Regressor, Gaussian Process Regression, and Deep Learning, as to their use in forecasting direct solar radiation across six climatically diverse regions in the Kingdom of Saudi Arabia. The models were evaluated using eight statistical metrics along with time-series and absolute error analyses. A key contribution of this work is the introduction of Trigonometric Cyclical Encoding, which has significantly improved temporal representation learning. Comparative SHAP-based feature-importance analysis revealed that Trigonometric Cyclical Encoding enhanced the explanatory power of temporal features by 49.26% for monthly cycles and 53.30% for daily cycles. The findings show that Deep Learning achieved the lowest root mean square error, as well as the highest coefficient of determination, while Artificial Neural Network demonstrated consistently high accuracy across the sites. Support Vector Regression performed optimally but was less reliable in some regions. Error and time-series analyses reveal that Artificial Neural Network and Deep Learning maintained stable prediction accuracy throughout high solar radiation seasons, whereas Linear Regression, Random Forest Regressor, and k-Nearest Neighbors showed greater fluctuations. The proposed Trigonometric Cyclical Encoding technique further enhanced model performance by maintaining the overall fitness of the models, which ranged between 81.79% and 94.36% in all scenarios. This paper supports the effective planning of solar energy and integration in challenging climatic conditions.

Forecasting doi: 10.3390/forecast7040057

Authors: Carlo Grillenzoni

Positive percentage time series are present in many empirical applications; they take values in the continuous interval (0,1) and are often modeled with linear dynamic models. Risks of biased predictions (outside the admissible range) and problems of heteroskedasticity in the presence of asymmetric distributions are ignored by practitioners. Alternative models are proposed in the statistical literature; the most suitable is the dynamic beta regression which belongs to generalized linear models (GLM) and uses the logit transformation as a link function. However, owing to the Jensen inequality, this approach may also not be optimal in prediction; thus, the aim of the present paper is the in-depth forecasting comparison of linear and beta autoregressions. Simulation experiments and applications to nonstationary time series (the US unemployment rate and BR hydroelectric energy) are carried out. Rolling regression for time-varying parameters is applied to both linear and beta models, and a prediction criterion for the joint selection of model order and sample size is defined.

Forecasting doi: 10.3390/forecast7040056

Authors: Cristian Lozano Tafur Jaime Orduy Rodríguez Pedro Melo Daza Iván Rodríguez Barón Danny Stevens Traslaviña Juan Andrés Bermúdez

This research introduces a system designed to predict three-dimensional airspace occupancy over Colombia using historical Automatic Dependent Surveillance-Broadcast (ADS-B) data and machine learning techniques. The goal is to support proactive air traffic management by estimating future aircraft positions—specifically their latitude, longitude, and flight level. To achieve this, four predictive models were developed and tested: K-Nearest Neighbors (KNN), Random Forest, Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM). Among them, the LSTM model delivered the most accurate results, with a Mean Absolute Error (MAE) of 312.59, a Root Mean Squared Error (RMSE) of 1187.43, and a coefficient of determination (R2) of 0.7523. Compared to the baseline models (KNN, Random Forest, XGBoost), these values represent an improvement of approximately 91% in MAE, 83% in RMSE, and an eighteen-fold increase in R2, demonstrating the substantial advantage of the LSTM approach. These metrics indicate a significant improvement over the other models, particularly in capturing temporal patterns and adjusting to evolving traffic conditions. The strength of the LSTM approach lies in its ability to model sequential data and adapt to dynamic environments—making it especially suitable for supporting future Trajectory-Based Operations (TBO). The results confirm that predicting airspace occupancy in three dimensions using historical data are not only possible but can yield reliable and actionable insights. Looking ahead, the integration of hybrid neural network architectures and their deployment in real-time systems offer promising directions to enhance both accuracy and operational value.