Search Results (10,822)

Accurate cryptocurrency price forecasting represents an important yet challenging problem in financial time-series analysis due to the highly volatile, nonlinear, and noise-sensitive nature of digital asset markets. Although transformer-based architectures have recently demonstrated strong capabilities in temporal sequence modeling, their behavior under high-frequency cryptocurrency dynamics and the role of explainability-guided feature refinement remain insufficiently explored. To address this gap, this study presents a comprehensive transformer-based forecasting framework for hourly cryptocurrency price prediction and investigates the impact of explainability-guided feature optimization on forecasting performance, robustness, and interpretability. Five transformer architectures—Vanilla Transformer, Informer, Autoformer, Reformer, and Temporal Fusion Transformer (TFT)—are systematically evaluated across five major cryptocurrency assets: Bitcoin (BTC), Ethereum (ETH), Solana (SOL), Dogecoin (DOGE), and Ripple (XRP). The experimental framework employs Open, High, Low, Close, and Volume (OHLCV) data together with a broad set of engineered technical indicators and evaluates model performance using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). To improve interpretability and reduce feature redundancy, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) are integrated directly into the forecasting pipeline. Based on the resulting explanations, asset-specific feature subsets are constructed, and all models are subsequently retrained using the refined feature representations. The results show that explainability-guided feature refinement provides compact, model-aware, and interpretable feature subsets with competitive forecasting performance; however, its effect on prediction accuracy is dependent on the cryptocurrency asset, transformer architecture, retained feature subset size, and market conditions. Additional robustness, sensitivity, alternative feature-selection, and statistical significance analyses indicate that the SHAP–LIME Top-15 subset should be interpreted as a conservative dimensionality-reduction strategy rather than a universally optimal feature-selection rule. The findings further reveal that transformer architectures incorporating sparse attention, decomposition mechanisms, or gating structures generally provide stronger performance than the Vanilla Transformer under highly volatile hourly market conditions. Overall, the proposed framework demonstrates that combining transformer-based forecasting with explainability-guided feature refinement can support interpretable and parsimonious high-frequency financial time-series modeling, while highlighting the importance of evaluating robustness, feature-selection sensitivity, and statistical variability alongside average forecasting errors. Full article

Electric vehicle (EV) parking lots can create concentrated charging demand that couples operating cost, grid capacity use, and carbon emissions. This paper proposes a reinforcement-learning-guided two-stage multi-energy optimization framework for carbon-aware EV parking lot charging. The reinforcement learning (RL) layer uses a Deep Q-Network (DQN) to generate a data-driven charging reference from state-of-charge (SoC), time-to-departure, vehicle presence, electricity price, and grid-load information. This profile is transferred to a two-stage optimization model as a behavioral reference rather than being used as the final dispatch schedule. The first stage determines baseline operation and residual grid headroom, while the second stage schedules EV charging together with Power-to-Gas (P2G) and Carbon Capture and Storage (CCS) decisions under capacity, carbon-budget, and multi-energy constraints. A soft-tracking formulation links the learned profile with the optimized schedule and allows the tracking coefficient to shape different operating regimes. The results show that the proposed framework improves grid feasibility, reduces peak charging stress, and enhances carbon-aware operation. CCS mainly supports carbon-budget feasibility, whereas P2G provides additional value when renewable surplus is available. Full article

International migration is often considered an important factor in housing market dynamics, yet its relationship with residential property prices remains uncertain, particularly in developing economies. This study examined whether changes in international migration are associated with changes in residential property prices in South Africa, both in the short term and over a longer period. Annual data covering 1976–2022 were analysed using an unrestricted error-correction model and the Pesaran, Shin and Smith bounds-testing approach. The results show that changes in the number of international migrants were not significantly associated with changes in residential property prices in the short term. The analysis also found no evidence of a lasting relationship between international migration and residential property prices over the longer term. In other words, the data do not provide sufficient statistical evidence to conclude that changes in international migration were linked to changes in South African residential property prices during the study period. In contrast, the results show that changes in residential property prices tended to persist over time, with price changes in one year being positively associated with changes in the following year. These findings suggest that migration should not be assumed to be an important explanation for national housing-price movements without supporting empirical evidence. They also highlight the importance of examining short-term and long-term relationships separately when studying migration and housing-market dynamics. Full article

The vehicle–battery separation (VBS) model offers a viable framework for promoting new energy vehicle (NEV) adoption. To facilitate its broader diffusion, this study examines a supply chain comprising a vehicle manufacturer and a third-party operator for the battery leasing and swapping (BLS) services under this business model, from the perspective of automakers. Within this framework, three service operation strategies are developed for the automaker: (1) self-operated swapping services (VYN strategy); (2) self-operated leasing services (VNY strategy); and (3) full outsourcing of both services (VNN strategy). A comparative analysis yields three main findings. First, the automaker achieves the highest profit under strategy VNY, whereas the third-party operator prefers to cooperate with an automaker that adopts strategy VNN. Second, whether VNN or VNY yields higher total supply chain profit is determined by swapping price sensitivity and the vehicle’s base price: the VNN strategy is optimal when both parameters are low, and the VNY strategy is optimal otherwise. Third, to promote the adoption of battery-swappable vehicles, automakers should outsource battery swapping services (i.e., adopt strategy VNY or VNN). By contrast, leasing operations are better outsourced (through VYN or VNN) when the goal is to grow the swapping service market. Full article

Live-streaming commerce has become a critical information channel through which consumers evaluate products and make purchase decisions, yet brand manufacturers face substantial uncertainty when selecting live-streaming partners. This study investigates how streamer influence and consumer sensitivity to live-streaming service quality jointly shape optimal cooperation structures in platform-based commerce. We develop a game-theoretic decision framework comparing three cooperation modes—Nash negotiation, manufacturer-led, and streamer-led—and derive closed-form equilibria for pricing, service quality, and profit allocation. The results show that manufacturer-led cooperation consistently maximizes brand profit by preserving incentives to provide service quality. In contrast, streamer-led cooperation can reduce overall efficiency when dominant streamers lack motivation to improve service quality. To corroborate our theoretical predictions, we employ live-stream sales data from the Douyin (TikTok) platform, covering multiple streamer tiers and two cosmetic brands, and find empirical support for the model’s key insights. The findings contribute to the literature on information processing and incentive alignment in platform-based live-streaming commerce, offering actionable guidance for manufacturers’ partnership strategies. Full article

To address the challenge of quantifying the coupling relationship between user-side reporting and bidding behavior and electricity market operational risks, this study proposes a method for electricity market risk assessment and propagation identification based on entropy-weighted TOPSIS and directed temporal-dependency analysis. First, a risk indicator system is constructed across four dimensions, including electricity supply and demand, price volatility, system operation, and user-side bidding credibility. Robust quantile standardization and the entropy-weighted TOPSIS method are employed to construct category-specific and composite risk-state indices and warning levels. Subsequently, the PCMCI+ method is used to identify candidate temporal relationships among risk indicators, and risk propagation pathways are screened through stability and temporal direction tests; on this basis, a dynamic model with pathway constraints is established to quantify the propagation strength between different risk categories based on impulse responses. The PJM case study shows that the framework can distinguish elevated risk-state periods and identify a sparse full-summer directed temporal-dependency pattern, while sensitivity and chronological split-sample analyses reveal that the detailed propagation structure is specification- and period-dependent. These results support the use of the framework for multidimensional risk-state monitoring while also defining the empirical boundary of its propagation interpretation. Full article

This study explores the drivers of income inequality using crude oil prices, consumer prices, economic development, trade openness, unemployment, and human capital as predictors. Despite a lot of research conducted on income inequality, there have been very few studies capable of simultaneously analysing the influence of energy prices, macroeconomic conditions, trade openness, conditions in the labour market, and human capital, as well as accounting for the dynamic persistence and potential endogeneity of inequality across countries. To fill this research gap, based on a panel of 230 observations, analyses are run using pooled ordinary least squares, as well as fixed and random effects derived from model-selection tests that favour the random effects specification. An analysis using a dynamic panel data model was also carried out in this study. The empirical results show that crude oil prices, trade openness, and unemployment are all positively correlated with income inequality and that GDP per capita and human capital have a significant deterrent effect on the expansion of income inequality. With the random effects model, a US$1 increase in crude oil prices increases the Gini index by 0.0186 points; meanwhile, one-percentage-point increases in trade openness and unemployment contribute to increases in inequality of 0.0142 points and 0.722 points, respectively. On the other hand, a US$1000 increase in GDP per capita lowers the Gini index by about 0.0699 points, and an improvement of 0.1 point in human capital decreases inequality by approximately 1.497 points. Controlling for country-specific effects does little to show an association with consumer prices. The model accounts for 43.11 per cent of the variation in income inequality. The findings show that inclusive growth strategies, investment in human capital and labour-market policies are evidently pivotal to addressing inequality fostered by economic transformation and energy-price volatility. Full article

This study examines whether a structure-preserving representation of the joint Open-High-Low-Close (OHLC) vector provides coherent forecasts for China’s INE crude-oil futures and how seven benchmark models compare within that representation. The sample contains 1547 trading days from the contract launch to 8 January 2025. Under the originally reported 80:20 setting, GRU has the smallest aggregate MAPE (0.7825%), MAE (4.6774), and RMSE (7.2957), together with the largest interval-overlap Success Ratio (0.5629). These figures establish a numerical ranking only: the archived materials do not verify the exact temporal split, training-only preprocessing, component-specific errors, or price-limit-event robustness. The economic section therefore contains only illustrative Close-to-Close and Open-to-Close mappings and does not claim executable profitability. The defensible contribution is the structure-preserving OHLC framework and a bounded numerical comparison. Full article

A key question at the interface of sustainable finance and market efficiency is whether capital markets discipline firms not only for what they do—emissions, violations, or green investment—but also for how they communicate their environmental positions. Disclosure research suggests that the quality and verifiability of environmental narratives matter for investors, yet the pricing of narrative fuzziness itself remains underexplored. This study examines whether capital markets respond to the way high-polluting firms narrate green constraints in their annual reports. We construct a large-language-model-based fuzzy disclosure index (GFIX) from Chinese MD&A texts. Using only pre-2008 annual reports for scale learning, we train a two-stage comparator–scorer system from sentence-level pairwise judgments and apply it to MD&A sentences from 2008 to 2025. The resulting firm–year GFIX is matched with announcement-quarter stock returns and ownership data for 1016 high-polluting A-share firms. Double/debiased machine learning and panel fixed-effects estimates show that higher GFIX is systematically associated with significantly lower announcement-quarter returns, suggesting that investors treat fuzzy green narratives as a signal of information risk under green transition. The discount is concentrated in high-polluting manufacturing industries and firms with higher non-institutional ownership, while higher GFIX also predicts subsequent declines in institutional ownership. In addition, major Chinese green-regulatory milestones are followed by systematic increases in GFIX. These findings indicate that narrative clarity is an economically meaningful channel through which green transition and environmental regulation are transmitted to asset prices and ownership reallocations. These findings carry direct implications for sustainable finance: disclosure verifiability appears to be a first-order determinant of how green-transition risks are priced, and strengthening the verifiability of mandatory environmental disclosure can lower information risk, reduce the discount applied to opaque narrators, and support the orderly pricing of transition risks, in line with SDG 12 (responsible consumption and production) and SDG 13 (climate action). Full article

Postharvest fruit quality differs among various fruits and alters during sorting, packaging, transportation and storage. Therefore, it is important to have an efficient and objective grading method that causes little mechanical damage to ensure the uniformity of products and decrease market price and losses in the supply chain. This paper describes the grading norms, detection techniques, processing algorithms, structural designs and practical uses in different types of fruits. It examines the external features, internal quality and hidden faults by using machine vision, visible-near-infrared spectroscopy, hyperspectral and X-ray imaging, acoustic and mechanical detection, electronic noses and the integration of multiple sensors. In addition, it also investigates conventional machine learning, deep learning, transfer learning and lightweight implementations. The research has developed from grading according to size, weight and color to a complete evaluation of ripeness, juice volume, hardness, internal defects and shelf life. Moreover, single detection devices are joined together to construct integrated systems including feeding, separation, inspection, classification, redirection, packaging and data management. However, the application is restricted by the discrepancy between the grading criteria and measurable results, the lack of cross-species and batch generalization ability, and the difficulty in coordinating multiple sensors in real time. The systems should maintain a balance between production speed, mechanical damage and costs. Some matters needing attention are to standardize the quality description, choose multi-source fusion, develop adaptive lightweight models, design modular structures and guarantee end-to-end traceability. Solving these problems will facilitate the transition from accurate laboratory identification to reliable, economic and extensive commercial grading. Full article

This paper studies the valuation of continuously monitored geometric Asian options under a two-factor stochastic volatility model with stochastic long-term variance levels. We derive the joint characteristic function required for continuous geometric averaging and obtain analytical pricing formulas for fixed-strike call and put options through Fourier inversion. The analytical formulas are verified by comparison with Monte Carlo simulation. Numerical experiments show that option values are particularly sensitive to the initial variance levels, while changes in the trends of the stochastic long-term means and leverage correlations also affect prices across strikes. The proposed approach extends double Heston Asian option pricing while maintaining computational tractability, and it provides a flexible method for examining the impact of time-varying long-term volatility expectations. Full article

Forecasting equity returns remains challenging because financial markets exhibit nonlinear dynamics, volatility clustering, and complex temporal dependencies that are difficult to capture using a single modelling approach. Traditional statistical models can capture dependence and volatility dynamics, while deep learning models are capable of learning nonlinear temporal patterns. This study evaluates the forecasting performance of traditional statistical models (ARIMA and GARCH), deep learning models (TCN and GRU), and hybrid models (ARIMA-TCN, ARIMA-GRU, GARCH-TCN, and GARCH-GRU) for forecasting the FTSE/JSE Top 40 index in South Africa. Daily closing prices comprising 4159 observations from 2010 to 2026 were analysed, with model performance evaluated on log returns using MSE, RMSE, and MAE, with the Naïve model used as a benchmark. The results show that all competing models substantially outperformed the Naïve benchmark, while the TCN outperformed the standalone GRU. GARCH-based models also demonstrated strong forecasting performance, highlighting the relevance of volatility dynamics in forecasting equity returns. Although GARCH-GRU recorded the lowest MSE, RMSE, and MAE among the models evaluated, its performance was very similar to that of the GARCH (2,2) model. The Diebold-Mariano test further showed no statistically significant difference in predictive accuracy between GARCH-GRU and $GARCH (2,2)$ (DM = 0.9495; p = 0.3424). These findings indicate that GARCH-GRU and $GARCH (2,2)$ provide statistically comparable forecasting performance, suggesting that the additional complexity of the hybrid framework does not necessarily result in a significant improvement in predictive accuracy. This study contributes to the financial forecasting literature by providing empirical evidence from an emerging African equity market and demonstrating the importance of volatility modelling and complementary statistical–deep learning approaches for forecasting FTSE/JSE Top 40 log returns. Full article

Ultrasound-assisted extraction (UAE) processes for the production of standardized polyphenol-rich powders were evaluated as a valorization route for second- and third-grade apples, figs, and pomegranates from Central Macedonia, Greece. Based on experimental results, the scaled-up process was modeled using SuperPro Designer to evaluate a continuous facility treating 7920 tonnes·yr⁻¹ of feedstock. Material and energy balances revealed high utility demands, which were heavily mitigated through optimized closed-loop steam and cooling water regeneration systems. The techno-economic analysis showed that, driven by its high extraction yield (14.63 kg Gallic acid equivalent [GAE]·t⁻¹), the optimized pomegranate extraction was the only profitable scenario, yielding an annual operating cost of EUR 6.35 M/yr and a Net Present value (NPV) of EUR 11.81 M at a EUR 80/kg selling price; low-yield apple and fig lines remained highly unviable. Capacity sensitivity analysis showed that high fixed capital costs limit smaller plants, while market price sweeps identified a break-even point at EUR 66.03/kg. Using 1 kg gallic acid equivalent (GAE) as the functional unit, the life cycle assessment confirmed that pomegranate processing achieved the best environmental performance, primarily governed by reduced feedstock and utility requirements per unit of active extract. Toxicity-related categories constituted the main hotspots across all configurations, with the fig line acting as the worst absolute environmental performer. However, because the regional residue availability of 8483 tonnes·yr⁻¹ is dominated by apples, a standalone pomegranate processing plant is constrained at this scale, indicating that future work must focus on flexible multi-feedstock campaign processing or inter-regional feedstock aggregation. Full article

Ammonia is a promising carbon-free energy carrier for decarbonizing existing fossil fuel power plants. However, its economic viability and life-cycle environmental performance remain uncertain across different power generation technologies. This study develops an integrated techno-economic analysis (TEA) and life-cycle analysis (LCA) framework to evaluate 50% ammonia co-firing in four power generation systems: subcritical pulverized coal (SubC-CP), supercritical pulverized coal (SC-CP), integrated gasification combined cycle (IGCC), and natural gas combined cycle (NGCC). Process models modified from National Energy Technology Laboratory (NETL) simulations were used to determine mass and energy balances, while TEA quantified total plant cost, levelized cost of electricity (LCOE), breakeven ammonia price, and CO₂ avoidance cost (CAC). LCA evaluated well-to-gate (WTG) CO₂ emissions considering both blue and green ammonia. Results show that under the current blue ammonia price of $318/tonne, all ammonia co-firing cases exhibit higher LCOEs than the reference plants, and the breakeven ammonia prices remain below current market values. Blue ammonia reduces WTG CO₂ emissions by approximately 30% for coal-fired plants and 19% for NGCC where upstream ammonia production is the major emission source. Replacing blue ammonia with green ammonia further reduces WTG CO₂ emissions by 27–36% and decreases the CAC by 38–60%. These results demonstrate that low-carbon, low-cost ammonia production is essential for realizing economically competitive and environmentally sustainable ammonia co-firing in future low-carbon power systems. Full article

This study proposes a Bayesian Hierarchical Semiparametric Center and Range (BHS-CRM) model to address nonlinearity and heterogeneity in interval-valued panel data. By incorporating Bayesian penalized B-splines (P-splines) into a hierarchical random effects structure, the proposed method utilizes a parallel dual-component framework to separately capture dynamic central tendencies and variability. Simulation studies demonstrate the model’s robust performance, particularly in capturing nonlinear dynamics under high-noise conditions. Empirically, the model characterizes nonlinear temperature patterns in China’s Carbon Emission Trading Scheme (ETS) and achieves competitive out-of-sample predictive accuracy. Full article