Search Results (3,279)

The financial early warning mechanism of listed companies has an important strategic value for maintaining the stability of the capital market and preventing systemic financial risks. This study proposes a hybrid model (EBWO-BP) based on the improved beluga optimisation algorithm (EBWO) and BP neural network for financial early warning research. Innovative T-SNE nonlinear dimensionality reduction technique is applied to the multidimensional evaluation system constructed by 23 financial and two non-financial indicators. The empirical evidence based on the data of A-share listed companies in 2022–2024 shows that the accuracy of the EBWO-BP test set reaches 86.51% (AUC = 0.83), which demonstrates a significant prediction advantage compared with the optimisation algorithm models such as GA-BP and PSO-BP, as well as the CNN and LSTM deep learning models; when the sample size is increased to 700 groups, the accuracy is improved to 89.05%, verifying the model robustness. The method achieves significant improvement of financial risk prediction through algorithm fusion innovation, and provides methodological innovation and practical reference for intelligent financial risk monitoring. Full article

Accurate and non-destructive volume estimation of agricultural products is essential for precision agriculture, yet remains challenging when transitioning from controlled laboratory conditions to complex orchard environments. Although 2D image-based volume estimation methods provide a cost-effective and scalable solution, existing studies are fragmented and lack a unified perspective on their real-world applicability. This review presents a systematic synthesis of 2D image-based volume estimation methods, explicitly framed through the laboratory-to-orchard transition. We categorized existing volume estimation approaches according to the sensing modality into monocular RGB-based approaches and depth-assisted methods, and further reviewed them based on the image processing methods. A key finding is that high-precision geometric estimation can be achieved in laboratory environments, whereas deep learning and RGB-D fusion have driven a shift from conventional geometric modeling toward data-driven and hybrid learning frameworks in orchard settings. However, 2D image-based volume estimation remains fundamentally limited by scale ambiguity, severe occlusion, and sensitivity to illumination and background variability in real orchard environment. Overall, this review provides a unified perspective for understanding volume estimation methodology across environments and offers guidance for developing robust, scalable, and field-deployable volume estimation systems for real-world agricultural applications. Full article

Detecting and identifying faults in industrial chemical plants is essential for safe and stable operation, and modern monitoring systems increasingly rely on deep learning to classify faults from multivariate sensor data. A practical obstacle to adoption is trust: most deep-learning diagnosers reach their decisions through a classification layer that operators cannot inspect, making it hard to see how the model maps process signals to a particular fault. This study targets fault diagnosis on the Tennessee Eastman (TE) process, a standard benchmark of simulated chemical-plant sensor data, and asks whether this final decision stage can be made directly inspectable without sacrificing accuracy. We propose CNN–Transformer–KAN (CTKAN), a hybrid model that learns local temporal patterns with a one-dimensional convolutional encoder, captures global inter-time-step dependencies with a Transformer encoder, and classifies faults with a Kolmogorov–Arnold Network (KAN) head whose learnable B-spline activations can be plotted and examined individually, in place of a conventional multi-layer perceptron (MLP). On the TE benchmark, CTKAN attains a Macro-F1 of 91.38 ± 0.26% over ten independent runs, comparable to a CNN + Transformer + MLP ablation (91.21 ± 0.32%) and a capacity-matched MLP-head variant (91.43 ± 0.37%) within seed-to-seed variability. The main finding is therefore not a higher score: at matched capacity the KAN and MLP heads are statistically indistinguishable in accuracy, so the KAN head’s value is to add a directly inspectable view of the classification stage at no measurable accuracy cost, helping process engineers sanity-check how the diagnoser separates faults in safety-critical settings. Full article

The rapid evolution of data-driven enterprises demands scalable and intelligent systems capable of managing substantial volumes of heterogeneous data in real time. However, traditional systems lack a holistic approach to managing distributed data engineering, real-time analytics, and intelligent decision-making. To address these limitations, this paper proposes a Cognitive Lakehouse Framework that integrates distributed data processing, transformer-based deep learning, real-time analytics, and autonomous decision intelligence. Data are gathered from high-velocity, heterogeneous streams using Apache Kafka. Subsequently, data are processed using the hybrid batch/streaming paradigm, implemented via Apache Spark and Apache Flink, providing low latency and scalability. For data storage, a unified lakehouse layer is created using Delta Lake and Apache Iceberg, both of which support ACID transactions and schema evolution. In addition, transformer-based Deep Learning (DL) algorithms are utilized to capture temporal dependencies for predictive analytics, anomaly detection, and adaptive learning. Model lifecycle management is handled by MLflow, while ClickHouse and Apache Druid are used for real-time analytics. The architecture uses microservices and an event-driven approach on Kubernetes, and the workflow is automated with Apache Airflow. The performance assessment is conducted using TPC-H, TPC-DS, and real-time stream data to measure latency, throughput, and accuracy. Data quality, security, and compliance are provided by governance layers consisting of Apache Ranger and Apache Atlas. Experimental results show that significant gains can be made in terms of performance, with an accuracy of 98.5%, a query response time of 120 ms, a peak throughput of 85,000 records/s, and an end-to-end latency of 95 ms. Full article

Air pollution negatively impacts human health and environmental sustainability, particularly in areas with high industrial activity. This study comparatively evaluated deep learning-based models for estimating PM₁₀ and SO₂ pollutants in Dilovası and Ereğli (Turkey), industrial areas with high pollutant loads. The study utilized Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), Gated Recurrent Units (GRUs), an RNN–GRU stacked hybrid model, an attention-based hybrid model, and the proposed GRU–LSTM dual-path hybrid model. The proposed method consists of four main stages: data conversion into a time-series format, data preprocessing and feature generation, model architecture development, and model training and performance evaluation. The dataset consisted of 365 daily PM₁₀ and SO₂ observations obtained from the Air Monitoring Center for the Dilovası and Ereğli monitoring stations. Model performance was evaluated using the coefficient of determination (R²), training time, root mean squared error (RMSE), mean squared error (MSE), and mean absolute error (MAE) metrics. The findings showed that the hybrid models provided higher accuracy compared to the single-track models. Specifically, the proposed GRU–LSTM dual-path hybrid model produced the highest R² and lowest error values for both pollutant parameters in both the Dilovası and Ereğli regions. In Dilovası, this model achieved R² = 0.97 for SO₂ and R² = 0.96 for PM₁₀; in Ereğli, it reached R² = 0.92 for SO₂ and R² = 0.98 for PM₁₀. Thus, it has been shown that the GRU–LSTM dual-path hybrid model, which models short-term and long-term temporal dependencies in parallel, is an effective and reliable method for air pollutant forecasting in industrial areas. These findings demonstrate the potential of the proposed model to support air quality monitoring, early warning systems, and environmental decision-making in industrial regions. Full article

Early fault detection in microsatellite thermal control systems (TCS) faces fundamental challenges: high-dimensional redundant telemetry channels, overlapping multi-scale periodicities that obscure anomaly signatures, and severely limited daily data downlink (1–2 passes per day) that restricts the temporal window for diagnosis. Existing data-driven approaches either rely on supervised learning, requiring labeled fault data that are scarce in practice, or employ univariate analysis that fails to capture inter-sensor spatial correlations. To address these limitations, this paper introduces a hybrid framework integrating information-theoretic feature selection and spatiotemporal deep learning. The Generalized Maximum Information Coefficient (GMIC) quantifies nonlinear dependencies between temperature channels for key channel selection, reducing dimensionality by 82% while preserving diagnostic information. A dual-level Seasonal Trend Decomposition (STL) method disentangles orbital-periodic dynamics from diurnal cycles, effectively isolating distinct thermal characteristics at multiple timescales. Each decomposed component is modeled using Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) networks to capture spatiotemporal dependencies for accurate temperature prediction. An adaptive threshold-based weighted error fusion mechanism enables early fault detection within a single day of telemetry data. Experimental validation on real satellite telemetry data demonstrates that the proposed framework achieves high-precision fault detection across multiple fault types using a minimal set of temperature channels, significantly outperforming existing benchmarks in both prediction accuracy and detection reliability. Full article

Photoacoustic spectroscopy (PAS) is a highly sensitive and non-destructive technique widely used for trace gas detection; however, the simultaneous quantification of methane (CH₄), ethane (C₂H₆), and ethylene (C₂H₄) remains challenging due to severe spectral cross-interference and non-linear responses across broad concentration ranges. In this work, we propose a high-precision, end-to-end detection framework based on a Deep Kernel Extreme Learning Machine (DKELM) optimized using a Mutation–Chaotic Particle Swarm Optimization (MCPSO) algorithm. To enhance diagnostic information in the photoacoustic signals, a multi-scale wavelet transform based on a db4 wavelet basis with 5-layer decomposition and a Heursure soft threshold strategy is first employed for denoising and enhancing absorption features. To address the hyperparameter sensitivity and local-optimum trapping inherent in deep models, the MCPSO algorithm integrates hybrid chaotic initialization, adaptive mutation probability control, Cauchy-based perturbation, temperature-controlled mutation amplitude, and elite-guided population updating. The proposed MCPSO-DKELM model is evaluated on an expanded dataset of 470 mixed-gas spectra and benchmarked against other frameworks, including the previously reported SVM-CPSO-KELM architecture. The experimental results demonstrate that MCPSO-DKELM achieves stable, segmentation-free quantification across the full dynamic range, with an average detection error below 3.5% and the maximum relative error constrained to under 15%, which represents a substantial improvement over existing approaches. Thus, the combination of deep kernel feature extraction and mutation–chaotic global optimization provides a robust and reliable solution for simultaneous multi-component hydrocarbon gas analysis in complex industrial environments. Full article

Saltwater intrusion (SI) threatens groundwater sustainability in nearshore regions, particularly in Bangladesh, where over-extraction and sea-level rise accelerate aquifer salinization. Accurate prediction of SI dynamics is therefore critical for effective groundwater management. This study developed and evaluated several deep learning and hybrid models, including CNN, DNN, LSTM, CNN–DNN, CNN–LSTM, DNN–LSTM, and CNN–DNN–LSTM, to predict SI in a nearshore aquifer system. Predictor–response datasets were generated using the three-dimensional density-dependent flow and solute transport model FEMWATER. This study presents the first comprehensive benchmarking of standalone and hybrid CNN–DNN–LSTM models for SI prediction in a Bangladesh nearshore aquifer, supported by CRITIC–EDAS-based model ranking. Model performance was assessed using RMSE, MAE, MAD, R, IOA, a-20, NRMSE, along with CRITIC weighting and EDAS ranking. Results indicate that hybrid models integrating LSTM outperformed standalone models. The CNN–LSTM model achieved the best performance at OW1 (RMSE = 1.57 mg/L, MAE = 1.26 mg/L, R = 0.99, IOA = 0.99). The DNN–LSTM model performed best at OW2 (RMSE = 2.87 mg/L, IOA = 0.98, R = 0.97) and OW3 (RMSE = 1.95 mg/L, IOA = 0.99, R = 0.99). In contrast, the DNN model showed poor performance, while the CNN model demonstrated moderate performance and the LSTM model underperformed. Overall, the hybrid CNN–LSTM and DNN–LSTM models demonstrated superior accuracy and robustness for reliable SI prediction and sustainable groundwater management. Full article

The study examines the evolution, knowledge structure, and trends in snowmelt runoff prediction models. It identifies research hotspots, future directions, and offers a theoretical basis for accurate simulation and prediction. Utilizing CiteSpace software, 556 core Chinese and English publications from 2010 to 2025 were visually analyzed. Research on snowmelt runoff simulation shows: (1) Chinese publications are prominent in core journals like “Journal of Glaciology and Geocryology,” while English publications appear in high-impact journals like “Water Resources Research.” (2) Institutions like the University of Chinese Academy of Sciences, the Northwest Institute of Eco-Environment and Resources, and the University of California have formed a cross-regional research network. (3) International collaboration involves 42 countries, with a focus on China, the United States, and India. However, domestic institutional cooperation needs improvement. (4) Research trends in snowmelt runoff simulation have progressed from empirical statistics to remote sensing and model-driven physical mechanisms, and now to the integration of artificial intelligence with physical models. (5) The Chinese literature focuses on cold regions, while the English literature emphasizes intelligent modeling. This shift indicates a move towards “physical–intelligent” hybrid modeling. Future research should address challenges like model applicability in data-scarce areas, improving interpretability of complex models, quantifying uncertainties, and developing physically constrained deep learning models. Collaboration among institutions is crucial for enhancing water resource management and disaster warning systems in cold regions. Full article

Traffic accidents significantly impact societies and economies. The risk of collision is highest in urban areas, leading to devastating loss of life and escalating socioeconomic costs. In this context, numerous studies have focused on accurately predicting accident risk, severity, and duration using various methodologies. This paper presents an overview of traditional statistical models for accident prediction and a comprehensive systematic review of the literature on statistical modeling, machine learning (ML), and deep learning (DL) techniques employed in this field. Different methodologies and techniques are compared by categorizing studies that adopt similar approaches and analyzing them comparatively. Furthermore, a distinction is made between temporal and spatiotemporal models to describe how these approaches influence the accuracy of future predictions regarding accident occurrence and the duration of impact. This review distinguishes itself from similar works by not only comparing models and approaches, but also by analyzing how external features, such as meteorological data, road geometric design, and land usage, affect the probability of accidents and the models’ accuracy in forecasting road safety. The study explores the performance levels and limitations associated with a set of forecasting approaches, offering an analytical discussion of their differences and similarities, and potential future developments in this research space, including the use of hybrid models and reinforcement learning (RL). The results of this review indicate that DL models tend to be better suited to complex forecasting problems due to their superior ability to represent features and extract non-linear spatiotemporal correlations. This article concludes by describing various directions for further research, ranging from optimizing model architectures to integrating real-time big data into proactive prediction systems. Full article

Non-Intrusive Load Monitoring (NILM) enables estimation of the energy use of individual appliances in smart buildings from a single aggregate meter. In practice, however, this task is not straightforward. Signals from different appliances can overlap, and the measured data may also include distortions such as spikes, drops, and noise. To address these issues, this study presents a multi-task triple-hybrid deep learning framework that handles appliance classification and anomaly detection together. The model brings together 1D-CNN, BiLSTM, and Transformer Attention so that local patterns, temporal dependencies, and wider contextual information can be learned within the same structure. It also uses a dual-output design to classify appliance categories and detect anomaly types simultaneously. Experiments were carried out on Building 1 of the UK-DALE dataset with four appliances: kettle, microwave, washer dryer, and fridge freezer. For the anomaly task, synthetic disturbances were added to segmented signal windows and grouped as normal, spike, drop, and noise. To check how well the proposed framework handled different scenarios, it was tested on both the UK-DALE and REDD datasets. Looking at the main UK-DALE results, the model correctly identified appliances 99.48% of the time and spotted anomalies with 98.80% accuracy. A secondary test on the REDD dataset yielded an 86.44% classification score. This proves the architecture can adjust to completely new power grid environments without losing its edge. On top of that, when pitted against standard benchmark models like Seq2Point, this triple-hybrid design clearly does a better job of mapping out complex signal changes. As a result, it yields much stronger anomaly detection metrics. Full article

Indoor air quality has a significant impact on occupant health, comfort, and productivity in residential and commercial indoor environments. This paper proposes an IoT-edge enabled deep–fuzzy hybrid framework for real-time IAQ prediction and adaptive control. The proposed system integrates IoT-based environmental sensing, Temporal Fusion Transformer-based multivariate forecasting, knowledge distillation, edge-deployed Bi-LSTM inference, and Mamdani fuzzy logic control within a unified IAQ management architecture. A composite Comfort Risk Index is introduced to combine environmental parameters and occupant discomfort feedback into a single adaptive control indicator. Experimental evaluation under varying indoor conditions demonstrated strong forecasting performance, with prediction accuracies reaching 96.3% for CO₂ and 95.7% for PM_2.5 prediction, while reducing inference latency from 575 ms to 295 ms. Comparative analysis against baseline threshold-based control strategies further indicated improved comfort stability, smoother actuator behavior, and reduced estimated actuator operating intensity during deployment. The proposed framework also demonstrated resilient operation under simulated sensor-failure conditions while maintaining low computational overhead suitable for resource-constrained IoT-edge environments. Overall, the results indicate that combining lightweight deep learning models with interpretable fuzzy control can provide an effective, scalable, and energy-aware solution for intelligent real-time IAQ optimization in smart indoor environments. Full article

Background/Objectives: The accuracy and reliability of automated clinical analyzers are fundamental to patient safety and effective medical decision-making. Traditional quality control (QC) methods, which rely on periodic manual calibration and reactive maintenance, are inherently limited by high latency and labor costs; furthermore, they fail to provide continuous, real-time monitoring. This paper introduces a novel hybrid ensemble learning framework for the automated quality inspection of medical devices through the analysis of time-series reaction curves. Methods: Our system integrates three heterogeneous anomaly detection paradigms: an Enhanced Dynamic Time Warping (DTW) detector for robust non-linear pattern matching, a Shape Template Matching (STM) detector that mimics expert clinical logic by analyzing morphological features in a normalized shape space, and a specialized Time-series Variational Autoencoder (TimeVAE) for deep representation learning. The outputs of these detectors are fused using a weighted ensemble strategy, which is specifically designed to prioritize the minimization of false negatives—a critical requirement in medical diagnostics. Results: We evaluate our framework on a comprehensive, multi-center real-world dataset comprising seven distinct biochemical assays. Experimental results demonstrate that our proposed method achieves superior performance, attaining a $0\%$ false negative rate on CRE and DBIL assays and outperforming all baseline methods on the other five datasets. An ablation study confirms the model’s robustness even with limited training data, and a comparative analysis against eight state-of-the-art baseline methods further validates the effectiveness of our domain-optimized ensemble approach. Conclusions: The system provides a robust, interpretable, and highly automated solution for transitioning from reactive maintenance to proactive, real-time quality assurance in clinical laboratories. Full article

The high prevalence of cardiovascular diseases and the extensive application wearable electrocardiogram (ECG) devices for long-term monitoring have posed significant challenges for the transmission, storage, and real-time processing of massive amounts of ECG data. Consequently, efficient ECG compression and reconstruction have become a research priority in remote ECG monitoring. Traditional compressed sensing is complex and has high computational overhead, while single deep learning models cannot simultaneously extract local waveforms and model temporal dependencies. To address these shortcomings in the reconstruction process, this paper presents a CNN-LSTM-Attention hybrid model. This model utilizes a convolutional neural network (CNN) to capture local ECG waveform features, employs a long short-term memory (LSTM) network to learn long-term temporal dependencies, and introduces an attention mechanism to weight and fuse key diagnostic features, enabling accurate focus on key components including the QRS complex and ST segment. Experimental results on the MIT-BIH Arrhythmia dataset demonstrate that across the full compression range of 0.1–0.9, the proposed model achieves favorable comprehensive performance. Its PRD is stabilized at 10–12%, the SNR stays above 20 dB, and the RMSE is mostly lower than 0.25 mV. In terms of reconstruction accuracy and stability, our model outperforms the single CNN and CNN-LSTM models by a large margin. Full article

Accurate short-term forecasting for energy consumption is crucial in modern energy network management, especially for cities such as Tetouan, where considerable climate variability and diverse usage patterns present significant challenges when it comes to making short-term forecasts. This paper proposes a hybrid residual learning framework that combines a long short-term memory (LSTM) network with eXtreme Gradient Boosting (XGBoost) to improve short-term load forecasting for the Tetouan electricity network. The novelty of the proposed approach lies in coupling temporal sequence modeling with residual error correction driven by exogenous meteorological and calendar-related information. The proposed model is validated using real electricity consumption data from Zone 2 of Tetouan City, with further validation across all three available zones confirming the model’s generalizability. The proposed model achieves a coefficient of determination (R²) of 0.984, an RMSE of 687.21 kWh, and a MAPE of 2.41%, representing a 121.3 kWh RMSE improvement over the standalone LSTM baseline. These results confirm that the hybrid model is better at tracking periods of high demand compared to conventional machine learning approaches and standalone deep learning models. Full article