Search Results (683)

This paper proposes a novel forecasting model termed U-Client, which integrates parallel cross-temporal and cross-variable attention branches with an adaptive gated fusion mechanism for ultra-short-term photovoltaic (PV) power forecasting. First, meteorological features are screened using the Pearson correlation coefficient to reduce input dimensionality. Second, parallel cross-temporal and cross-variable attention branches are designed to extract long-range temporal trends and nonlinear interaction features among meteorological variables, respectively. Third, a gating mechanism is introduced to adaptively fuse the two types of features based on input conditions. Finally, a linear module is combined to generate the final forecasting results. Experiments based on measured datasets from a photovoltaic station in Ningxia, China, demonstrate that the proposed U-Client model outperforms classical models such as Long Short-Term Memory (LSTM) and Informer across all evaluation metrics for 1–4 step forecasting tasks. Ablation studies and statistical significance tests further verify the effectiveness of each component. The proposed model provides reliable support for ultra-short-term power dispatching in new-type power systems. Full article

Objectives: Secondary progressive multiple sclerosis (SPMS) is a natural transition from relapsing-remitting multiple sclerosis (RRMS) in many cases. However, whether and how these phenotypes differ on an individual basis is not fully understood, limiting timely diagnosis and management for SPMS. This study aimed to investigate how deep learning using 3-dimensional (3D) frameworks including VGG19, ResNet152, and DenseNet-121 helped differentiate SPMS from RRMS based on routine clinical datasets, and what brain areas mostly contributed to this differentiation using model explanation techniques. Methods: We examined 140 participants (70 each for RRMS and SPMS) as part of an ongoing study comprising prospectively collected clinical and imaging data from routine healthcare. The data was curated to improve consistency and completeness using different strategies and were then randomly split by subject into training (n = 120) and held-out testing (n = 20). The former was used for model development through five-fold cross validation. Deep learning used T1-weighted, T2-weighted, and FLAIR brain MRI, with optional clinical variables (n = 6). A 3D gradient-weighted class activation mapping (Grad-CAM) technique was applied to identify brain areas of significance followed by ablation studies for additional insight. Results: Among the 3D frameworks validated, VGG19 was deemed the best. Based on MRI and the best 3D VGG19 model, different data curation strategies showed largely similar results. Additionally, the models combining clinical variables with MRI achieved equivalent or slightly greater performance than MRI-only models, with an average testing area under the receiver operating characteristic curve of 0.84 when datasets were fused at the flatten layer, best at 0.92, versus 0.82 and 0.89. Model explanation indicated brain regions of significance in distinguishing SPMS from RRMS individuals, including bilateral frontal lobes, left occipital and temporal lobes, and cerebellum. Conclusions: Overall findings suggest the potential of 3D deep learning models such as VGG19 for distinguishing SPMS from RRMS using routine brain MRI and clinical data, which, along with 3D Grad-CAM, could facilitate discovery of new biomarkers underlying disease worsening. Full article

Accurate monitoring of hydrological dynamics in complex perennial landscapes is a cornerstone for tropical agricultural sustainability. Traditional energy balance models based on orbital optical data often face methodological bottlenecks due to cloud cover and the “greening myth,” where optical indices fail to capture immediate water stress due to the non-linear decoupling between stomatal closure and pigment loss. This study developed a cloud-integrated multisensor framework to estimate actual evapotranspiration (ET_a) at a refined 100 m resolution in mountain coffee systems, utilizing active microwave proxies from Sentinel-1. We fused polarimetric metrics—Degree of Polarization (DoP) and Shannon Entropy (SE)—with land surface temperature and soil moisture data. Multiple Linear Regression (MLR) was compared against non-linear algorithms (Random Forest and SVR) to prioritize model parsimony and physical interpretability. The results show that MLR emerged as the most parsimonious and suitable model within this localized dataset scope (R² = 0.872; RMSE = 2.916 mm/8-day), outperforming complex “black-box” architectures. Soil moisture emerged as the dominant environmental driver of ET_a variability, while SAR-based metrics served as sensitive mechanical proxies for canopy geometric heterogeneity and macro-structural variations. Cross-correlation analysis revealed a 16-day lag, empirically indicating that biophysical water shifts temporally precede geometric canopy alterations. Operationally, this framework ensures temporal continuity under persistent cloud cover and provides high-fidelity spatial detailing for precision water management. This approach offers an auditable and scalable tool for watershed planning and climate resilience in tropical agriculture. Full article

Accurate wind-power forecasting is essential for grid scheduling when renewable generation becomes highly variable. This study developed WindPower-SAFusion, an Informer-inspired forecasting model designed for long wind-power sequences. The framework is built around three complementary designs. First, ProbSparse self-attention is used to lower the attention cost from $O\left(L^2\right)$ to $O\left(L\log L\right)$ while retaining informative temporal dependencies. Second, convolutional distillation is embedded in the encoder to summarize local fluctuations and form multi-scale representations. Third, historical theoretical power and wind speed are fused in a recursive forecasting scheme for multi-step prediction. The model is evaluated using measured data from the Daliang Wind Farm in Guazhou, Gansu Province, China. Experiments conducted using 1-day, 3-day, and 7-day horizons show that WindPower-SAFusion obtained lower errors and higher explanatory ability than the selected statistical and deep learning baselines. The ablation results further confirm the contributions of sparse attention, convolutional feed-forward extraction, and sequence distillation. These findings indicate that the proposed framework can provide an effective data-driven tool for wind-farm dispatching and power-management applications. Full article

Predicting landslide displacement is important for geological-hazard early warning. In reservoir areas, displacement evolution is affected by rainfall, reservoir water level, vegetation variation, and the intrinsic non-stationarity of the displacement sequence, which makes accurate prediction difficult for conventional single-sequence models. To address this problem, this study proposes a residual-increment-oriented landslide displacement prediction framework that fuses multi-source monitoring variables. The displacement sequence is first processed into trend and periodic-related fluctuation representations, and the residual increment is used as the prediction target. Rainfall, reservoir water level, and the normalized difference vegetation index (NDVI) are incorporated as external monitoring variables. A cross-branch attention mechanism models interactions among heterogeneous feature branches, and a sparse MoE-based fusion module is introduced to adaptively adjust branch contributions under different deformation conditions. The model predicts the displacement residual increment, from which the final displacement is reconstructed. A case study using the Baishui River (Baishuihe) landslide monitoring dataset was conducted, together with additional validation on the related Bazimen Z110 landslide monitoring dataset and comparisons against conventional recurrent, convolutional, statistical, and Transformer-based baselines. The results show that the proposed model achieves lower RMSE and MAE than the compared methods on the tested datasets. These findings suggest that residual-increment modeling, multi-source monitoring variables, and condition-dependent branch fusion can improve short-term displacement prediction for the tested reservoir-area landslide cases. Full article

Fusing spatial observations from sensors with heterogeneous error characteristics is a persistent challenge in geostatistics. Classical kriging assumes a Gaussian likelihood for all observations, an assumption that fails when sensors exhibit one-sided or asymmetric noise. We present a Variable Rank Kriging (VRK) formulation that supports per-observation heterogeneous likelihoods where each observation may define its own likelihood function, thus enabling principled fusion of sensors whose noise structures are significantly different in terms of distribution family and magnitude. Within this framework, we use the exponential (one-sided) likelihood as a case study to demonstrate the method and compare it with sampling-based numerical alternatives for general likelihoods without closed forms. A non-collocated RTK calibration workflow uses kriging predictions from a sparse high-accuracy reference to characterise sensor-specific likelihood parameters without requiring co-located paired observations. Synthetic 1-D and 2-D experiments show that correct per-point likelihood specification reduces RMSE by up to 92% (1-D) and 57% (2-D) relative to a misspecified Gaussian model while also eliminating systematic positive bias. A demonstration using NEON Airborne Observation Platform lidar data at Harvard Forest confirms these findings in a practical, real-world scenario. Across multiple subsamples of the lidar dataset, the exponential likelihood reduces vegetated-zone RMSE by $20.6\%$ (open zone: $18.6\%$) and mean absolute bias by $26.5\%$ relative to a heteroscedastic Gaussian baseline. The open-source vrk Python ($>=$3.10) package provides a reproducible implementation that can be applied to any spatial domain that requires multi-sensor spatial fusion with heterogeneous error structures. Full article

Background/Objectives: Survival prediction in non-small cell lung cancer (NSCLC) remains challenging, particularly in limited-sample settings where end-to-end deep learning models may suffer from limited generalization. This study aimed to develop a data-efficient, multimodal, and explainable framework integrating computed tomography (CT)-derived imaging information with clinical variables for NSCLC survival prediction. Methods: CT images, tumor segmentations, and clinical data from the publicly available NSCLC Radiomics (LUNG1) dataset (377 patients) were used. Tumor-focused regions were extracted using segmentation masks, and pretrained RadImageNet-InceptionV3 embeddings were obtained from the largest tumor-containing slice and neighboring-slice summaries. Deep imaging embeddings, engineered imaging features, and clinical variables were fused into a unified tabular representation. To improve robustness under limited-sample conditions, feature blocks were compressed using principal component analysis. CatBoost, XGBoost, and LightGBM models were trained on a development set and evaluated on a strictly held-out final validation set. Results: In three-class survival stratification, assigning censored/non-event patients to the upper survival group produced the strongest ordinal prognostic performance. Under the EX_PLUS_NON_EX_TOP setting, CatBoost achieved the best holdout score-based class C-index of 0.655. In continuous survival regression, LightGBM achieved the best holdout event-patient C-index of 0.576. Clinical variables provided the dominant prognostic signal, while compact deep image embeddings contributed complementary information, particularly in separating short- and long-survival groups. SHAP analysis confirmed contributions from both clinical and image-derived features. Conclusions: The proposed framework provides a proof-of-concept demonstration of a data-efficient and explainable image-to-tabular approach for NSCLC survival prediction under strict internal holdout validation. The results suggest that pretrained CT embeddings, clinical variables, gradient-boosted trees, and SHAP-based interpretation can be combined in a feasible, limited-sample survival modeling pipeline, while external validation remains necessary before clinical translation. Full article

To address the insufficient accuracy of short-term load forecasting caused by the strong randomness of distributed energy output, variable electricity consumption patterns, and complex meteorological factors, this study proposes a load forecasting method that integrates K-means scene clustering and a Transformer–BiGRU–Attention (CTBA) hybrid deep learning architecture. Different from conventional Transformer–BiGRU hybrid forecasters that train a single global predictor across all operating conditions, the proposed CTBA framework first partitions daily load curves into representative scenes and then routes each sample to a scene-specific Transformer–BiGRU–Attention predictor, thereby reducing distributional heterogeneity before temporal modeling. First, the K-means algorithm is used to perform scene clustering on historical daily load curves, and the optimal number of clusters is selected according to the silhouette coefficient and downstream prediction performance. Subsequently, the CTBA model is trained separately for each clustering subset. The Transformer encoder captures the long-range global dependencies of load sequences through the self-attention mechanism, the BiGRU module extracts local bidirectional temporal fluctuation features, and the Attention mechanism further focuses on key time nodes such as morning and evening peaks while fusing multi-source data including historical load, day-ahead electricity price, and multi-dimensional meteorological factors. Experimental results based on the German ENTSO-E power dataset show that the coefficient of determination R² of the proposed model reaches 0.9893, with MAE, RMSE, and MAPE as low as 0.0141, 0.0187, and 3.92%, respectively, which are significantly improved compared to benchmark models such as SVR, LSTM, CNN, and TCN-BiGRU. Ablation experiments further demonstrate that removing the clustering, Transformer, BiGRU, or attention layer will degrade performance, thus verifying the effectiveness and superiority of the method in short-term load forecasting and providing an accurate solution for the short-term load forecasting of power systems. Full article

Reasoning-intensive multimodal retrieval suffers from a counter-intuitive bottleneck: on MM-BRIGHT multimodal-to-text (Query+Image → Documents), the strongest dense multimodal encoder reaches only 27.6 nDCG@10 and the rest of the dense vision–language retrievers cluster between 10.0 and 23.0. The visual signal, encoded as a dense vector, adds noise rather than evidence; even augmenting strong text retrievers with raw image captions degrades performance by up to 12.0 points. We propose VISA, a Visual Symbolic Agent that re-casts multimodal-to-text as text retrieval over three parallel streams. A Vision LLM is dispatched in three roles via separate prompts: a zero-shot router that classifies the query image into up to three parser types from a fixed taxonomy of nine (chart, circuit, equation, screenshot, code, figure, diagram, map, photograph); typed parsers that extract structured text per type; and a holistic captioner. The agent constructs three text streams (raw query, query ⊕ symbolic, query ⊕ caption), scores each with a single frozen 4B-parameter retrieval LLM, and fuses the per-document scores via Reciprocal Rank Fusion or a confidence-weighted linear combination. The whole agent contains no trainable parameters. The key novelty is a change of substrate: rather than projecting the query image into a dense multimodal vector that competes with text, VISA is, to our knowledge, the first retrieval system to convert the image into typed symbolic text and keep retrieval entirely text-side, so that a frozen text retriever can match the literal tokens (axis values, variable names, function signatures) that answering documents actually contain. Across all 29 MM-BRIGHT multimodal-to-text domains, VISA achieves 32.4 nDCG@10, an absolute improvement of +4.8 over the strongest dense multimodal encoder and substantially larger margins over the remaining six dense vision–language baselines. Per-domain analysis shows VISA maintains its margin across STEM and software domains where image content is structure-heavy. In practical terms, VISA is training-free and model-agnostic: it requires no fine-tuning, reuses any off-the-shelf vision LLM and text retriever, caches all per-image parsing so re-runs cost only three query encodes, and can therefore be dropped into an existing text-retrieval stack to add reasoning-intensive multimodal capability without building or training a multimodal encoder. Full article

Soil organic carbon (SOC) is closely linked to soil fertility, agricultural carbon cycling, and the functioning of cotton field ecosystems, and it provides essential information for sustainable soil management. Rapid and accurate SOC estimation is therefore important for assessing carbon sequestration potential and supporting low-carbon agricultural management. This study focused on cotton fields in northern Shawan City and used optical imagery, Synthetic Aperture Radar (SAR) imagery, and 140 ground-collected SOC samples to estimate SOC content with three machine learning models: Random Forest (RF), Light Gradient Boosting Machine (LightGBM), and Extreme Gradient Boosting (XGBoost). The Kennard–Stone algorithm was applied to partition the 140 SOC samples into training and validation subsets at a 7:3 ratio, ensuring a more representative distribution of samples. Model performance was evaluated using the coefficient of determination (R²) and root mean square error (RMSE), and SHapley Additive exPlanations (SHAP) was used to interpret feature contributions and SOC spatial variability. The results showed that: (1) optical features performed better than SAR features, while fused optical-SAR features achieved the highest accuracy; (2) XGBoost consistently outperformed RF and LightGBM, with the optimal model achieving R² = 0.726 and RMSE = 1.252% on the validation set; (3) SHAP analysis confirmed the dominant contribution of optical features to SOC estimation; and (4) the predicted SOC distribution showed higher values in the central study area, lower values in the northern and southern parts, and high-value zones mainly along both sides of the Manas River. By comparing optical, SAR, and fused features for SOC estimation in arid-zone cotton fields, this study provides methodological support for rapid SOC monitoring and sustainable soil management, and offers practical guidance for variable-rate fertilization and soil carbon sequestration planning along the Manas River corridor. Full article

Accurate prediction of football pass outcomes is important for tactical analysis, decision evaluation, and skill-oriented feedback in student football training and physical education. However, event-level pass outcome prediction remains challenging because pass success is jointly influenced by spatial context, defensive pressure, receiver-related cues, and historical coordination between players. To address this issue, this study proposes an information-guided multilayer perceptron (IGMLP) based on grouped feature representation and gated feature fusion using structured event data. In the proposed framework, input variables are organized into interpretable semantic feature groups, including contextual features, pressure-aware features, historical coordination features, and receiver-related features. These groups are encoded through separate branches and adaptively fused by a group-level gating mechanism for nonlinear pass outcome modeling. Unlike conventional gated neural architectures that usually apply generic gates to hidden units, channels, or sequential states, the proposed gated design operates at the semantic feature-group level and adaptively weights football-specific information sources according to their relevance to each pass event. Using the StatsBomb open-event dataset, both prediction and recognition paths were constructed, and the proposed model was compared with standard multilayer perceptron (MLP), residual neural network (ResNet), boosting tree (BT), convolutional neural network (CNN), and long short-term memory network (LSTM). In the prediction path, IGMLP achieved an Accuracy of 0.9184, Precision of 0.9295, Recall of 0.9837, F1-score of 0.9558, and AUC of 0.9325. In the recognition path, IGMLP achieved an Accuracy of 0.9808, Precision of 0.9882, Recall of 0.9902, F1-score of 0.9893, and AUC of 0.9925. These results indicate that semantic feature grouping and gated feature fusion are effective for event-level football pass outcome prediction. Full article

Accurate fault classification of electro-hydrostatic actuators (EHAs) remains challenging because multivariate fault signals contain local transient variations, inter-variable coupling, and dynamic temporal dependencies that are difficult to capture simultaneously using a single model. To address this problem, this paper proposes a multi-view temporal feature collaborative heterogeneous ensemble learning model (MTF-HEM) for EHA multivariate time series fault classification. MTF-HEM integrates a representative subsequence-guided time series forest (RSG-TSF), XGBoost, and a lightweight LSTM to extract local morphological, global statistical, and temporal dependency features, respectively. The outputs of these heterogeneous base learners are fused using a bootstrap-driven out-of-bag probability binning stacking (BOPB-stacking) strategy. The proposed method was evaluated on an AMESim-based simulated EHA plunger pump fault dataset containing one normal condition and six fault conditions. Under the present simulation setting, MTF-HEM achieved an accuracy of 99.52% and outperformed the tested deep time series classification models, ensemble models, and individual base learners. These results suggest that multi-view heterogeneous feature fusion can improve the classification of simulated EHA fault time series and provide a methodological reference for intelligent actuator fault diagnosis. However, the current validation is based on data generated from a single AMESim simulation model, and further evaluation on real EHA systems is needed to assess the practical applicability and generalizability of the proposed approach. Full article

Accurate vessel traffic flow prediction provides an important data basis for intelligent shipping management, including maritime traffic monitoring, navigational risk awareness, waterway organization, and emission-related assessment. Although recent studies have advanced spatiotemporal, graph-based, and hybrid forecasting methods, improving the predictive ability of a conventional ConvLSTM backbone without introducing substantially more complex model structures remains underexplored in grid-based waterway scenarios. This study proposes a lightweight CA-ConvLSTM framework for grid-level vessel inflow and outflow prediction. AIS-derived flow data and MERRA-2 meteorological variables are rasterized onto a common spatial grid and fused at an early stage. A residual dilated convolution module with dilation rates of 1, 2, and 4 is used to extract multi-scale spatial dependencies, and a channel attention mechanism is applied before ConvLSTM-based temporal prediction to adaptively reweight the fused flow-meteorological feature channels. Experiments using AIS and MERRA-2 data from the northern Bohai Strait waterway show that the proposed framework improves baseline ConvLSTM performance. Compared with ConvLSTM, CA-ConvLSTM reduces MSE and MAE by 24.93% and 12.55% for outflow prediction, and by 24.80% and 12.82% for inflow prediction. These results suggest that spatially aligned meteorological fusion, multi-scale spatial feature extraction, and channel-wise feature weighting can effectively enhance ConvLSTM-based grid-level vessel traffic flow prediction without relying on complex model fusion or heavy graph-based architectures. Full article

Mobile RFID inventory in libraries must be planned and evaluated under noisy observations, configuration-dependent read regimes, and incomplete supervision. This paper presents an uncertainty-aware analytics framework for robot-assisted RFID inventory using the public RFID Location dataset. The framework has three phases. Phase 1 converts irregular list-encoded logs into atomic RFID events and quantifies how operating configuration changes read density and signal variability. Phase 2 performs map-constrained Bayesian shelf inference by synchronizing RFID reads with robot trajectory and antenna geometry and by fusing RSSI and carrier phase over feasible shelf candidates. Phase 3 translates posterior spread and non-convergence into proxy review workload and cost, enabling configuration comparison and certainty–throughput trade-off analysis when strict EPC-to-item linkage is unavailable. Across 688,073 aligned RFID observations, the pipeline produces 18,190 posterior tag estimates from five inventory runs. The empirical results show strong run dependence: the best run achieves a mean posterior spread of 0.906 m with a convergence rate of 0.553, whereas a degraded run reaches only 0.004 convergence with a mean spread above 2.1 m. Because EPC-to-item linkage is unavailable, these values are posterior concentration and workload indicators rather than ground-truthed localization-accuracy metrics. A saved phase-weight ablation further shows that adding phase information substantially sharpens posterior concentration relative to an RSSI-only baseline. Under the proxy workload model, autonomous-S1-P30 provides the most favorable balance among posterior certainty, scan effort, and implied review burden. Full article

Autoencoders (AEs) have been widely used for industrial process fault detection owing to their ability to learn nonlinear representations from normal operating data. However, conventional AE methods rely heavily on reconstruction errors and may miss weak faults due to overgeneralization. In addition, insufficient modeling of temporal evolution and operating condition variations may reduce their sensitivity to dynamic faults. To address these issues, this study proposes a memory-enhanced and prediction-assisted conditional variational autoencoder named MI-CVAE for unsupervised fault detection. In the proposed framework, statistical features extracted from sliding windows are used as condition information to describe variable operating states. A memory module stores representative normal prototypes to constrain reconstruction and reduce overgeneralization to faulty samples. Meanwhile, an Informer branch captures temporal dependencies and provides complementary prediction residuals. Reconstruction and prediction residuals are fused to construct squared prediction error and squared Mahalanobis distance statistics, with control limits determined by kernel density estimation. The proposed method is validated on the Benchmark Simulation Model No. 1 wastewater treatment benchmark and a real papermaking process dataset. The results show that MI-CVAE outperforms the evaluated comparison methods, particularly in detecting weak and dynamic faults, while maintaining a low false alarm rate. Full article