Risk prediction models for imported infectious diseases: A systematic review.

A state-of-the-art review of scientific, industrial, and commercial developments in water leakage management for water supply systems.

MARINE-Transformer: A General-purpose framework for multivariate ocean time series analysis.

Maximum mean discrepancy enhanced Informer for accurate cross-domain dual-timescale effluent prediction in wastewater treatment plants.

Physiological data-driven models for motion sickness prediction.

The growing use of renewable energy sources (RES), particularly wind power (WP), in deregulated electricity markets has resulted in large operating uncertainties because of the erratic and uncertain behavior of wind resources. Differences between actual wind velocity (AWV) and forecasted wind velocity (FWV) cause a large price imbalance (PIMB) that has implications for market execution scheduling, operational reliability, and economic profitability. To solve these problems, this paper presents a comprehensive techno-economic model comprising Long Short-Term Memory (LSTM)-based wind forecasting coupled with an American Zebra Optimization (AZO) algorithm for profit maximization in wind-integrated deregulated power systems. The proposed approach is validated on the IEEE 14-bus test system by collecting real wind datasets from Mina Sultan Qaboos, Muscat, Oman, and Duqm, Oman. This work shows that the LSTM model works in modelling nonlinear temporal dependencies in wind speed data, leading to a significant improvement in forecasting accuracy relative to traditional forecasting techniques and Random Forest (RF)-based approaches. Using Locational Marginal Pricing (LMP), imbalance settlement mechanisms, generator operational limits, and transmission constraints, the forecasted outputs are integrated into an Optimal Power Flow (OPF)-based market framework. The proposed LSTM-based framework has achieved a significant reduction in imbalance pricing by 30-35% in Muscat and 28-29% in Duqm when compared to traditional forecasting techniques. Moreover, the AZO algorithm outperforms Sequential Quadratic Programming (SQP) in convergence performance and overall system profitability by approximately 6%. Profit improvement of the obtained result of wind potential is 5.9% for Muscat and 6.4% for Duqm. The proposed framework successfully reduces uncertainty in the market and enhances renewables integration capacity, as well as offers a robust decision-support framework for the electricity market operation and a decision support tool under the RES-driven high-level market conditions to ensure economic competition for competitive electricity market operations.

In growing 6G-enabled Internet of Vehicles (IoV) environments, in-car networks (IVNs) are more susceptible to sophisticated cyber threats, especially zero-day attacks that avoid signature-based detection. The centralised data dependency, lack of geographical awareness, and poor generalisation in heterogeneous and non-IID conditions are the limitations of current intrusion detection systems. This paper suggests a cooperative intrusion detection system that combines a spatio-temporal graph convolutional long short-term memory (ST-GConvLSTM) model with hierarchical personalised federated learning. The suggested method uses dynamic vehicle-to-vehicle graph structures to capture both intra-vehicle temporal patterns of CAN communications and inter-vehicle spatial dependencies. Scalable training is made possible by a three-tier learning architecture (vehicle, edge, and cloud) that protects data privacy and reduces concept drift. Furthermore, under limitations of latency, energy consumption, communication overhead, and privacy budget, a multi-objective reinforcement learning technique is used to dynamically optimise client involvement. In comparison to state-of-the-art federated baselines, experimental evaluations on real-world CAN datasets under realistic 6G-IoV settings show that the proposed framework achieves high detection performance for both known and zero-day attacks while significantly improving convergence speed and lowering system overhead. These findings demonstrate how well hierarchical federated optimisation and spatiotemporal graph learning can be combined to create safe and scalable vehicular networks.

Cryptocurrency markets have gained significant global attention due to their decentralized nature and high financial value. Among various cryptocurrencies, Bitcoin is the most widely traded and exhibits highly volatile price behavior. Accurate analysis and prediction of Bitcoin price trends are challenging because the market is influenced by rapid trading activities, large data streams, and complex temporal patterns. This paper presents a streaming data collection and analysis system for Bitcoin using the Long Short-Term Memory (LSTM) deep learning algorithm. The proposed system continuously collects real-time Bitcoin market data from online cryptocurrency exchanges through streaming APIs. The collected data is then preprocessed and analyzed using an LSTM-based predictive model capable of learning long-term dependencies in time-series data. The LSTM network processes sequential historical price data to forecast future market trends and provide analytical insights into Bitcoin price movements. The system integrates data acquisition, preprocessing, deep learning-based prediction, and visualization modules to create an efficient cryptocurrency analysis framework. The proposed approach focuses on improving prediction accuracy by combining real-time streaming data with advanced neural network models. This system can assist researchers, financial analysts, and investors in understanding cryptocurrency market behavior and making informed trading decisions. The proposed design demonstrates the feasibility of integrating streaming data technologies with deep learning models for real-time financial market analysis. Keywords— Cryptocurrency, Bitcoin, Streaming Data, LSTM Algorithm, Deep Learning, Time-Series Prediction, Financial Data Analysis.

Identifying tumor-specific T-cell antigens is essential for advancing cancer immunotherapy and enabling precision-driven, AI-assisted discovery. While artificial intelligence (AI) and machine learning (ML) have significantly impacted healthcare and biotechnology, existing approaches often struggle with the inherent complexity and sequence dependency of antigen data, resulting in suboptimal predictive performance. In this study, we propose a Deep Neural Network (DNN)-based framework specifically designed to address these challenges in computational tumor T-cell antigen identification. The proposed framework employs hybrid sequence encoding techniques, including Position-Specific Scoring Matrix with Discrete Wavelet Transform (PsePSSM-DWT) and Protein Bidirectional Encoder Representations from Transformers (ProtBERT-BFD). To enhance efficiency, a Shapley Additive exPlanations (SHAP)-based global feature selection strategy is applied to select the most informative feature set before model training. The optimized feature set is subsequently used to train the DNN. Experimental evaluation demonstrates that the proposed model achieves an average accuracy of 96.16% with a Matthew's correlation coefficient of 0.923. These results significantly outperform conventional machine learning and state-of-the-art methods. The proposed framework not only establishes a robust computational baseline for antigen identification but also provides a foundation for potential integration with multi-omics data and real-time immunotherapy workflows.

ABSTRACT Assessing the robustness of epidemiological associations requires diverse methodological approaches. This study introduces the Repeated Acute Exposure (RAE) design, a survival-based framework utilising an extended Cox model with time-varying coefficients to evaluate the short-term association between PM10 and daily mortality in Valencia, Spain. By incorporating a time-varying coefficient to capture the dynamic intensity of the risk across a 7-day window, this approach conceptualises every exposure increment as contributory to the mortality risk. We adjusted for temperature, relative humidity, and day of the week, incorporating a Ridge penalty for the date to account for the structural dependency and redundancy of the lag-stratified data. The analysis yielded a Hazard Ratio of 1.025 (95% CI: 1.009–1.040) per 10 µg/m3 increase in PM10. While numerically analogous to the Relative Risk (RR), the HR specifically reflects the instantaneous risk profile. The estimated effect size was consistent in direction and significance of the association with the one obtained using a Generalised Additive Model (GAM)-based distributed lag non-linear model (RR: 1.035; 95% CI: 1.005–1.066), and other time-series methods. The RAE design offers a rigorous sensitivity tool for validating associations through methodological triangulation. The dataset and R script for reproducibility are available on Zenodo at https://doi.org/10.5281/zenodo.19115821

The rapid expansion of single-cell RNA sequencing (scRNA-seq) has made accurate cell type annotation a critical bottleneck for biological discovery. Existing computational methods are often limited by reference data dependency, while emerging single Large Language Model (LLM) approaches are susceptible to model-specific biases and provide insufficient uncertainty quantification. To address these limitations, we introduce mLLMCelltype, a framework that harnesses collective intelligence-the emergent problem-solving capacity arising when multiple independent agents interact through structured deliberation to produce solutions exceeding individual capabilities-of multiple LLMs through an iterative deliberation process. Across 49 diverse datasets, our framework achieves a mean accuracy of 77.2%, a 15.7-percentage-point improvement over the best-performing single-LLM baseline (61.5%). The consensus mechanism demonstrates high robustness to noisy input and generalizes to datasets released after the LLMs' training. By providing transparent reasoning chains and robust consensus-based confidence metrics, mLLMCelltype minimizes manual annotation effort and enables reliable interpretation of complex cellular landscapes. The framework is available as an open-source package and an accessible web server.

This study proposes a hybrid CNN-LSTM-XGBoost model that integrates shale oil seismic attributes with macroeconomic indicators to predict global oil prices. The model extracts spatial features from seismic volumes using 3D CNNs and captures temporal dependencies in economic data via LSTM, with fused features processed by XGBoost regression. It achieves an RMSE of 3.61 USD/barrel, R2 of 0.847, and MAPE of 6.98%, outperforming standalone models by 12.7-23.1%. SHAP analysis shows that seismic attributes contribute an average marginal impact of 0.19 on the model output. The model uses seismic attributes as input features to forecast oil price trends, providing a transparent and interpretable framework for integrated geophysical-economic analysis. At the local scale, the model assists shale oil operators in optimizing drilling and production decisions by linking seismic-derived reservoir quality to medium-term supply expectations. At the broader global scale, the framework demonstrates how geological information can be systematically incorporated into energy market forecasting, offering a new paradigm for geoeconomic modeling under supply‑side uncertainties.

Accurate weather prediction is crucial in agriculture, disaster prevention, and public safety. Challenge: Traditional numerical models have high computational costs and struggle with atmospheric nonlinearity and chaos, while existing deep learning methods face limitations in handling spatial heterogeneity and non-Euclidean data. Solution: This paper introduces the STGLDWeather method. It combines multi-scale spatiotemporal graph neural networks (MS-ST-GNN) and latent diffusion models (LDM) to capture multi-scale spatiotemporal dependencies in weather data and model the temporal evolution of weather conditions in latent space. Conclusion: Experiments on real weather datasets show that STGLDWeather significantly outperforms existing state-of-the-art baselines in prediction accuracy and computational efficiency, particularly excelling in temperature, geopotential height, and wind speed forecasts.

Emotion recognition from EEG signals has been one of the most promising areas due to its potential in enhancing human-computer interaction, especially in adaptive systems. This paper proposes a novel emotion recognition system that improves classification accuracy through advanced signal processing, adaptive channel selection, and deep learning techniques. The system starts with the multi-scale wavelet transform to break down EEG signals effectively followed by Kalman filtering with Wavelet denoising that enhances the quality of signal. Emotional peaks are established using Spectral Entropy analysis for accurate epoch determination. For adaptive channel selection, the utilization of Reinforcement based Deep Q-Networks to pick the most informative EEG channels dynamically for each emotional state. In feature extraction, the hybrid model based on Spatio-Temporal Attention Networks (ST-ANs) is adopted to capture spatial and temporal dependencies in the EEG data. Multi-Scale Feature Fusion is utilized to fuse short- and long-term dependencies. Final emotion classification is done through an Ensemble of Graph Neural Networks (GNNs) and Memory-Augmented Neural Networks (MANNs) providing robust adaptability across different subjects and emotional states. The benchmark dataset is collected from kaggle repository such as EEG brainwave, DEAP dataset and Computer game based EEG dataset. The proposed work achieves 98.5% accuracy on the EEG Brainwave Dataset. On the DEAP Dataset, it achieves 94.5% accuracy. For the EDA on Emotion Recognition (S01G1AllChannels), the model reaches 97.6% accuracy, outperforming the existing frameworks in emotion classification.

Power management is a cornerstone for many Cyber-Physical Systems (CPSs), which relies on low-power circuits, dynamic power management algorithms and energy-aware software to match their requirements in terms of energy. As CPSs evolve towards data-centric designs to more promptly accommodate AI models and integration, traditional power management techniques must also be improved. In this paper, we build on SmartData to introduce a data-centric Power Manager (PM) framework that allows CPSs to model energy in terms of data. SmartData defines a high-level interface for sensing, actuation, and control in data-centric CPSs. It abstracts the myriad of features of modern embedded platforms related to processing, scheduling, synchronization, and communication. These Energetic SmartData encapsulate the components of a CPS, which interact in a publish–subscribe fashion, declaring interest on other SmartData and responding to other SmartData interests. We introduce an algorithm to extract a Directed Acyclic Graph (DAG) from these Interest relationships, with vertices representing the involved components and edges representing the associated cost in terms of energy. We also introduce a Power Manager that uses such DAGs to monitor the state of the system, eventually overriding low-priority Interests to reach the specified lifetime. We evaluated the proposed framework through a case study with Ocean-Bottom Nodes (OBNs) under realistic, dynamic energy conditions. Results show that without any power management, the system fails 12 days before its target operational lifetime. The proposed data-driven PM was then benchmarked against a fixed-schedule Static PM and a reactive Threshold PM. Our approach was the only strategy to guarantee a 365-day lifetime in all scenarios. With an ideal initial battery capacity of 260 Ah, it achieved a high utility of 23.1%. It also proved its adaptability in an energy-deficit scenario with an initial capacity of 257 Ah, where it reduced utility to 2.8% to survive, a condition in which the other strategies failed. • Data-driven Cyber-Physical-Systems (CPS) design allows for Power-State Estimation from the data dependencies, which dictate the operation mode for CPS components. • Realistic energy supply models impose relevant challenges on Power Management approaches. • Non-adaptive Power Management solutions tend to fail to meet requirements under complex energy supply models. • Machine-Learning-based solutions are often Component-based and fail to model the multi-component nature of modern data-driven CPS. • Data-driven Power Management approaches lead to a direct relationship between QoS and Power Configuration.

• Hybrid physics-AI approaches for modelling bio-based building envelopes are reviewed. • Challenges of coupled heat and moisture transfer in bio-based materials are discussed. • PINN-based methods are critically compared with UDE and graph-based formulations. • Model robustness, interpretability and data dependency are assessed. • Perspectives for advanced hygrothermal modelling in building applications are outlined. The hygrothermal behavior of bio-based building materials plays a central role in determining indoor comfort, energy performance, and durability. Conventional physical models, grounded in conservation laws, provide interpretability and robustness but often struggle with hysteresis effects, heterogeneity, and computational cost. Conversely, data-driven machine learning (ML) approaches offer flexibility and efficiency but lack physical consistency and interpretability. In response to these limitations, hybrid physics-AI models have recently emerged as transformative tools. This review critically examines three ML paradigms: Physics-Informed Neural Networks (PINNs), Physics-Informed Graph Neural Networks (PIGNNs), and Universal Differential Equations (UDEs), and evaluates their potential for simulating coupled heat and moisture transfer in porous bio-based envelopes. PINNs demonstrate high accuracy under sparse data conditions but remain limited by training cost and scalability. PIGNNs offer scalability and adaptability to irregular geometries, enabling large-scale or real-time simulations of building envelopes, but face challenges in representing hysteresis effect. UDEs provide balanced trade-off by embedding physics while correcting unmodeled nonlinearities such as sorption hysteresis and multiscale porosity. By offering a critical state-of-the-art analysis of recent advances, this review identifies current limitations, experimental requirements, and future directions for the deployment of hybrid AI–physics approaches in hygrothermal analysis. It concludes by positioning these methods as a roadmap for next-generation digital twins of bio-based materials, supporting predictive design, performance monitoring, and sustainable building practices.

The smart city is a shifting promise defined by current objectives. This means that smart cities redefine themselves as issues such as connectedness, sustainability, or equality become politically, commercially, and or socially more or less appealing. Certain seemingly invisible factors can impact the extent to which the promise of smart governance and infrastructure come to fruition. This paper examines Amsterdam's internal “smart” soft infrastructure, from politics, to policies, to practice. This overview, stemming from interviews and reports, uncovers how policies, funding, and government structure determine what the smart city is and the extent to which it can be inclusive and accessible for all. The Amsterdam case reveals that similar smart cities have the potential to suffer from short term projects, ethical issues unextractable from data dependency, and conflicting needs of diverse city-dwellers. These concerns make it such that independent self-funded forces with long-term visions may wield comparative power in smart urban development. • Qualitative analysis of Amsterdam reveals soft infrastructure roadblocks. • Political shifts drive Amsterdam to reject the specific smart city label. • Short-term pilot projects hinder sustainable, long-term urban inclusion. • Diverse citizen needs create conflicts in smart infrastructure design.

Cross-validation strategies under data dependency: An example with anemia prediction in sheep using ocular conjunctiva images.

Oceans exhibit complex dynamics influenced by climate change, anthropogenic activities, and natural phenomena. Understanding these dynamics is critical for ensuring the sustainability of marine environments and their optimal utilization. This research aims to study and monitor upwelling phenomena in the South Sea of Java. Upwelling, the exchange of nutrient-rich, cold water from deeper layers to the surface, enhances marine biological productivity; Sea Surface Temperature (SST) serves as a key indicator for its detection. To achieve these objectives, this study employs both ConvLSTM and 3D-CNN. ConvLSTM, a deep learning architecture that integrates convolutional structures within LSTM units, effectively captures spatiotemporal dependencies in sequential data. 3D-CNN, a deep learning model extending traditional 2D convolutional neural networks, processes volumetric data, enabling the extraction of spatial features across three dimensions. Analysis reveals that ConvLSTM outperforms 3D-CNN in modeling upwelling data in the South Sea of Java. This is evidenced by lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). The ConvLSTM method was then used for forecasting, and the results were validated with data obtained from local fishermen regarding their fishing expeditions. Visual analysis confirms that the ConvLSTM method accurately models upwelling data in the South Sea of Java with fishermen's schedules. ConvLSTM and 3D-CNN methods were comparatively evaluated for modeling Sea Surface Temperature (SST) data, considering wind speed, sea surface salinity, and the El Niño-Southern Oscillation (ENSO) phase as influential factors. Based on Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values, the ConvLSTM method exhibited lower values, indicating superior performance compared to the 3D-CNN approach. Specifically, RMSE and MAE values for ConvLSTM were 0.4161 and 0.3017, respectively, while for 3D-CNN, the corresponding values were 0.6095 and 0.4259. Upwelling data forecasting results were validated against local fishermen's schedules, with data collected in July 2022. Visual inspection confirmed alignment between the forecasted upwelling patterns and the fishermen's activity.

Respiratory diseases remain a challenge in Brazil due to socioeconomic inequalities and environmental risks that intensify population vulnerability. This study compared XGBoost with a deep learning model using stacked Gated Recurrent Units (GRU), trained with morbidity data from respiratory diseases and exogenous variables such as per capita GDP, population density, urbanisation index and greenhouse gas emissions (1999-2023). These data were normalised and temporally disaggregated using synthetic data to refine time-series granularity. Results showed regional heterogeneity: the GRU achieved superior performance in Porto Alegre (R2 = 0.529), São Paulo (R2 = 0.518) and São Luís (R2 = 0.313), while XGBoost showed mostly negative R2 values. These findings demonstrate the potential of recurrent neural networks to capture temporal dependencies in health data and support morbidity forecasting. By anticipating fluctuations, such models can guide resource allocation and inform evidence-based policies. The study underscores the value of integrating socioeconomic and environmental indicators into predictive frameworks and positions deep learning as a promising tool for precision public health in unequal contexts.