Introduction/Objective: Fixed feature sets and static coupling assumptions in existing short-term multi-load forecasting fail to capture seasonal variations in load interactions and meteorological sensitivity within integrated energy systems. This paper proposes an adaptive hybrid framework to improve point-forecasting accuracy for electric, cooling, and heating loads. Methods: Autocorrelation and Kendall rank analyses quantify seasonal periodicity, inter-load coupling, and load-weather relationships; XGBoost then selects load- and season-specific features. These features feed into a BiLSTM network, whose hyperparameters are optimized by an enhanced Newton-Raphson-Based Optimizer (NRBO) with a quasi-opposition-based initialization, following a multi-input single-output scheme. Results: On hourly data from Arizona State University, the model achieves MAPE values of 0.87%-1.70% (average 1.14%) across twelve season-load tasks, reducing MAPE by 54.5%- 68.1% (average 64.8%) compared with BiLSTM-XGBoost. Diebold-Mariano and Wilcoxon tests yield p < 0.05 against ARIMA, BiLSTM, SVR, and BiLSTM-XGBoost. Discussion: The statistically significant error reductions indicate that the proposed framework is robust and effective under seasonal fluctuations. The improvement is mainly attributed to the coordinated use of season- and load-specific feature construction, XGBoost-based feature screening, BiLSTM temporal modeling, and improved NRBO initialization, which together capture differentiated internal and external load characteristics. These findings support more refined energy management in integrated energy systems. Conclusion: The proposed framework, integrating XGBoost-based selection, BiLSTM modeling, and NRBO optimization, substantially enhances multi-load forecasting performance, offering a practical solution for refined energy management in integrated energy systems.
Introduction: The rapid adoption of the Internet of Medical Things (IoMT) has significantly enhanced healthcare services through continuous patient monitoring and intelligent clinical decision-making. However, the increasing interconnectivity and resource constraints of IoMT devices expose healthcare infrastructure to a wide range of cyberattacks, making the development of accurate, computationally efficient intrusion detection systems a critical challenge. Materials and Methods: This study proposes an integrated CSA-DNN intrusion detection framework that combines Clonal Selection Algorithm (CSA)-based affinity-driven feature selection with a lightweight Deep Neural Network (DNN) classifier. results: Experimental results demonstrate that the proposed CSA-DNN model achieves consistently high detection performance on both datasets. The model attains an accuracy of up to 99.6% and an F1-score of 98.4%, while maintaining low inference latency suitable for real-time IoMT environments. Compared to baseline models, including Random Forest, conventional DNN, CNN, and PSO-based approaches, CSA-DNN shows superior performance in terms of detection accuracy, recall, and computational efficiency. Statistical validation using paired t-tests confirms that the observed improvements are statistically significant. Results: The CSA identifies compact and discriminative feature subsets by reducing feature redundancy while preserving relevant intrusion information, whereas the shallow DNN provides efficient classification with low computational complexity, making the proposed framework suitable for resource-constrained IoMT environments. Discussion: The framework was evaluated on two recent public benchmark datasets, CICIoMT2024 and WUSTL-EHMS-2020, and compared with conventional machine learning methods, deep learning models, representative published approaches, and alternative metaheuristic feature selection algorithms. Comprehensive validation was further conducted through ablation studies, statistical significance analysis, attack-wise evaluation, and computational efficiency assessment. Experimental results demonstrate that the proposed CSA-DNN framework achieved an accuracy of 99.32% and an F1-score of 98.60% on CIC-IoMT2024, and an accuracy of 99.61% with an F1-score of 98.47% on WUSTL-EHMS-2020. Moreover, the proposed framework consistently outperformed the compared methods while maintaining low inference latency and reduced computational overhead. Conclusion: These results demonstrate that the proposed CSA-DNN framework provides an effective balance between detection performance, feature compactness, and computational efficiency, making it a promising solution for real-time intrusion detection in resource-constrained IoMT healthcare systems.
Introduction: The increasing complexity and dynamic behavior of next-generation wireless networks motivate network-management mechanisms capable of predictive and adaptive operation. This study develops and evaluates an AI-driven digital twin framework that integrates traffic forecasting, anomaly detection, network-state synchronization, and adaptive resource management within a closed-loop architecture. Methods: An LSTM model was used for traffic forecasting using the Abilene backbone traffic dataset, while a CNN was trained on synthetic intrusion-like traffic for anomaly detection. The outputs of the forecasting and anomaly detection modules were integrated with a simulationbased resource management layer and compared with an equal-allocation baseline. Results: For the evaluated forecasting scenarios, the LSTM achieved an R² of 0.984 and a MAPE of 1.92% for the 5-min horizon. Across the evaluated anomaly scenarios, CNN F1 scores ranged from 0.917 to 0.939, and ROC-AUC values ranged from 0.945 to 0.971. Relative to the equal-allocation baseline, the integrated framework reduced the reported latency metric by 34.4%, increased reliability by 16.5%, increased throughput by 25.3%, reduced packet loss by 42.6%, and reduced energy consumption per transmitted data unit by 26.6%. Discussion: The results indicate that jointly incorporating predictive and diagnostic information can improve resource-management performance under the evaluated benchmark and simulation conditions. However, the experiments use backbone IP traffic and synthetic anomaly data; therefore, they do not constitute validation in an operational 5G/6G radio access network. Conclusion: The proposed framework provides proof of concept for closed-loop AI-assisted digital twin management, while evaluation on representative cellular network datasets and operational testbeds remains necessary.
Artificial Intelligence (AI) is a computational technique that enables machines to perform tasks associated with human intelligence, such as decision-making. AI is a crucial and effective tool for addressing the challenges of climate change. AI techniques overcome numerous challenges associated with traditional methods of addressing climate change, such as efficiently managing large datasets and reducing computational complexity. Various AI- and machinelearning- driven methods are used to address climate-related problems, including predictive modeling, image recognition, and geospatial analysis. Various algorithms, such as Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Random Forests, and Generative Adversarial Networks (GANs), and their applications in climate research are discussed. This study explores the factors contributing to climate change and analyzes research gaps in existing studies. This research analyzes the impacts of climate change, including sea level rise, extreme weather events, cyclones, and reduced crop production. The rapid growth in climate-related data from 2010 to 2024 is also discussed in this study. Additionally, this study examines which countries provide more funding in this integrated field of research at both the global and Indian levels. This study also discusses the growth of publications at the intersection of AI, climate change, and Green AI.
Background: The proposed system can be used to detect underwater objects, which is essential for autonomous underwater exploration, biodiversity monitoring, and the fulfilment of the United Nations Sustainable Development Goal (SDG) 14. Even though deep learning models trained on terrestrial imagery work well, severe visual degradation stemming from light scattering, colour absorption, turbidity, and changing illumination in underwater environments fundamentally limits their applicability. Methods: This structured literature review compares the CNN-based models (YOLOv8, YOLONAS, SSD, RetinaNet, Faster R-CNN, Mask R-CNN), transformer-based models (Swin Transformer, DETR, DINO) and hybrid CNN-Transformer architectures, and evaluates their results in a unified four-dimensional framework: detection accuracy (mAP), environmental robustness under compound underwater disturbances, computational efficiency (FPS, power consumption), and edge deployability on AUV hardware. Results: The comparative analysis shows that various architectural paradigms have various advantages and disadvantages in the underwater environment. Transformer-based architectures tend to achieve better detection performance and robustness in visually degraded environments, thanks to stronger contextual modelling. The YOLO-based architectures have shown a good trade-off between detection accuracy, computational efficiency, and deployment suitability, making them attractive for real-time AUV applications. Two-stage detectors are well suited for precision-oriented inspection and segmentation applications, while hybrid CNN-Transformer networks offer greater robustness in turbid, visually complex underwater environments. Discussion: Model selection must be environment- and mission-specific. Four research gaps are identified: architecture design, data engineering, deployment optimisation, and robustness evaluation, each with specific unsolved problems, including the absence of cross-paradigm benchmarks, compounded dataset biases, and a lack of compound-condition robustness testing. Conclusion: This structured literature review provides condition-specific model selection guidelines aligned with SDG 14 and identifies future research priorities including selfsupervised learning, sensor fusion, and lightweight AUV-optimised architectures.
