In recent years, wireless sensor networks (WSNs) have been growing rapidly because of their ability to sense data, communicate wirelessly, and compute data efficiently. These networks contain small and low‐powered sensor nodes that organize and configure themselves to carry out their functions. Even though WSNs are cheap, easy to deploy, flexible, and efficient, there are some challenges in terms of energy efficiency and network lifetime. Clustering in WSNs is the most reliable solution for the challenges, in which nodes are grouped into few clusters, and a cluster head (CH) is selected for data aggregation and data transfer to the base station (BS). However, there are still many challenges such as energy hole and isolated node problems that exist because of inefficient CH selection and cluster formation methods. In this work, we comprehensively reviewed various nonmetaheuristic and metaheuristic methods for CH selection and cluster formation that are used in networks from various environmental settings, for a better understanding of how the aforementioned problems are tackled by some authors. Moreover, the methods’ parameter settings, advantages, limitations, and future directions are presented with a brief performance summary of the approaches.

A good exploration ability can ensure that the method jumps out of local optimum in multimodal problems and a good exploitation can ensure an algorithm converge faster to global optimum values. So, this paper proposes a new hybrid sperm swarm optimization and genetic algorithm to obtain global optimal solutions termed HSSOGA which is developed based on the concept of balancing the exploration and exploitation capability by merging Sperm Swarm Optimization (SSO), which has a fast convergence rate, and a Genetic Algorithm (GA) that can explore a search domain efficiently. To ensure that the proposed method delivers good performance, it is evaluated with 11 standard test function problems consisting of 5 unimodal and 6 multimodal functions. The proposed HSSOGA set is compared with conventional GA and SSO methods, as well as with several hybrid methods such as Hybrid Firefly and Particle Swarm Optimization (HFPSO), hybrid Simulated Annealing and Genetic Algorithm (SAGA), Hybrid Particle Swarm Optimization and Genetic Algorithm (HFPSO), hybrid Particle Swarm Optimization and Grey Wolf Optimization (PSOGWO), and closely related Hybrid Sperm Swarm Optimization and Gravitational Search Algorithm (HSSOGSA). The results are evaluated in terms of each method’s best fitness, mean, standard deviation, and convergence rates. The numerical experiment results show that HSSOGA has better convergence towards the true global optimum values as compared to the conventional and existing hybrid methods in most unimodal and multimodal test function problems.

Stock movement prediction remains challenging due to the complexity of integrating heterogeneous data sources and capturing intricate temporal-semantic relationships. Existing approaches suffer from incomplete semantic preservation, shallow fusion mechanisms, and inadequate multi-scale temporal modeling. In this paper, we propose a multimodal deep fusion framework that addresses these limitations through three synergistic components: 1) An SRTE module based on pretrained BERT that preserves 100% of textual content by segmenting long sequences and applying average pooling, recovering 17.82% more tokens than conventional truncation; 2) An MsLSTM architecture that simultaneously captures short-term price dynamics and long-term market trends through parallel processing of original and temporally smoothed sequences, and; 3) A CMCF mechanism that performs dual-level integration via across-temporal interval fusion and along-temporal step aggregation. We compare our proposed method against baseline models on the StockNet dataset, achieving superior predictive performance. Additionally, we conduct comprehensive ablation studies, back-testing analysis, long-text modeling comparisons, and interpretability analysis. The results validate the effectiveness of deep multimodal fusion for financial forecasting and provide actionable insights for data-driven decision-making in volatile market environments.

Wireless sensor networks (WSNs) support environmental monitoring, industrial control, healthcare, and innovative city services, where early discovery of abnormal behavior is vital for reliability and safety. Practical deployments still face a core problem in which strong detection often demands memory and energy that sensor nodes cannot spare, while light models struggle when data patterns shift. This review examines the current landscape and compares five method families in terms of accuracy, computational and energy cost, adaptability, and deployability on constrained hardware. The evidence shows that traditional methods are efficient but brittle under change, machine learning (ML) is lean but feature dependent, deep learning (DL) attains high accuracy but is heavy in memory and latency, bio-inspired search aids tuning but can be unstable, and mixed strategies narrow the gap at the cost of added engineering. Findings indicate that lightweight DL models refined with bio-inspired optimization achieve a practical balance between accuracy and resource limits. This review offers a concise taxonomy with a mixed-model workflow to guide implementation. Future work should focus on energy-aware optimization of LightYolo and sperm swarm optimization (SSO), using shared benchmarks that report accuracy, latency, memory, and energy consumption, bridging theory, and practice toward scalable and efficient anomaly detection in next-generation WSNs.