Swarm Intelligent Data Clustering Performance Improvement by Social Spiders’ Optimization

Researchers have focused on the collective intelligence of animals, birds, and insects over the last decade. Swarm intelligence is an artificial intelligence field that implements intelligent systems by studying the joint behaviour of social bugs and other animal cultures. Meta-heuristic procedures using swarm behaviour and complicated interactions without supervision have been used to address challenging optimization issues. Data cluster puts related data into groups. In addition, it may create discontinuous clusters. The term “disjoint” refers to a condition in which two or more elements or sets the crucial metrics in data clustering are accuracy and efficiency. Recent research has elucidated the characteristics of bio-inspired systems, highlighting their capacity as information-processing systems with specific cognitive capabilities. Nevertheless, widely used bio-inspired algorithms for data clustering still need to adequately consider the crucial trade-off between exploration and exploitation to achieve more optimal clustering outcomes. This article presents a bio-inspired procedure called social spider optimization (SSO) for gathering. The program aims to achieve a desirable equilibrium between exploration and exploitation by leveraging the distinct roles of female and male spiders. In this study, we conduct a proportional scrutiny of the outcomes found by the proposed procedure SSO about those achieved by K means, as well as other nature-inspired algorithms like colony optimization (IBCO) particle swarm optimization (PSO), ant colony optimization (ACO). The robustness of the method is seen to provide improved clustering outcomes. While Single-Source Optimization (SSO) effectively addresses the issue of being trapped in a local optimum, it necessitates adaptation to effectively identify the optimal solution within the vicinity of the created global solution. Therefore, SSO with K-means is used, yielding favorable outcomes in localized search operations. Our findings indicate that the suggested hybrid algorithms consistently outperform the alternative approaches regarding gathering outcomes. The presentation and efficiency of the suggested clustering approaches are measured and validated using the (SICD), regular cosine resemblance, precision, and inter-cluster reserve.