Comparative Analysis of Feature Selection Methods Using Swarm-Based Optimization Techniques

Feature selection is a vital data preprocessing technique aimed at selecting informative features from a dataset, which is essential for improving classification performance. Optimization involves deriving an optimal solution from feasible options, often drawing inspiration from nature. Particle swarm optimization (PSO) is highlighted as an effective metaheuristic algorithm for feature selection, capable of achieving satisfactory results with reduced computational complexity. Techniques are categorized into evolutionary, swarm intelligence, physics-based, and human-based approaches, with swarm intelligence particularly notable for its emergent intelligent behavior from simple agents. Swarm intelligence gains traction due to its speed, adaptability, and self-organization. These optimization techniques are highly efficient for tasks like optimal feature selection. In datasets of high dimension, quantity of noisy, irrelevant, and redundant features degrade clustering performance, necessitating methods to maximize accuracy, accelerate output, and filter out irrelevant features.