Hybrid Feature Selection

Applying evolutionary algorithms to feature selection issues in high-dimensional spaces has proven challenging due to the ”curse of dimensionality” and the high computing cost. Our project implements and then tries to extend the HFS-CC-PSP, a three-phase hybrid feature selection algorithm. The approach simultaneously tackles processing cost and dimensionality problems by combining correlation-guided clustering and particle swarm optimization (PSO). The HFS-CC-PSO algorithm combines three distinct feature selection techniques, each with benefits. The search space is condensed in the first and second phases using a filter and a clustering-based method. The ideal feature subset is located in the third phase using an evolutionary method with global searchability. The algorithm also incorporates a rapid correlation-guided feature selection approach, a symmetric uncertainty-based feature deletion method, and other features to enhance the performance of each phase.