Genetic Algorithm Selection Strategies based Rough Set for Attribute Reduction

摘要

Attribute reduction is considered a vital topic for studies that consider actual data intricacy. The attribute reduction problem aims to find a minimum attribute set from a large set of attributes while avoiding information loss. The problem is denoted as an NP-hard, which is the non-deterministic polynomial time optimization problem. Researchers have widely used many heuristic and meta-heuristic approaches to optimize this problem in rough set theory. Numerous studies have utilized meta-heuristic methods to address the attribute reduction problems, prompting this research to suggest an improved one-population meta-heuristic method. This paper presents the implementation of the genetic algorithm on an attribute reduction-based rough set utilizing different selection strategies: roulette wheel, tournament and rank-based selections. An experiment was performed on 13 datasets from the public domain available in the UCI repository. The results demonstrated that the tournament selection strategy performed better than the roulette wheel and rank-based selection strategies and other published meta-heuristic algorithms.