Analysis of Improved Evolutionary Algorithms Using Students’ Datasets

Abstract

Evolutionary Algorithms (EAs) are powerful heuristic search approaches which relies on Darwinian evolution that capture global solutions to complex optimization problems which has powerful features of reliability and versatility. (EAs) such as Particle swarm optimization (PSO) is a global optimization method that is extremely effective. PSO's flaws include slow convergence, premature convergence, and getting stuck at local optima. In this paper, chaotic map and dynamic-weight Particle Swarm Optimization (CHDPSOA) are combined with PSO to enhance the search strategy through adjusting the inertia weight of PSO and changing the position update formula in the (CHDPSOA), resulting in efficient balancing for local and global PSO feature selection processes. The performance of CHDPSOA was compared to that of three metaheuristic techniques: Differential Evolution (DE) and the original PSO, using eight numerical functions. The validation of this technique is carried out on four different datasets. The results show that the CHDPSOA is a good feature selection technique that balances the exploration and exploitation search processes to produce good results. The proposed CHDPSOA method performed well in correctly categorizing features using the KNN Classifier for all four datasets.