A Proposed Framework for Optimum Feature Selection using Improved Chicken Swarm Optimization Algorithm for Face Recognition system

Abstract

Feature selection is a significant assignment in data mining and pattern detection as it lessens the largeness of the data sets and at the same time preserves the classification exploit. Standard Chicken Swarm Optimization (CSO) has been universally employed for feature selection considering its efficacy. The standard CSO algorithm, however, experiences the challenges of falling at local optima and high computational cost due mainly to the large search space. The study proposed a framework to increase the accuracy of a face recognition system. The Improved Chicken Swarm Optimization technique will be formulated from standard CSO and chaotic map by introducing chaotic gauss map and chaotic tent map equations into the rooster and hens update equations of CSO respectively and will be employed for feature selection. Local binary pattern (LBP) will be used for feature extraction. Finally, the classification of individual images based on input images will be recognized using a Support Vector Machine (SVM) classifier. The evaluation will be done by comparing the combination of the ICSO-LBP technique with the combination of CSO-LBP technique based on recognition accuracy and will serve as our performance metrics. Based on the proposed evaluation, this study believed that the ICSO-LBP technique would have a high recognition accuracy than the CSO-LBP and would also avoid being trapped at the local optimum and improve the convergence speed of the algorithm.