Swarm Intelligence Based Algorithm for Feature Selection in High-Dimensional Datasets

Abstract

In this era of ever-expanding data complexity, feature selection plays a paramount role in-to reducinge the issues associated with high dimensionality. The purpose of feature selec-tion is to discover a minimal optimal subset of features from the set of all available fea-tures with high accuracy. For datasets with numerous features, it is ubiquitous to determine the smallest possible subset of features to elevate performance accuracy. Prudently select-ed feature subset provides outstanding results, as compared to complete set of features, thereby making it convenient for the analysts to explore patterns and relationships among data. This research examines a temperature variable’s inclusion to the existing Ant Colony Optimization algorithm that demonstrates a significant increase in classification perfor-mance. The objective of this research is to evaluate the performance of hybrid Ant Colony Optimization- Simulated Annealing (ACO-SA) algorithm with other algorithms on several benchmark datasets using a Decision tree classifier.