Feature selection with centre of gravity method using ant colony optimization

Abstract

The high dimensional dataset with irrelevant, redundant and noisy features has much influence on the performance of machine learning problems. In this work, an existing Ant Colony Optimization (ACO) based feature selection algorithm is modified by attaching a dimensionality reduction method as a data pre-processing step. This is achieved by introducing the concept of Centre of Gravity (CoG) of a set of points. After reducing the dimension, the ACO algorithm is used to generate the optimal subset of features. The performance of the proposed algorithm is evaluated using Artificial Neural Network (ANN) classifier. The performance comparison using various dataset shows that the proposed method outperforms the existing ACO based feature selection methods.