Rank based binary particle swarm optimisation for feature selection in classification

Abstract

Feature selection (FS) is an important and challenging task in machine learning. FS can be defined as the process of finding the best informative subset of features in order to avoid the curse of dimensionality and maximise the classification accuracy. In this work, we propose a FS algorithm based on binary particle swarm optimisation (PSO) and $k$-NN classifier. PSO is a well-known swarm intelligent algorithm that have shown to be very effective in dealing with various difficult problems. Nevertheless, the performance of PSO is highly effected by the inertia weight parameter which controls the balance between exploration and exploitation. To address this issue, we use an adaptive mechanism to adaptively change the value of the inertia weight parameter based on the search status. The proposed PSO has been tested on 12 well-known datasets from UCI repository. The results show that the proposed PSO outperformed the other methods in terms of the number of features and classification accuracy.