An Improved Discretization-Based Feature Selection via Particle Swarm Optimization

Feature selection (FS) aims to remove the redundant or irrelevant features of the data, which plays a very important role in data mining and machine learning tasks. Recent studies focus on integrating the data discretization technique into the FS process to help achieve superior performances in classification. In this paper, we proposed an improved discretization-based FS method via particle swarm optimization to obtain a higher classification accuracy with a smaller size of feature subset. In our approach, we use a novel encoding and decoding way for particle swarm optimization (PSO) which can efficiently select multiple cut-points for discretization. In addition, a new updating strategy and a local search procedure is proposed to strengthen the searching ability and avoid being trapped into local optimal. Experimental results on benchmark datasets demonstrate the efficacy of our proposed methods both in the classification accuracy and the feature subset size.