Applying Weighted Particle Swarm Optimization to Imbalanced Data in Software Defect Prediction

Abstract

Imbalanced data typically refers to class distribution skews and underrepresented data, which affect the performance of learning algorithms. Such data are well-known in real-life situations, such as behavior analysis, cancer malignancy grading, industrial systems’ monitoring and software defect prediction. In this paper, we present a W-PSO method, which comprises weighting of instances in a dataset and the Particle Swarm Optimization algorithm. The presented method was combined with classification methods C4.5 and Naive Bayes, respectively, and tested experimentally on ten freely accessible software defect prediction datasets. Based on the results achieved, the presented W-PSO method creates better classification models than classification methods C4.5 and Naive Bayes in the majority of the cases.