Software Defect Prediction: An Ensemble Learning Approach

Abstract

Software defect prediction plays an increasingly critical role in emerging software systems. However, existing software defect prediction approaches typically suffer from low accuracy due to the under/over fitting problems. To address this problem, we propose an ensemble learning approach to achieve the accurate defect prediction, where various machine learning algorithms, i.e., artificial neural network, random forest, k-nearest neighbour methods are integrated together. The proposed software defect prediction workflow is introduced. Experiments are conducted to verify the effectiveness of the proposed method. Extensive experiment results verify that our proposed method can improve the defect prediction accuracy when compared with existing methods.