A Defect Level Assessment Method Based on Weighted Probability Ensemble

Lixia Xie, Siyu Liu, Hongyu Yang, Liang Zhang

Abstract: In order to solve the problems that the existing defect prediction methods lack the assessment of the potential defect level of samples and do not fully consider the cost impact of misclassification, a defect level assessment method based on weighted probability ensemble (DLA-WPE) is proposed. Firstly, the greedy selection method is used to select features. Then, according to the number of samples in different categories, the unequal punishment of different misclassification is calculated to obtain the misclassification punishment (MP). The weighted probability ensemble (WPE) model is built. Finally, the voting weight of each base classifier is calculated according to the MP. According to the dichotomous probability of base classifiers, the defective quantification value is calculated to obtain the defect assessment results, and the potential defects of modules are assessed. The experimental results show that the defective quantitative values and defect levels are consistent with the actual situation of the samples.