Abstract: Classification problem on imbalanced data is a key issue in the machine learning field，obtaining data is unbalanced in many real applications, such as the defect prediction for software modules.The classification methods based on support vector machine for imbalanced data is one of the effective classification approaches, many researchers focus on these methods. Duc to the software modules defect metric datasets have the characteristics, such as class imbalance and noise, the prediction models based on the normal support vector machine (SVM) can’t get satisfactory results.Therefore, in this paper, we make a relatively in-depth study on support vector machine for predicting software module defects. Based on the previously proposed fuzzy support vector machine for imbalanced
data classification(FSVM CIL)，integrating sampling technology, in this paper we introduce two improved algorithms; Onc is FSVM CIL RUS, which combines FSVM CIL algorithm with random under sampling algorithm. Before building software module defect prediction models usingFSVM CIL, we balance the datasets
using random under sampling. And the other is an ensemble algorithm called FSVM丛aI. RBBag. This algorithm combines the FSVM CII. algorithm with roughly balanced bagging algorithm. Using FSVM CIL algorithm to build base classifiers, and then we ensemble the base classifiers to improve the prediction performance. Appling two algorithms to two NASA software module datasets CMl and KC3，and experimental results show the effectiveness of the newly proposed algorithms, and the combination between FSVM CIL. and the sampling technology or ensemble technology can improve the performance of the module.