A study on software reliability prediction based on support vector machines

Support vector machines (SVMs) have been successfully used in many domains, while their application in software reliability prediction is still quite rare. A few SVM-based software reliability prediction models have been proposed in the literature; however, the accuracy of prediction can still be improved. In this paper, we propose an SVM-based model for software reliability prediction and we study issues that affect the prediction accuracy. These issues include: 1. Whether all historical failure data should be used; 2. What type of failure data is more appropriate to use in terms of prediction accuracy. We also compare the prediction accuracy of software reliability prediction models based on SVM and artificial neural network (ANN). Experimental results show that our proposed SVM-based software reliability prediction model could achieve a higher prediction accuracy compared with ANN-based and existing SVM-based models.