DYNAMIC SOFTWARE RELIABILITY PREDICTION: AN APPROACH BASED ON SUPPORT VECTOR MACHINES

A support vector machine (SVM) modeling approach for software reliability prediction is proposed. Based on the structural risk minimization principle, the learning scheme of SVM is focused on minimizing an upper bound of the generalization error that eventually results in better generalization performance. The SVM learning scheme is applied to the failure time data, forcing the network to learn and recognize the inherent internal temporal property of software failure sequence. Further, the SVM learning process is iteratively and dynamically updated after every occurrence of new failure time data in order to capture the most current feature hidden inside the software failure behavior. The performance of our proposed approach has been tested using four real-time control and flight dynamic application data sets and compared with feed-forward neural network and recurrent neural network modeling approaches. Experimental results show that our proposed approach adapts well across different software projects, and has a better next-step prediction performance.