Research on Software Reliability Based on Forecasting Model Combined by LS-SVM and Markov Chain

Abstract

To improve the accuracy of prediction on software failure data, one combine forecasting model is proposed based on least square support vector machine (LS-SVM) and Markov chain. First, LS-SVM optimized by simulated annealing algorithm (SA) is adopted to establish the time series forecasting model on software failure data. Then, in order to improve the prediction accuracy, the prediction interval is narrowed by means of Markov chain. Finally, after applying the combination forecasting model to the predicting of one commercial software failure data, the results indicate that the model has a certain precision and reliability.