关键词: 软件可靠性预测, 相关向量机, Mann-Kendall方法, Sen的非参数方法

Abstract: Due to the common knowledge in software testing that early failure behavior of the testing process may have less impact on later failure process, the Relevance Vector Machine (RVM) learning scheme was applied to model the failure time data to capture the most current feature hidden inside the software failure behavior. Then the development of Average Relative Prediction Error (AE) series was studied while the value of m changed, so that it can be determined whether recent failure history could contribute to a more accurate prediction of near future failure event or not. Non-parametric statistical methods were applied toward detecting and estimating the trends in the data sets of AE value. Finally, Sen's slope estimator was applied to estimate the trend degree in the data sets so that suitable values of m can be got.

Key words: software reliability prediction, Relevance Vector Machine (RVM), Mann-Kendall method, Sen’s slope estimator