摘要： 利用递归最小二乘支持向量机(RLSSVM)构造软件可靠性失效模型，通过失效数据集对模型进行反复训练，提高模型学习能力。模型依据递归计算方法，可动态反映软件可靠性的变化，对软件失效有准确的预测能力。使用模拟退火(SA)算法对RLSSVM的参数进行寻优，得到改进的RLSSVM，实现对模型结构的优化。与常用的非齐次泊松过程模型相比，利用RLSSVM与SA算法构造的可靠性模型具有更好的拟合和预测能力。

关键词: 软件失效模型, 递归最小二乘支持向量机, 拟退火算法

Abstract: Recurrent Least Squares Support Vector Machines(RLSSVM) is proposed to construct the failure model of software reliability and train it recurrently by using failure data set to improve its learning ability. The recurrent algorithm is used in model to reflect dynamically the change of software reliability so the model can predict accurately the software failure. Simulated Annealing(SA) algorithm is used to optimize the parameters of RLSSVM to construct more optimal failure model. After comparing with the models of NHPP category, the model based on RLSSVM and SA algorithm has better ability of fitting and prediction.

Key words: software failure model, Recurrent Least Squares Support Vector Machines(RLSSVM), Simulated Annealing(SA) algorithm

中图分类号: 

• TP311