基于信息熵的软件可靠性多模型动态预测方法

Multi-model dynamic prediction method of software reliability based on information entropy

针对经典软件可靠性预测模型中的不一致问题和"变点"问题,研究了一种基于信息熵的软件可靠性多模型动态预测方法,采用固定步长滑动窗口的动态赋权策略,运用信息熵对多个经典模型预测性能进行动态融合,实现了软件可靠性的综合预测。试验结果表明,在一定步长情况下,基于信息熵的软件可靠性多模型动态预测方法比单个经典模型有更好的预测性能。

Multi-model dynamic prediction method of software reliability based on information entropy was proposed to solve the inconsistency problem and ＂change points＂ of classical prediction models. Based on information entropy theory, using dynamic weighted strategy by fixed step sliding window, the combined prediction method of software reliability was obtained which enhanced the prediction results by fushing the prediction performances of classical models dynamically. Test results show that the multi-model dynamic prediction method of software reliability based on information entropy is better than classical ones under the right step condition.