Abstract: Gene expression programming (GEP), which is a new evolutionary algorithm based on genetic algorithm and genetic programming, has been acknowledged as a powerful machine learning technique and widely used in the field of data mining. Thus, this paper applies GEP into the non-parametric software reliability modeling due to its unique and pretty characters. This new GEP-based modeling approach considers some important characters of reli-ability modeling in several main components of GEP, i.e. function set, terminal criteria, fitness function, and then obtains the final model (named GEP-NPSRM) by training on the failure data. Finally, for several real failure data-sets, four case studies are proposed by respectively comparing GEP-NPSRM with several representative existing software reliability models in term of the fitting and prediction powers. The results show that the GEP-NPSRM pro-vides better fitting and prediction performances.

Key words: software reliability modeling, gene expression programming (GEP), non-parametric software reliability model, machine learning

摘要： 基因表达式编程是一种基于遗传算法和遗传编程的新型机器学习技术, 其具有更为优秀的数据挖掘能力, 已被成功应用于函数发现领域。提出一种基于基因表达式编程的非参软件可靠性建模方法, 该方法将基因表达式编程算法中的若干关键步骤(如初始种群函数集、适应度函数、终止条件等)与软件可靠性建模的若干重要特征相融合, 在失效数据集上进行训练, 从而获得基于基因表达式编程算法的非参软件可靠性模型。在若干组真实失效数据集上, 将所提出的模型与若干典型的基于人工神经网络以及遗传编程的非参软件可靠性模型进行对比实例研究。实例结果表明, 基因表达式编程算法的非参软件可靠性模型具有更为显著的模型拟合与预计性能。

关键词: 软件可靠性建模, 基因表达式编程(GEP), 非参数化软件可靠性模型, 机器学习