基于隐马尔可夫链模型的软件维护性评估方法研究

Software Maintainability Modeling and Assessment Method Research Based on Hidden Markov Chain Model

摘要:

针对当前软件可维护性评估主观性强,可操作性弱等问题,提出了定量描述维护性的维护时间统计概率描述方法,引入隐马尔可夫链(HMC)模型对维护性状态变迁过程进行描述,以可度量的维护性内部属性影响因素集量化值为观测序列,以维护时间统计概率为状态序列,构造了反映可维护性状态转移的HMC模型；收集配置管理库中软件模块历史维护时间从而确定完成维护任务频率来估计软件维护性初始状态,利用复杂网络特性计算软件维护性影响因素集的量化值,理论上即可评估出当前软件所处的维护性状态,最后运用实例对模型进行了训练与评估；结果表明,利用模型评估出的概率与实际维护任务统计出的可维护性概率基本一致,说明该方法可行且可重复,具有一定实践意义和研究前景。

Abstract:

In order to meet the demand of software maintainability assessment, three-state probability description method was put forward. Hidden Markov Chain model was introduced to estimate maintainability of three-state software, and state transform model was built up. Historical maintenance time of software module was collected from configuration management database to ascertain frequency of success maintenance so as to estimate software maintainability. Affecting factors set value was computed through Fuzzy inference theory. The estimation model was trained and validated through real software case, and the result shows that this method is feasible and repeatable, and can be further studied.