基于决策树的软件可维护性预测方法研究  

Research on Software Maintainability Prediction Based on Decision Tree

作　　者：朱佳俊 王炜[1,2] 李彤 唐季[1] ZHU Jia-jun;WANG Wei;LI Tong;TANG Ji(School of Software,Yunnan University,Kunming 650091,China;Key Laboratory for Software Engineering of Yunnan Province,Kunming 650091,China)

机构地区：[1]云南大学软件学院,云南昆明650091 [2]云南省软件工程重点实验室,云南昆明650091

出　　处：《计算机技术与发展》2018年第8期66-70,共5页Computer Technology and Development

基　　金：国家自然科学基金(61462092,61379032);云南省自然科学基金重点项目(2015FA014)

摘　　要：软件维护是软件全生命周期中一项高难度、高成本、长周期的活动,准确预测软件可维护性对降低软件维护成本、提高软件可用性具有重要意义。软件可维护性分析历经20多年的研究,当前的预测分析性能和准确率仍然不高,甚至达不到模型预测是准确的标准;而总结相关研究发现,软件可维护性数据还普遍存在数据分布不均衡问题,该问题将直接影响到模型预测的性能。针对上述问题,基于采样方法利用决策树建立软件可维护性预测模型,并通过UIMS和QUES数据集对模型进行实验验证。结果表明,与基线方法和现有的可维护性预测方法相比,文中方法在UIMS数据集和QUES数据集的平均误差率(MMRE)分别提高了84%和61%,且Pred(0.25)都达到了该评价标准的最优值1,表明该方法具有更优的综合性能。Software maintenance has been one of the most difficult,costly and long-term tasks in the software development lifecycle. Accurate prediction of software maintainability can be useful to reduce the cost and improve the usability of software. The process of software maintainability research has gone through more than 20 years,but predicting performance and accuracy is still not high,and even fail to reach the standards that the model prediction is accurate; and summary of the study found that the common problem of imbalance of data distribution exist in software maintainability, which will directly affect the performance of the model prediction. For this,we apply decision tree to construct the software maintainability prediction model based on the sampling method,and use the UIMS and QUES datasets to conduct the experiment. The result shows that compared with the baseline method and the existing maintainability prediction method, the MMRE of the UIMS data set and QUES data set is improved by 84% and 61% respectively. And both of datasets achieve the optimum value 1 of Pred （0.25）. The result suggests the proposed method has a better comprehensive performance.

关 键 词：面向对象 软件可维护性 可维护性预测 采样方法 决策树 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]