基于FARM降低系统性偏差的缺陷修复时间预测方法  

BUG FIX-TIME PREDICTION BY REDUCING SYSTEM DEVIATION BASED ON FARM

作　　者：吴含宇 肖俊超[2,3] 王青[2,3] 

机构地区：[1]中国科学院大学,北京100149 [2]中国科学院软件研究所互联网软件技术实验室,北京100190 [3]中国科学院软件研究所计算机科学国家重点实验室,北京100190

出　　处：《计算机应用与软件》2015年第7期5-9,15,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(91218302;91318301;61303163;61202064);国家重大科技专项(2012ZX01039-004)

摘　　要：缺陷修复时间预测能够帮助开源社区管理者精确地估计缺陷修复工作量,实现缺陷修复人员的优化分配。缺陷追踪库Bugzilla是开源社区缺陷修复数据的主要来源。在实际使用中,缺陷追踪库除被用作记录缺陷之外,还被用于记录用户对软件产品不足之处的投诉,以及软件功能相关讨论。缺陷追踪库的混合使用给缺陷修复时间预测带来系统性偏差。为降低系统性偏差对预测结果的影响,给出基于模糊关联规则挖掘FARM(Fuzzy Association Rules Mining)降低系统性偏差的缺陷修复时间预测方法。该方法通过分析历史数据生成模糊关联规则来过滤系统性偏差。实验结果显示经过该方法处理后的新预测模型结果优于原有模型。Predicting bug-fix time is able to help the mangers of open source communities precisely estimate the workload of bug-fix and realise the optimised distribution of bug-fix personnel. Bug tracking repository Bugzilla is the principal source for bug-fix data in open source communities. In actual use, the bug tracking repositories are used to record the complaints of users in regard to disadvantages of software products apart from recoding the bugs themselves, and to the correlated discussions about software functions. The mix usage of bug tracking repositories brings systematic deviation to the prediction of bug-fix time. To reduce the impact of systematic deviation on prediction results, this paper presents a bug-fix time prediction method which reduces system deviation based on fuzzy association rules mining （FARM）. The method filters the systematic deviation by analysing historical data and generating fuzzy association rules. Experimental result shows that the results of new prediction model treated by the method are better than that of previous models.

关 键 词：缺陷追踪库 模糊关联规则挖掘 缺陷修复时间预测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]