一种软件特征定位研究结果的评估方法

摘要

软件特征定位是软件演化得以成功实现的重要前提条件,且软件特征定位实验结果的评估标准直接关系到软件演化活动的适用范围.当前软件特征定位结果的评估方法通常采用实验检索结果的10%~15%检索量来进行下一步的定位实验,会不可避免地造成查找范围过大且查准率过低的情况.同时,由于特征定位相关领域新技术的出现,急需一种新的特征定位研究结果的评估方法来提升特征定位的效率.通过多次实验的总结,结合波及效应分析及当前使用的一些评价实验结果的方法,认为每次定位出一个源代码文件,恰好该源文件也是特征相关的源代码文件为最理想的特征定位.为此,提出一种新的实验结果评估方法,将检索量从实验数据总数的10%~15%降到1个实验数据.数据分析结果表明,由所提出的评估方法所获得的特征定位结果可以有效地缩小软件演化的范围并提高软件演化的效率.%Software feature localization is an important prerequisite for the successful development of software evolution,and the evaluation criteria of software feature localization experiment results are directly related to the scope of software evolution.The evaluation method of current software features location results of experiments using the 10%~15% retrieval of retrieval results are used for the next step experiment,which could inevitably cause the too large search range and low precision.Meanwhile due to the emergence of new technologies in the field of feature localization,a new evaluation method of research results is urgently necessary to improve the efficiency of feature location.Through the summary of many experiments,combined with the ripple effect analysis and the current use of some experimental results of the evaluation methods i.e.each time a source code file is located exactly that the source file is also a source code files related feature for the most ideal feature location.For this purpose,a new experiment result evaluation method has been proposed,which has reduced the retrieval amount from the total number of experimental data of 10%~15% to 1.The experimental results show that the feature localization results obtained by the proposed evaluation method have effectively narrowed the scope of software evolution and improve the efficiency of software evolution.