For large open source software projects, users submit a large number of defect reports, manual distribution of defects will be a lot of misallocation.By calculating the similarity between the new defect report and the historical defect report, K historical defect reports with the highest similarity and the corresponding list of repair persons are obtained based on the K nearest neighbor algorithm, and then based on frequency and social network map of the indicators, the OSDR (Open Software Developer Recommendation) method proposed in this paper evaluates the developer''s professional competence.The real experiment data were collected from the Mozilla Firefox database to compare the accuracy and recall of different social network indicators when recommending the human.The results show that the recommended performance index is the frequency and out, and its accuracy is about 0.6 or so;Betweenness and Closeness recommended effect is the worst;Degrees, in-degree and PageRank recommended effect is good.%对于大型开源软件项目来说,用户提交了海量缺陷报告,人工分发缺陷时会出现大量的错误分配.提出OSDR(Open Software Developer Recommendation)方法通过计算新缺陷报告和历史缺陷报告之间的文本相似度,基于K最近邻算法得到相似度最高的K个历史缺陷报告及其对应的修复人列表,再基于频率和社交网络图的各项指标对开发者专业能力进行评价.从Mozilla Firefox 缺陷库中采集真实实验数据,比较不同社交网络指标在推荐修复人时的准确率与召回率.结果表明,推荐性能最高的指标是频率和出度,其准确率大约在0.6左右;Betweenness和Closeness的推荐效果最差;度、入度以及PageRank推荐效果良好.
展开▼
