Abstract:

Software defect prediction can help developers predict defect programs in advance and allocate limited test resources reasonably. Software defect prediction accuracy depends not only on prediction methods choice but also on software metrics. Therefore, ?it has become a research hotspot to predict defects by combining?multi-metrics. Starting from the metrics, the research progress of traditional metrics, multiple metrics, and defect prediction combined with multiple metrics is introduced systematically. The main work includes：firstly, the traditional code and process metrics, the software defect prediction model based on traditional metrics and the factors that affect data quality are introduced; secondly, the semantic structure metrics are discussed; then the analysis lists the current software defect prediction evaluation indicators; finally, the future research trend of multi-metrics software defect prediction is prospected by combining prediction granularity, traditional metrics, semantic structure metrics, and cross-project software defect prediction.

Key words: multi-metrics, traditional metrics, semantic metrics, structure metrics, software defect prediction

摘要：

软件缺陷预测可帮助开发人员提前预测缺陷程序，合理分配有限的测试资源。软件缺陷预测的准确度不仅依赖于预测方法的选择，更依赖于软件的度量指标。因此，结合多元度量指标进行软件缺陷预测已成为当前的研究热点。从度量指标出发，对传统度量指标、多元度量指标以及结合多元度量指标的缺陷预测的研究进展进行了系统介绍。主要工作包含：介绍了传统的代码和过程度量指标、基于传统度量指标的软件缺陷预测模型以及影响数据质量的因素；阐述了语义结构度量指标；分析列举了当前用于软件缺陷预测的评价指标；结合预测粒度、传统度量指标、语义结构度量指标、跨项目软件缺陷预测对多元度量指标软件缺陷预测未来的研究趋势进行了展望。

关键词: 多元度量指标, 传统度量, 语义度量, 结构度量, 软件缺陷预测