改进的基于PCA的软件缺陷预测方法

An Improved Approach to Software Defect Prediction Based on PCA

为了避免在变量多重情况下基于PCA的软件缺陷预测出现明显失误,对传统PCA方法的缺陷加以改进,利用UML软件工程组织网站上公布的某中等规模软件公司项目功能点数据,采用改进的PCA方法对项目综合性能进行软件缺陷预测,并与传统PCA方法的预测结果进行了对比.结果表明,改进的PCA方法对软件系统的综合性能有较好的预测能力.

This paper aims to improve the traditional principal component analysis （PCA for short） so as to avoid apparent errors in connection with software defect prediction based on PCA made under conditions of multiple variables. The improved PCA was used to predict software defects with regard to the comprehensive performance of the data from a middle - sized software company released by UML software engineering organization website. Then the results were compared with those predicted with the use of traditional PCA, indicating that the improved PCA is more effective in terms of predicting the comprehensive performance of software systems.