软件质量综合评价的投影寻踪模型

Comprehensive evaluation method for software quality based on projection pursuit

软件质量评价直接影响到软件开发的质量,软件质量是由多维指标因素决定的,投影寻踪模型能够将多维指标综合投影成一维投影指标,根据该投影指标值的大小可以对软件质量进行优劣排序。为了有效地利用投影寻踪模型降维分析和处理数据的能力、粒子群算法和多智能体遗传算法全局快速搜索的能力,文中将粒子群算法和多智能体遗传算法应用于投影寻踪模型,建立了软件质量的综合评价模型。实验结果表明投影寻踪模型为软件质量综合评价提供了一种新的客观、正确可靠的综合分析方法。

Software quality evaluation directly affects the quality of software development and software quality is determined by many indexes. Projection Pursuit （PP） model can integrate multi-dimensional indexes to one-dimensional projection index, and based on the value of it the comprehensive evaluation of samples can be given. To integrate the dimensionality reduction analysis capability of PP and the global fast search capability of particle swarm optimization （PSO） algorithm and multi-agent genetic algorithm （MGA）, the PSO and MGA are used in build PP model to achieve the comprehensive evaluation. The empirical study shows that PP model provides a new objective, accurate and reliable method for software quality comprehensive evaluation.