Improving Predictive Models of Software Quality Using an Evolutionary Computational Approach

Abstract

Predictive models can be used to identify components as potentially problematic for future maintenance. Source code metrics can be used as input features to classifiers, however, there exist a large number of structural measures that capture different aspects of coupling, cohesion, inheritance, complexity and size. Feature selection is the process of identifying a subset of attributes that improves a classifier's performance. The focus of this study is to explore the efficacy of a genetic algorithm as a method of improving a classifier's ability to identify problematic components.