Machine-learning techniques for software product quality assessment

Integration of metrics computation in most popular Computer-Aided Software Engineering (CASE) tools is a marked tendency. Software metrics provide quantitative means to control the software development and the quality of software products. The ISO/IEC international standard (14598) on software product quality states, "Internal metrics are of little value unless there is evidence that they are related to external quality". Many different approaches have been proposed to build such empirical assessment models. In this work, different Machine Learning (ML) algorithms are explored with regard to their capacities of producing assessment/predictive models, for three quality characteristics. The predictability of each model is then evaluated and their applicability in a decision-making system is discussed.