Supervised Machine Learning Approach for Software Maintainability Assessment

EasyChair Preprint 4590

Abstract

In the highly competitive product market, the importance of quality is no longer an advantage, but a necessary factor for the success of the company. The work presented in this paper focuses on software quality assessment using a supervised machine learning approach. Software quality is a vague and multifaceted concept and varies from person to person. Thus, assessing quality depends on the perspective we have, making direct quality assessment very difficult. In this document we evaluate quality from the developer’s point of view. The ISO 9126 standard defines six factors (internal and external) of high level to characterize the quality of a software product. In the following we use the standard to identify the quality requirements. The software being an abstract entity, the essential element for its evaluation is its source code. We use static code extractor to extract the metrics it contains. These metrics are used as inputs to the machine learning system. The machine learning system is used to discover the knowledge that is hidden in the data. This is done by making the best possible approximation to reality. Our work proposes a formula to calculate the quality of the software. To do this we use supervised machine learning algorithms to optimize the quality formula. Our quality formula thus offers developers an objective and independent view of the concept of quality on the one hand, but also to be able to build good quality products on the other hand.

Keyphrases: Quality factor, Supervised Machine Learning, quality metrics, software product, software quality

@booklet{EasyChair:4590,
  author    = {Stephane Nkeuga Ngueliekam and Mathurin Soh},
  title     = {Supervised Machine Learning Approach for Software Maintainability Assessment},
  howpublished = {EasyChair Preprint 4590},
  year      = {EasyChair, 2020}}