An improved software defect prediction algorithm using self-organizing maps combined with hierarchical clustering and data preprocessing

Abstract

An improved software defects prediction algorithm based on combination of Kohonen map and hierarchical clustering is presented in this paper. The need for software reliability assessment and analysis growths rapidly due to increasing dependence of our day-to-day life on software-controlled devices and systems. Software reliability prediction is the only tool available at early stage of software development lifecycle when the debugging cost risk of faulty operation is minimal. Artificial intelligence and machine learning in particular are promising techniques to solve this task. Various classification methods have been used previously to build software defect prediction models, ranging from simple, like logistic regression, to advanced methods, e.g. multivariate adaptive regression splicing. However, the available literature still does not allow to make unambiguous conclusion concerning the choice of the best classifier and trying different dimensions to overcome potential bias is suggested. The purpose of the paper is to analyze the software code metrics to find dependences be-tween software module’s defect-proneness and its metrics. JM1 public NASA dataset from PROMISE Software Engineering Repository was used in this study. To increase the classification accuracy, we combine self-organizing maps with hierarchical clustering and data preprocessing.