SOFTWARE DEFECT PREDICTION USING CLASS IMBALANCE LEARNING

Abstract

Software testing is important part of the software development. Testing requires maximum resources and effort. Defective modules in the software can risk the development and maintenance costs. It is important to detect the defect ìn the starting stages of applìcation development lìfe cycle. Defect prone modules should be identified and quality assurance activities are enforced. Earlier Predicting of software faults improves the efficiency, reliability, software qualìty and reduces the cost of software. To facilitate software testing and reduce testing costs, various machine learning approaches are explored to predict faults in software modules. Class imbalance learning specializes in solving classification problems of "balanced distributions", which can be useful in predicting defects, but have not yet been widely studied. In thìs article, I have outlined whether and how learning methods for class imbalance will benefit from software defect predìctìon in order to find better solutions. Software defect prediction is defined as the process of finding faulty modules in software using certain historical data and software metrics to improve software quality, and this software defect prediction process reduces the number of modules to be tested. In software development, the prediction of software faults can be framed as a learning task that is gaining "increasing attention" both in academia and in industry. The characteristics of the static code are taken from previous versions of the product with error logs and are used to create models to predict bad modules for the next version.