Software Defect Prediction: An ML Approach-Based Comprehensive Study

Abstract

Software testing is a vital phase in the software development life cycle. Testing validates a developed software against the input test cases by identifying the defects present in the system. The phenomenon of testing is not only time-consuming but also a costly affair. Although there are automated tools available that reduce the effort of testing up to some extent, the high maintenance cost of these tools only increases the cost. Earlier defect prediction in software significantly reduces the effort and cost without affecting the constraints. It identifies the defect-prone modules that require more rigorous testing. A practical and effective defect prediction mechanism is the need of the hour due to the challenges, namely dimensionality reduction and class imbalance, present in software defect prediction. Lately, machine learning (ML) has emerged as a powerful decision-making approach in this regard. This research work aims to do an extensive study on the implementation of ML techniques in software defect prediction. This comprehensive report is based on two different aspects named feature selection/reduction techniques and ensemble learning methods that have been used in software defect prediction. This study has also discussed the widely used software and performance measure metrics used in software defect prediction. This concise work would guide future researchers in this emerging research area. Further, this paper also emphasizes the need to identify a suitable feature selection approach that could enhance the model's predictive performance when applied with ensemble learning.