IDENTIFING SOFTWARE BUGS OR NOT USING SMLT MODEL

Defect classifiers are widely used by many large software corporations. Defect classifiers are commonly interpreted to uncover insights to improve software quality. Such insights help practitioners formulate strategies for effective testing, defect avoidance, and quality assurance [8, 9]. Therefore, it is pivotal that these generated insights are reliable. When interpreting classifiers, prior studies typically employ a feature importance method to compute a ranking of feature importance (a.k.a., feature importance ranks). These feature importance ranks reflect the order in which the studied features contribute to the predictive capability of the studied classifier. These feature importance methods can be divided in two categories: classifier-specific (CS) and classifier-agnostic (CA) methods. The objective is to propose an improvement to the existing study on the importance of feature ranking in terms of defect classifiers..