SBCDetector: a hybrid approach to detect second-order similarity or change

Abstract

Software Configuration Management (SCM) involves tracking similarities/changes during software evolution. Efficient comparison for tracking requires two perspectives—Granularity: comparing the entities at file level, class level, and method level. Second, Robustness should be prominent to detect renaming and shifting that occur as a part of restructuring. Even GIT repository, which is widely used, allows such comparison with renaming and shifting details but is limited to file level only, along with its own limitation of default similarity criteria of above 50%. In this study, the proposed technique named SBCDetector detects similarity/change status with both perspectives that is lacking in the existing literature. Result shows that one-fourth of entities have been found renamed/shifted at three granularities for eight subject systems improving tracking, understandability, and onboarding. Hybrid technique involving fuzzy logic derives classification model with .99 f-score to detect first- and second-order similarity/change.