Software Refactoring Prediction Using SVM and Optimization Algorithms

Test suite code coverage is often used as an indicator for test capability in detecting faults. However, earlier studies that have explored the correlation between and effectiveness not addressed this evolutionally. Moreover, some of these works only small sized systems, or systems from same domain, which makes result generalization process unclear other domain systems. Software refactoring promotes a positive consequence terms software maintainability understandability. It aims to enhance quality by modifying internal structure without affecting their external behavior. identifying needs level should be executed still big challenge developers. In paper, authors explore employing support vector machine along with two optimization algorithms predict at class level. particular, SVM was trained genetic whale algorithms. A well-known dataset belonging open-source (i.e., ANTLR4, JUnit, MapDB, McMMO) study. All experiments achieved promising accuracy rate range 84% SVM–Junit system 93% McMMO − GA + Whale SVM. clear added value gained merging F-measure SVM–Antlr4 system’s 86% 96%. results proposed approach were compared four well known ML (NB-Naïve, IBK-Instance, RT-Random Tree, RF-Random Forest). The outperformed prediction performances studied MLs.