A Novel Four-Way Approach Designed With Ensemble Feature Selection for Code Smell Detection

Purpose: Code smells are residuals of technical debt induced by the developers. They hinder evolution, adaptability and maintenance software. Meanwhile, they very beneficial in indicating loopholes problems bugs Machine learning has been extensively used to predict Smells research. The current study aims optimise prediction using Ensemble Learning Feature Selection techniques on three open-source Java data sets. Design Results: work Compares four varied approaches detect code performance measures Accuracy(P1), G-mean1 (P2), G-mean2 (P3), F-measure (P4). found out that values did not degrade it instead either remained same or increased with feature selection Learning. Random Forest turns be best classifier while Correlation-based selection(BFS) is amongst techniques. aggregators, i.e. ET5C2 (BFS intersection Relief Forest), ET6C2 union ET5C1 Bagging) Majority Voting give results from all aggregation combinations studied. Conclusion: Though good, but needs a lot validation for variety sets before can standardised. also pose challenge concerning diversity reliability hence exhaustive studies.