Optimising SVM with Improved Grey Wolf to Detect Code Smell Severity

Code smells indicate design problems that deteriorate the system's quality over time. They may introduce errors and make the system hard to maintain and understand. Smells can be removed through refactoring, the process of making changes to software code that enhance the internal organisation of the system without affecting its external behaviour. To be refactored, smells must be identified and ranked according to severity. The severity ranking aims to eliminate the most dangerous code smells. The Support Vector Machine(SVM) is optimised in this work using the metaheuristic technique, Improved Grey Wolf Optimization (IGWO). Performance metrics were examined before and after tuning the hyperparameters of the SVM classifier, and the result obtained was 98.79% Accuracy, 99.32% Precision, 97.20% Recall and 98.72% F-measure in Feature Envy. This integration solved a recurring software engineering problem, improved classifier performance, and automated code smell severity detection.