Analysis of Code Smells Using Machine Learning Algorithms

A code smell is a sign or indication that is introduced in the source code of a software program during the design or implementation phases, which may lead to more significant problems during software maintenance. The existing methods have limitations in extracting sufficient rich information from source code, as they frequently reflect various code smells using an easy-to-understand code representation. The objective of this research is examining the code smells with machine learning classifiers. This research suggests the use of metrics extraction software named as WekaNose for the analysis of code smells and training different machine learning models for determining the efficient algorithm. In this research we have used ML algorithms like Random forest, Decision Tree, SVM, Naive Bayes and boosted algorithms such as LightGBM and AdaBoost. The proposed approach performs remarkably well in single code smell identification according to the experimental results obtained.