Empirical Analysis of Machine Learning Classifiers for Architectural Smells Using Critic-Topsis Approach

Code Smells are resultants of poor coding techniques adopted by programmers which not only depreciates the quality of the software but also makes it tedious to maintain. We report best Machine learning classifier for the detection of the most severe smell of the Architectural category. The architectural category needs further investigation thereby this category of smell is considered. The most severe architectural smell is analyzed to rank eight Machine Learning (ML) classifiers. Eight performance measures of the ML classifiers are used during the study. Criteria Importance Through Inter Criteria Correlation and Technique for Order Preference by Similarity to an Ideal Solution (CRITICTOPSIS) method is used to analyze the performance measures of the classifiers for the severe architectural smell which are then ranked. J48 classifier is reported to be the best classifier for the analysis of the Feature Concentration smell of the architectural category.