Code Smell Detection using Machine Learning Classification Algorithm

Code smell indicates a poor implementation choice that affects software quality attributes (Pérez, 2013). Fowler (1999) also describes it as an internal code-level problem where the code becomes complex, the design broken, and eventually worsens software quality. Jose (2020) has reported that most applied existing approaches for code smells detection are search-based (30.1%), metric-based (24.1%), and symptom-based approaches (19.3%). However, these existing approaches can only apply to simpler detection; the greater the complexity of code smell, the lower the results for code smell detection (Mantyla M, 2004). Kessentini (2014) also has reported that detecting the problems of code smell is difficult and the performance is not effective using the existing approaches such as search-based, symptom-based, visualization-based, probabilistic, cooperative-based, manual, metrics-based, and rule-based. As a result, many of these approaches extend to the application of machine learning classifiers in software code smell detection. Fontana (2016) reported that a supervised machine learning strategy can be used to forecast the value of the dependent variable using machine learning classifiers to address the problem. In this project, we propose a machine learning supervised Gaussian processes algorithm for JAVA open-source code smell detection. The Gaussian process is a highly interpretable supervised machine learning algorithm used in regression testing to quantify prediction uncertainty. A code smell detection application prototype will be developed to implement the proposed work. The effectiveness of the proposed work in terms of detection accuracy will be evaluated further.