SLR Machine Learning Techniques for Code Smell Detection

Objectives
1. (i) the types of code smells taken into account by previous research
2. (ii) the dependent and independent variables proposed in literature to identify code smells
3. (iii) the types of classifiers exploited by researchers
4. (iv) the training strategies used to train and evaluate the machine learning techniques
Research Questions
1. RQ1 - Code Smells Considered
  1. God Class, Long Method , Feature Envy, Spaghetti Code, Functional Descomposition
2. RQ2 - Machine Learning Setup
  1. RQ2.1 Independent Variables: The CK metric suite is the most used one
  2. RQ2.2 Dependent Variable: Binary, Probability and Severity Level
  3. RQ2.3 Machine Learning Algorithms: Decision tree, Support Vector Machines, Random Forest, Naive Bayes
3. RQ3 - Evaluation Setup
  1. RQ3.1 Validation Techniques: k-fold cross-validation
  2. RQ3.2 Evaluation Metrics: precision, recall
  3. RQ3.3 Code Smells Datasets: Qualitas Corpus dataset, Gantt Project, ArgoUML, Eclipse JDT
4. RQ4 - Performance Meta-Analysis
  1. RQ4.1 - Impact of Independent Variables
  2. RQ4.2 - Impact of Machine Learning Algorithm on Performance: JRip and Random Forest (most effective)
  3. Rq4.3 - Impact of Training Strategies

Resource summary

	Created by Guisella Aa over 6 years ago