Code Smell Detection Using Deep Learning Models to Enhance the Software Quality

Code Smell Detection Using Deep Learning Models to Enhance the Software Quality

연구 분야: Verification

학회: International Conference on Computing, Communication and Learning

High-quality software is a major priority in the area of software development. This study goes into this goal by concentrating on two major areas: code smell detection and defect prediction. It also incorporates findings from different software project analyses to provide a complete approach to improving software quality. The objectives of this research paper are two-fold. To enhance the software quality through code smell detection to identify problematic coding patterns and structures that indicate potential issues in software code. In this paper, we have used the two deep learning models Bi-LSTM Model and GRU Model to detect the code smell present. In this paper, we have collected data from a various software project and used advanced methods to identify code smells. We also considered the three different classes namely: God class, Data class, and Feature Envy. Various software measures, static analysis tools, and machine learning and deep learning algorithms to develop efficient models for code smell. The statistical analysis reveals a strong correlation between the early detection of code smells and the subsequent reduction in software defects. It also uses the data analysis and validation of the proposed code smell detection. Statistical tests will be employed to determine the significance of the results obtained and to assess the reliability of the models. This research highlights that finding code issues early is important to make software better. The research aims to enhance software reliability, performance, and maintainability, resulting in higher-quality software systems and more efficient development processes. The highest accuracy of the Data class utilizing the Bi-LSTM is 0.99, precision is 0.97, recall is 0.94 and f-measure is 0.96.