ML-BASED SOFTWARE DESIGN PATTERN DETECTION IN SOURCE CODE USING NEURAL NETWORK

Abstract

Software design patterns play a vital role in improving code reusability, maintainability, and scalability in software development. However, identifying design patterns within large and complex source codebases remains a challenging and time-consuming task, especially when patterns are implemented in diverse ways by different developers. Traditional static analysis tools and rule-based approaches often fail to detect patterns accurately due to variations in coding styles, incomplete documentation, and the presence of anti-patterns. To overcome these limitations, this study proposes a machine learning (ML)-based approach using neural networks to automatically detect software design patterns in source code. The proposed model leverages source code features such as class relationships, method invocations, inheritance structures, and object interactions to train a neural network that can recognize underlying design patterns. Deep learning techniques, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), are employed to capture both structural and sequential aspects of the code. Experimental results demonstrate that the proposed model achieves higher accuracy and robustness compared to traditional rule-based systems. This approach not only automates the detection process but also assists developers in understanding and refactoring legacy code, ultimately improving software quality and maintainability.