Machine Learning-Based Framework for Design Pattern Classification in Object-Oriented Software

Abstract

Design patterns serve as reusable solutions to common software development challenges, enabling better organization, maintainability, and scalability of code. Despite their advantages, selecting the right design pattern during development can be a complex and subjective task, particularly for beginner programmers. This paper introduces an intelligent approach that utilizes machine learning to automatically recommend appropriate design patterns based on the structural characteristics of Java source code.The proposed framework begins by analyzing Java files to extract key structural features, which are then encoded into numerical vectors. These vectors capture the essential aspects of the code design and are used as input to machine learning algorithms such as Support Vector Machine (SVM), Decision Tree, and Random Forest. Additionally, an ontology-based similarity ranking method is employed to enhance the precision of predictions by measuring the closeness between the input code and existing pattern examples in the dataset.To make the solution user-friendly and accessible, the model is deployed as a RESTful API. Users can submit their source code through a web interface and receive instant feedback on the most likely design pattern classification, along with model confidence levels. Experimental evaluations indicate that the Random Forest model consistently delivers high accuracy in predicting one of the 13 predefined design patterns, outperforming the other classifiers tested.This system not only supports developers in making more informed design decisions but also contributes to the automation of software architecture practices. The integration of machine learning with software engineering principles creates a valuable resource for both academic research and industry application.