关键词: 设计模式, 设计模式挖掘, 子图同构, 机器学习

Abstract: Mining design patterns from source code is of great significance. It can improve the comprehensibility and maintainability of software systems. Based on machine learning, a design patterns detection method is proposed. A class diagram is used to represent software systems and design patterns. Subgraph matching is utilized to extract design pattern candidates from software systems and filter numerous non-design pattern instance combinations to improve design patterns mining efficiency. Metrics are proposed tovec-torized software systems and design patterns. Machine learning is used to classify the design pattern instance candidates to improved design pattern mining accuracy. Experimental results show that the efficiency and effectiveness of this method have been improved to some extent.

Key words: design pattern, design pattern mining, subgraph isomorphism, machine learning