Code Comprehensibility Evaluation for Java Class based on LDA Topic Model

The rapid development of the current software industry has produced a large amount of software, to evaluate the quality of a software code is difficult, which is also not conducive to software maintenance and iteration. Among them, software comprehensibility is one of the most important factors for software maintainability, and it is one of the indispensable elements for software quality evaluation. However, most of the existing software quality evaluation only stay in the concept level, lack of corresponding evaluation methods. From the perspective of java class code, this paper constructs a comprehensibility evaluation method for java class code based on LDA topic modeling, implements the software class code intelligibility evaluation, and quantifies the experimental results using clustering algorithm and WordNet similarity calculation method. Experiments show that the proposed method can clearly distinguish the comprehensibility of class code and obtain reliable evaluation results of software class code intelligibility.