Rochester Institute of Technology
Improving Automatic Refactoring Candidate Identification
Abstract
dc:description.abstractExtract method refactoring is pivotal for enhancing code readability, maintainability, and modularity by segmenting complex code into clearer, isolated methods. Identifying opportunities for such refactorings necessitates a deep understanding of the codebase’s evolution and its intricate relationships. Current methodologies utilize developer commit messages, advanced graph analysis, and diverse machine learning approaches to automate this identification process. This research delves into the application of deep learning-based Large Language Models (LLMs) to tackle the complexities inherent in extract method refactoring. We introduce innovative approaches, including the use of LLMs to cluster code blocks based on complex patterns and dependencies, and the analysis of developer commit messages to infer the intent behind refactorings. These methods aim to enhance the precision of identifying refactoring opportunities by leveraging historical code data and contextual insights. Through rigorous experiments, we compare the efficacy of our proposed methods against traditional refactoring tools using metrics such as precision, recall, and F1-score. Our findings reveal the significant potential of integrating deep learning techniques into the refactoring workflow, enhancing the automation and efficacy of software maintenance. This study not only validates the use of deep learning-based approaches for code refactoring but also paves the way for future research aimed at the continuous improvement of automated software maintenance tasks.
Degree
thesis:*- Name thesis:degree_name
- Software Engineering (MS)
- Level thesis:degree_level
- Thesis
- Discipline thesis:degree_discipline
- Software Engineering, Department of
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Devoe, Ryan
- Contributors dc:contributor
-
- Mohamed W. Mkaouer
Subjects
dc:subject × 5Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.rit.edu/theses/11737