An IT Community Code Quality Ranking Model Based on Sentiment Analysis and Fuzzy Multi-Attribute Decision-Making

• Published in: Proceedings - 2024 IEEE 24th International Conference on Software Quality, Reliability and Security Companion, QRS-C 2024
• Main Author: Qin H.; Tang K.; Ma X.; Niu X.; Zheng Y.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2024
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85209791574&doi=10.1109%2fQRS-C63300.2024.00091&partnerID=40&md5=d5df3f25aecad96d174dcf68c2e4b716

With the rapid development of IT technology, the IT community has accumulated a large number of code snippets, but the quality of these codes varies, which makes it inconvenient for users to find and use codes. If these codes can be ranked according to their quality to provide reference for users, it will greatly improve user experience. Existing methods use machine learning techniques to rank code quality, but the extraction of features often carries the author's patterned subj ective judgment, and the evaluation standard is relatively single, in order to solve this problem, this paper proposes an intuitionistic fuzzy multi-attribute decision-making method based on user comments for code quality evaluation, which is divided into two steps, one is to take aspect-oriented sentiment analysis based on BERT pre-training model on user comments to obtain the fuzzy decision matrix. analysis to get the fuzzy decision matrix. The second is to use the TOPSIS fuzzy multi-attribute decision-making model and give different weighting schemes according to the content of different comments, and finally arrive at the answer ranking of a specific question. This study is based on the CSDN technical Q&A(Question and Answer) community data for instantiation verification, and the results show that the use of fuzzy multi-attribute decision-making can better adapt to the problem of the level of the commenters and differences in opinions, and the different weighting scheme better adapts to the different needs of the users who ask questions. © 2024 IEEE.