Portal de Periódicos da CAPES

Developing classifiers by considering sentiment analysis of reported bugs for priority prediction

2023; Springer Science+Business Media; Volume: 15; Issue: 5 Linguagem: Inglês

10.1007/s13198-023-02199-2

0975-6809

Anisha Singh, P. K. Kapur, V. B. Singh,

Software System Performance and Reliability

Software systems behave abnormally due to bugs when it comes into the operational phase. Lack of proper understanding of customer requirements, implementation, knowledge, wrong algorithmic designing, and other issue is also the reason for bug production. To fix those flaws, developers request to the users for feedback. Users have had issues with the software systems that have been released. Users are encouraged to submit their issues to issue-tracking systems such as Bugzilla, Mantis, Google Code Issue Tracker, GitHub Issue Tracker, and Jira to improve the next version of the product and meet user needs. Manual prioritization is time-consuming and inconvenient. In this research paper, we propose using sentiment analysis to anticipate the report's priority. This is the first time the sentiment-based approach used for a bug report to prioritize prediction on open-source projects. First, we take the bug report summary and use natural language pre-processing techniques to clean the text and pre-process the bug report. Second, sentiment analysis is applied to clean texts that contain sentiments of terms. Third, we use TF-IDF to construct a feature vector for bug reports, fourth, we used resampling techniques to balance the dataset, and then we used different machine learning classifiers to train historical data namely Bugzilla open-source projects to forecast their priority. The proposed method we have used improves the performance of the classifier with sentiment comparison to without sentiment on average f-score 2–10%.