Scopus Indexed Publications

Paper Details

Categorizing Code Review Comments Using Machine Learning
Yeasir Arafat, Hossain Shamma, Syeda Sumbul Hossain,

Code review turns into a progressively mainstream method to detect early defects in the codebase. These days experts are rushing towards peer-investigating the codebases written by any co-located team members or other authors from distributed or dispersed teams. Chipping away at a circulated or scattered team, reviewing a codebase is required to inspect the patches before consolidating. Code looking into can likewise be a structure of approving practical and non-useful necessities. In certain circumstances, analysts do not invest enough time to comment in an organized manner, which turns into a bottleneck to other developers for tackling the discoveries or recommendations remarked by the peer-reviewers. To make the review process more progressively successful and well-organized, productive remarks are compulsory. We have extricated 2185 human code review comments of five marketed projects by mining respective projects’ repositories. Six machine learning classifiers have been utilized to train our model. Stochastic Gradient Descent (SGD) vector machine accomplishes a higher accuracy of 63.89% among the others. This work will assist the specialists with building up organized and viable code review culture among worldwide programmers or software engineers by categorizing code review comments.

Mining software repositories Modern code review Classification Machine learning Global software development
Journal or Conference Name
Lecture Notes in Networks and Systems
Publication Year