Automation of Program Code Analysis Using Machine Learning Methods
Vadim Moshkin, Vladislav Kalachev, Зарубин Антон Александрович · 2022 International Russian Automation Conference (RusAutoCon) · 2022
The paper presents a description of the developed approach and service for analyzing source code in Python. The service allows you to reduce the time spent on the code review process by partially automating this process. After publishing a changeset to a remote Git repository, the service starts checking the changeset and publishes reports of bugs and duplicates found in changeset comment format. A neural network with an original architecture is used for automatic code reviews. PyTorch, Scikit-learn and Transformers libraries are also used. The testing performed did not reveal any errors that affect the operation. All the main functions of the system are performed correctly.