Research group

Machine Learning Methods in Software Engineering

Automatic Classification of Error Types

Project supervisor: Timofey Bryksin
Status: Completed

One of the problems with online programming courses is the lack of feedback. The main idea of this project is to cluster submitted fixes of incorrect solutions to detect frequent error types, label several largest clusters, and then use this labeled dataset to train a classifier. Such an approach allows us to show human-written hints to users who made a frequent mistake, making the education process more personalized and, therefore, effective.

Project on GitHub