Software Analytics aims to make software developers life easier with building software tools, developing heuristic/ml/dl models using software artifacts (e.g. source code, software repositories, feature specifications, issue/bug reports, test cases, execution traces/logs, and real-world user feedback). To have a better understanding, you may refer to the link microsoft-software-analytics, it is a short but well-defined description of Software Analytics.

The idea behind the project is on bug report assignee recommendation. One of the most famous bug report tools is JIRA, and there are many open source project data available e.g. apache-jira. I chose Apache Kafka as the dataset, it is an open-source distributed event streaming platform, commonly used by Data Engineers. A bug report consists of several fields (Description, Assignee, Reporter, Status, Priority, Resolution, Affects Version/s, Fix Version/s, Component/s) and many more. A bug description is a text field that information related with the bug is defined and an assignee is a developer who is usually assigned by a Team Lead and responsible for fixing that bug. The main idea behind the project is to predict an assignee to a bug report by using its' bug description. This project can be helpful for Software Team Leads while assigning bugs to Software Developers in a team.

Bag-of-Words classification model (simple one) to serve as a baseline, test accuracy is around 26%, considering there are 36 possible assignees (less than 3% random choice), the model consists the baseline. Secondly I experimented several BiLSTM models and the best one is performing 41% on test set. Lastly, I did the same for BERT and achieved 56% test accuracy.