Just-In-Time defect prediction for Android app for the paper Just-in-Time Defect Prediction for Mobile Applications: Using Shallow or Deep Learning? by Raymon van Dinter, Cagatay Catal, Görkem Giray, and Bedir Tekinerdogan.
The datasets can be retrieved from:
G. Catolino, D. Di Nucci, and F. Ferrucci. (2019) Cross-project just-in-time bug prediction for mobile app: An empirical assessment - online appendix https://figshare.com/s/9a075be3e1fb64f76b48.
- Clone this repository and inspect main.py
- Install scikit-learn==0.24.2, pytorch-tabnet==3.1.1, pytorch==1.9.0, and xgboost=1.4.2.
- Run main.py
The folder metrics/ contains the sklearn and TabNet metrics for evaluating the model. The folder models/ contains each of the models from the paper and a function to train the model for a single fold. DatasetLoader is used for loading and preprocessing the dataset. DefectDataset is a PyTorch data object used by the MLP model.