This is an example of dormant customer detection. At first, we generate dataset with various features (generate_dataset.ipynb). Then we fit LightGBM model on 80% this data, and use remaining 20% for evaluation (train_model.ipynb).