For efficiency purposes, Betsson is trying to predict which customers are going to call the Customer Service, based on their past behaviour.
For a given day, predict whether a customer will call the Customer Service in the following 14 days.
The main notebooks are numbered in the folder "notebooks":
- 0-eda.ipynb
- 1-feature-selection.ipynb
- 2-predictive-modeling.ipynb
- pycaret.ipynb
The predictions for the test set can be found at predictions/test_predictions.csv.
Additionally, I've created a web app and an API to consume the model. I used Streamlit and FastAPI, respectively.
Links:
GitHub Repositories:
Docker images:
Note: I'm using the free resources from Render. Downtime might occur (there's limited memory usage as well).