ML Hotel Cancellation.
Booking cancellation model with an interpretable approach for day-to-day operational decisions.
AUC 0.95
- Python
- CatBoost
- pandas
- scikit-learn
PROBLEM
Late cancellations break staffing and inventory planning; the front office needs to know which bookings are at risk, and why.
DATA
Historical booking records, real and noisy: cleaning, missing values and feature preparation before any modelling.
APPROACH
Several supervised models trained and compared, CatBoost retained, with explanatory-variable analysis so results can be argued, not just trusted.
RESULT
AUC-ROC of 0.95, the main cancellation risk drivers identified and formalised into operational recommendations.
