Hotel Booking Cancellation Prediction (Machine Learning Project)
Tools: KNIME, RapidMiner | Techniques: EDA, Decision Tree, Random Forest | Date: July 2025
Developed a predictive model for INN Hotels Group to identify and reduce booking cancellations that were negatively impacting revenue and operational efficiency. Using customer reservation data (lead times, deposits, room types, and guest behaviors), I conducted exploratory data analysis and built Decision Tree and Random Forest classifiers to forecast cancellation risk.
Key findings showed lead time as the strongest predictor of cancellations, with deposit type and special requests also influencing guest commitment.
Decision Tree Model: 85.4% test accuracy after MDL pruning for interpretability.
Random Forest Model: 87.4% test accuracy with stronger recall and precision, selected for deployment.
Delivered actionable insights to refine pricing, improve staffing forecasts, and apply dynamic cancellation policies.
Outcome: Enabled data-driven strategies for revenue recovery and improved operational planning through predictive analytics.