Hotel Booking Cancellation Prediction Using Machine Learning for Improved Reservation Management (Published)
Hotel booking is a vital part of hotel operations that supports the information needed for better management of room inventory and availability, occupancy, and resources. However, cancellations can complicate expected occupancy levels and have a continuing impact on reservation planning and revenue management. The objective of this study is to improve and compare four machine learning (ML) models (XGBoost, Random Forest (RF), Decision Tree (DT), and K-Nearest Neighbors (KNN) to identify the most effective model for hotel booking cancellation prediction. The analysis was performed based on the Hotel Booking Demand dataset containing a total of 119,390 records with 31 attributes. This involved missing value treatment, feature engineering, and data type conversion for encoding categorical variables. Models were assessed using accuracy, recall, precision, F1-score, and confusion matrices. XGBoost obtained an accuracy of 98.39%, precision values of (0.99 and 0.97), recall values of (0.98 and 0.98), and F1-scores of (0.99 and 0.98) for Classes 0 and 1respectively, outperforming the other models. It was followed by RF with 98.24%, DT with 97.95%, and KNN with 97.82%. These results show that XGBoost achieved the strongest predictive performance among the evaluated models and can be utilized for reservation planning, resource allocation, capacity planning, and revenue management.
Keywords: CRISP-DM, hotel booking hotel booking demand, machine learning, reservation management