Develompent of Machine Learning Model to Predict Hotel Room Reservation Cancellations

Abstract

The frequent cancellations of hotel room reservations have become a pressing issue for the hospitality industry, especially in high-tourism areas such as Borobudur, Indonesia. This research develops a predictive machine learning (ML) model to identify cancellation probabilities to support proactive decision-making for hotel management. Using datasets from Borobudur-based hotels, key variables such as booking lead time, arrival month, and reservation outcomes were analyzed. Random Forest demonstrated the best performance, achieving an accuracy of 86.36% with a precision of 88.06%, recall of 93.65%, and F1-score of 90.77%. Logistic Regression demonstrated moderate effectiveness, while Bayesian Networks underperformed, highlighting the importance of robust algorithms for such tasks. The findings underscore the potential of ML models, particularly Random Forest, to reduce financial losses and enhance operational efficiency in the hospitality sector by anticipating cancellations and facilitating better resource allocation