Develompent of Machine Learning Model to Predict Hotel Room Reservation Cancellations

Abstract

The increasing frequency of hotel room reservation cancellations poses significant challenges to the hospitality industry, particularly in tourist-heavy regions like Borobudur, Indonesia. This study aims to develop a machine learning (ML) model to predict reservation cancellations and provide actionable insights for hotel management. The data used in this research was collected from hotels around Borobudur, focusing on factors such as booking lead time, arrival month, and reservation status. Three ML models were implemented: Random Forest, Logistic Regression, and Bayesian Networks. Among these, the Random Forest model demonstrated superior performance with an accuracy of 86.36%, precision of 88.06%, recall of 93.65%, and an F1 score of 90.77%. The Logistic Regression model also performed reasonably well, but the Bayesian Network model underperformed in generalization tasks. These findings suggest that predictive models, particularly Random Forest, can effectively mitigate the financial risks associated with last-minute cancellations, enabling hotel operators to optimize pricing strategies, manage inventory, and improve operational efficiency.