The hotel industry, closely tied to the tourism sector, plays a critical role in supporting national economies by offering a range of accommodations to guests. Effective hotel management, encompassing finance, marketing, distribution and human resources, is essential for ensuring steady growth. However, a persistent challenge in the industry is managing reservation cancellations, which significantly impact revenue and the accuracy of demand forecasting. One of the main issues is the ignorance of what causes cancellations and how to predict them. In order to address this gap, this study uses machine learning algorithms to predict cancellations of hotel reservations. The research uses the CRISP-DM framework to assess and predict cancellations using a real-world customer transaction dataset from a 3-Star Hotel in Indonesia. Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Decision Tree and Random Forests are among the machine learning techniques that are tested in this study. With an F-measure of 94.60% among them, the Random Forest model showed the best prediction ability. This research gives hoteliers a mechanism to better predict cancellation patterns and modify cancellation and overbooking rules to minimize costs by utilizing machine learning.