Cancellation of bookings puts considerable pressure on management decisions, in this case from the hospitality industry. Cancellation of bookings limits the correct prediction and is, therefore, a very important tool for revenue management performance. However, in recent times, thanks to the availability of considerable computing power through machine learning approaches, it has become possible to create more accurate models for predicting booking cancellations compared to using more traditional methods. Previous research has used several machine learning approaches, such as Decision Tree, Support Vector Machine, Deep Neural Network, Logistic Regression, and Random Forest to predict hotel cancellations. However, they have not addressed the class imbalance problem that exists in predicting hotel cancellations. In this study, we have provided a solution by introducing an oversampling technique to solve the class imbalance problem, together with the k-nearest neighbors algorithm to predict hotel booking cancellations better. The results of this study show that an increase in the performance of the method's accuracy increased by 3.88%, precision increased by 9.00%, recall increased by 10.00%, and F1-Score increased by 10.00% in the hotel booking dataset. It can be concluded that the SMOTE method with KNN has a better performance than only using the KNN method in predicting the cancellation of hotel reservations.
Copyrights © 2022