Reservation Systems Cancellation Determination Using Machine Learning
Abhishek Kumar, Archana Kumari, Ravi Kumar Burman, Muskan Pandey, Nishant Kumar, Kamaldeep Kamaldeep
Abstract
The increasing reliance on online reservation systems across various industries, including hospitality, airlines, and e-commerce, has heightened concerns over last-minute cancellations, may lead to monetary losses and ineffective operations. This study examines the application of machine learning techniques, particularly the Random Forest classifier (a bagging ensemble method), to predict reservation cancellations. In order to enable businesses to put proactive measures in place to reduce these risks, the research attempts to identify important variables that affect cancellations, such as customer demographics, booking lead times, and cancellation policies. The model was trained using 80% of the datasets, with the remaining 20% being utilized for testing. The model was assessed using key performance metrics such F1-Score, precision, recall, and accuracy. The model obtained a 0.37 F1-Score, recall of 0.38, precision of 0.37, and accuracy of 42%, indicating reasonable performance. The model's ROC-AUC score of 0.42 indicates that it has to be improved in order to better differentiate between canceled and non-canceled reservations. Overall, the findings show that Random Forest and other machine learning techniques, can be useful for predicting reservation cancellations, though further improvements are needed. Future research could concentrate on improving the model through hyper parameter tuning, feature engineering, and exploring more advanced methods to enhance predictive performance across different industries.