This study/paper aims to reach the core objective of hospitality order cancellation prediction (HOCP), that is, to identify potential cancellers from many customer bases, thereby enhancing the effectiveness of customer retention campaigns. However, few studies have focused on predicting hospitality order cancellation.
A novel profit-driven model for predicting hospitality order cancellation is proposed to bridge this research gap. The authors construct profit-driven extreme gradient boosting (XGBoost) based on a grid search on HOCP to maximize profit by selecting optimal hyperparameters of XGBoost.
Real-world data set is analyzed, and the proposed model yields more profits than other predictive models. Sensitivity analysis proves that the proposed model is robust to the key hyperparameter and application scenario. Furthermore, some preventive measures based on visual analysis results are provided to reduce the cancelled probability of orders.
This research will help hotel managers to transfer the modeling goal to profit orientation and encourage relevant researchers to interpret the prediction results of models for hotel order cancellation prediction in a post hoc manner. Besides, the proposed model can be applied to various enterprises with different average order profits and help managers optimize revenue management.
This research expands the relevant literature and offers guidance for predicting hospitality order cancellation from a profit-driven perspective at the customer level. The proposed model can provide macro-control to hotel managers and obtain the most satisfactory profits in micro-control.
