Predictions of Hotel Dynamic Pricing for Revenue Management System

Hotel room prices are usually set and changed dynamically by the Revenue Manager (RM). RMs daily monitor the Key Performance Indicators (KPIs) recorded for future Days Of Stay (DOS), along with the market conditions and other external factors, and adjust the selling prices to maximize the revenue of the receptive structure. Having a well-established dynamic pricing strategy, this manual adjustment of prices performed by the RM is costly and time-consuming. In this work, we propose an approach to automate dynamic pricing for hotel rooms. To this aim, we designed a dataset structure that integrates the typical logic of time series input into the tabular structure used to run ML models. The dataset pertains to actual accommodation facilities, is annotated by revenue management experts and incorporates static, dynamic, and engineered features. We benchmark five machine learning models to automatically predict the price that a Revenue Manager (RM) would dynamically set for an entry-level room over the next 90 days. Our approach has been tested and assessed across three different hotels, demonstrating potential adaptability to other room types. Furthermore, to allow RM understanding the predictions given by the model we propose the use of a technique of Explainable Artificial Intelligence (XAI), that is the SHapley Additive exPlanations (SHAP). We also built the architecture of the entire process to show how the system should work when integrated in a Revenue Management System. To the best of our knowledge, the problem addressed in this paper is understudied and the results obtained in our study can help further research in the field.