Modeling and forecasting hotel room demand based on advance booking information

This study develops a stochastic approach to the short-term forecasting of hotel booking arrivals. We investigate the key characteristics of booking arrivals, specifically the time-varying arrivals rates, high variability in the final demand, and the strong positive correlations between arrivals in different time periods. We examine three Poisson mixture models to capture these salient features of booking arrivals. In particular, the presence of strong inter-temporal correlations can be leveraged for forecasting future arrivals based on the early realizations. We suggest a new forecasting method that exploits the intrinsic correlations between early and late bookings and then present validation results of data from a major hotel chain along with a comparison to benchmark models. Our empirical study confirms that our dynamic updating method leveraging inter-temporal correlations can significantly improve the short-term forecasting accuracy of hotel room demand.