Addressing complex seasonal patterns in hotel forecasting: a comparative study

Accurately forecasting demand poses challenges for revenue managers, especially amid supply and demand uncertainties increased by the recent global pandemic. In addition, demand forecasting is particularly challenging in the hotel industry due to anomalous days and repeating seasonal patterns. This study investigates techniques like TBATS, MSTL, and STL Decomposition against Linear Regression in hotel demand time series analysis, focusing on daily occupancy and average daily rate seasonalities. Using a 5-year dataset from an Upper Upscale branded property, the study employs in-sample data for model development and a rolling window approach for testing. Results highlight the robust performance of TBATS and MSTL across different forecasting horizons, consistently outperforming Seasonal-Trend Decomposition (STLF) and linear regression, providing insights crucial for revenue optimization and strategic decision-making in the hotel industry.