Forecasting hotel occupancy in São Paulo: evaluating and combining forecasting models in an event driven city
Carregando...
Arquivos
Data
2025-04-02
Orientador(res)
Cogo, Gabriel
Métricas
Título da Revista
ISSN da Revista
Título de Volume
Resumo
The competitive landscape of the hotel industry in São Paulo, Brazil, driven by corporate events and irregular demand patterns, presents significant challenges for effective revenue management and demand forecasting. This study conducted a comprehensive analysis of demand forecasting models, aiming to identify the most accurate methods for predicting room night demand in a midscale hotel located near Paulista Avenue, in the center of the city of São Paulo. The research compared the performance of traditional forecasting models—ARIMA, ARIMAX, SARIMAX, and Holt-Winters—and proposed a novel combined model to improve forecasting accuracy. Daily hotel performance data, spanning over a year, were analyzed alongside external variables such as event occurrences and day-of-week patterns. Key metrics, including MAE, RMSE, and MAPE, were utilized to evaluate model performance. The results demonstrated that Holt-Winters excelled in short-term forecasts, while ARIMAX performed better over extended periods, particularly in event-driven demand scenarios. The combined model, leveraging the strengths of Holt-Winters and ARIMAX, achieved superior accuracy, with reductions in forecasting errors across multiple metrics. This study highlights the applicability of forecasting models for independent hotels with limited resources and underscores the importance of integrating external factors like events into demand predictions. By bridging the gap in forecasting research specific to the São Paulo market, the findings contribute to advancing hotel revenue management practices and suggest opportunities for further exploration, including the incorporation of AI-driven approaches.