Tourism is globally accepted as one of the largest business industries that can easily be affected by a variety of external factors and events. Assessing and anticipating the impact of certain events on tourism has become a necessity in this increasingly dynamic and demanding world. According to TURIHAB, there are no mechanisms that allow objective forecasts for habitational tourism, namely its brands Solares de Portugal and Casas no Campo . This paper documents the techniques, models and platforms explored to develop the expected forecast mechanism. The arduous process of obtaining real data series to train the regression model, together with the exploration of two statistical models ( Random Forest and SARIMAX ) resulted in the tools for forecasting the TURIHAB association’s monthly occupancy rates for the years 2024 and 2025. In addition, adjustments on the training/test sets and hyperparameter tuning techniques were explored in order to improve the model. The performance evaluation followed specific metrics that fit the type of forecast expected. Finally, a comparison was made between the predicted and actual values as a validation strategy. The resulted dataset is in the process of being published. For any interest on accessing the developed models and used datatset, contact the authors.
Rocha et al. (Thu,) studied this question.