Modeling and Forecasting of Tourism Time Series Data using ANN-Fourier Series Model and Monte Carlo Simulation

Danbatta, S. J.; Varol, A.

Tourism is counted as one of the most sensitive sectors to crises such as the COVID-19 pandemic. By the first quarter of 2020, it brought the foreign visitors' travels to a sudden and unexpected halt. This has negatively affected the tourism sector. Due to the perishable nature of the tourism industry products, many researchers are calling for urgent development and implementation of a rescue plan that will help in predicting the future number of foreign visitors. In this paper, we proposed an approach to modeling and forecasting a tourism time-series data that have both trend and seasonality. This approach is a combination of the Fourier series and artificial neural network methods to capture the seasonality and trend components in data. We applied this method to the monthly foreign visitors to Turkey dataset. We studied the data for the periods before, and during the COVID-19 pandemic. To account for uncertainties in the model prediction during the COVID-19 pandemic, we employed the Monte Carlo simulation method. We run 100 Monte Carlo simulations within ±2σ from the model curve. The mean of these 100 Monte Carlo simulation paths is computed and used for presenting the Monte Carlo forecast result values of the data. To test the feasibility of this approach, we compared the model predictions with some other existing models in the literature. In each case, the model has demonstrated a decent prediction and outperformed the benchmarked models. The proposed model produces a statistically good fit and acceptable result that can be used to forecast other tourism-related attributes. © 2021 IEEE.