Combining deep neural network and fourier series for tourist arrivals forecasting

Abstract

Accurate tourist arrivals forecasting is essential for governments and the private sector to formulate policies and allocate funds more effectively. In this paper, the modeling of tourist arrivals time series data was introduced in a hybrid modeling that combines the deep neural network (DNN) with the Fourier series method. The proposed model approach applies the DNN to get the forecasted value and then employs the Fourier series to fit the residual error produced by the DNN. To verify the accurate prediction of the proposed model, different single models such as ARIMA, ANN and DNN, and modified ARIMA and ANN models using Fourier series are investigated. Historical data on monthly tourist arrivals to Langkawi Island with high trend and strong seasonality is used to compare the efficiency of the proposed model. A series of studies demonstrates that the performance of the single model can be further improved by taking into account the residual modification by Fourier series. The result shows that the proposed model is capable of forecasting tourist arrival series with higher reliability than other models used.