Forecasting Tourist Daily Arrivals With A Hybrid Sarima–Lstm Approach

Abstract

Timely predicting tourist demand is extremely important for the tourism industry. However, due to limited availability of data, most of the relevant research studies have focused on data on a quarterly or monthly basis. In this article, we propose a novel hybrid approach, SARIMA + LSTM, that is, seasonal autoregressive integrated moving average (SARIMA) combined with long short-term memory (LSTM) to forecast daily tourist arrivals to Macau SAR, China. The LSTM model is a novel artificial intelligence nonlinear method which has been shown to have the capacity to learn the long-term dependencies existing in the time series. SARIMA + LSTM benefits from the predictive power of the SARIMA model and the ability of the LSTM to further reduce residuals. The results show that the SARIMA + LSTM forecast technique outperforms other methods.