Forecasting monthly tourist arrivals to Sri Lanka

Abstract

The accurate forecasts of tourist arrivals have a significant impact on the economy of a country. The patterns of arrivals from different countries may be varied in a same region. Thus, forecasting the tourist arrivals to Sri Lanka is important for decision making processes. On the view of the above, the study has focused to forecast monthly tourist arrivals from highest tourist generating countries such as United Kingdom and India. In this context three forecasting techniques, namely Seasonal ARIMA, Holt Winters (HW) Multiplicative model and Holt Winters Additive model were employed to find the most appropriate model with least forecasting error. The models were trained from monthly tourist arrivals for the period from November 2010 to August 2017 and validated using data from September, 2017 to February, 2018. The Seasonal ARIMA(0,1,1,)×(0,1,1)12 was identified as the best fitted model for both countries. Among Holt Winters models, Holt Winters multiplicative model with smoothing constants 𝛼=0.3,𝛽=0.1 and 𝛾=0.1 was found to be the most suitable model for both countries. In both models errors were found to be white noise. The forecasts of monthly tourist arrivals from the UK and India from both models have high accuracy as corresponding values of percentage errors were within ±10 % and MAPE is less than 10% for independent set. By comparing percentage error for both training and validation set it was found that SARIMA is more superior than HW. The percentage changes of monthly tourist arrivals reveal that it can be expected an increment of monthly tourist arrivals in the coming months. The models developed in this study are recommended to use for policy decisions in medium term forecasting and which would be useful for the tourism industry in Sri Lanka.