Authors : Maria Elena Nor, Azme Khamis, Sabariah Saharan, Mohd. Asrul Affendi Abdullah, Rohayu Mohd. Salleh, Norhaidah Mohd Asrah, Kamil Khalid, Fazlina Aman, Mohd. Saifullah Rusiman, Harliana Halim, Muhammad Hisyam Lee and Eliza Nor
Abstract: In the case of tourism demand, better forecast would help directors and investors to make operational, tactical and strategic decisions. Besides that, government bodies need accurate tourism demand forecasts in the planning of the required tourism infrastructures such as accommodation, site planning, transportation development and other needs. Error magnitude measurements are commonly used to assess various forecasting models or methods. However, accuracy in terms of error magnitude alone is not enough especially in the field of economics. The information on the directional behaviour of the data is very important since if the forecast fails to predict the directional change effectively, it could cause huge negative impact on economic activities. Thus, in assessing economic forecast value, it is important to consider both the magnitudes and directional movements. This research aims to demonstrate the application of time series forecasting on Malaysia tourism demand data. Several time series methods were used, that are Box Jenkins, time series regression and Holt Winters. The forecast accuracy were evaluated by using MAPE, MAD, RMSE, Fishers exact test, mean directional accuracy and mean directional value. It was found that Holt Winters gave the most accurate forecast in terms of error magnitude. Meanwhile, in terms of directional accuracy, time series regression gave the most accurate forecast. The best model in terms of error magnitude does not necessarily give the most accurate directional forecast and vice versa.
Maria Elena Nor, Azme Khamis, Sabariah Saharan, Mohd. Asrul Affendi Abdullah, Rohayu Mohd. Salleh, Norhaidah Mohd Asrah, Kamil Khalid, Fazlina Aman, Mohd. Saifullah Rusiman, Harliana Halim, Muhammad Hisyam Lee and Eliza Nor, 2016. Malaysia Tourism Demand Forecasting by Using Time Series Approaches. The Social Sciences, 11: 2938-2945.