| Summary: | Tourism is an important industry for many nations including Malaysia. A tourism forecasting model is necessary in order to optimize resource allocation to maximize facilities and services to tourists. In realization of this issue, this paper proposes a comparison between two models (Nonlinear Auto-Regressive (NAR) and Nonlinear Auto-Regressive Moving Average (NARMA)) to forecast Malaysian tourism influx based on the volume of internet searches of the keyword 'tourism Malaysia' in Google Trends, based on proven strong correlatedness between the volume of internet searches with tourism in a particular area. Both models were constructed using two-stage Multi-Layer Perceptron (MLP) neural networks. The first stage involves the prediction of the NAR model, while the second stage involves the construction of the Moving Average (MA) part. The resulting NARMA model is a combination of both the MLPs. Results suggest that the NARMA model is more suited to approximate the tourism data due to its relatively better Mean Squared Error (MSE) and fitting results. © 2014 IEEE.
|
|---|