Artificial Neural Network for Indonesian Tourism Demand Forecasting

Abstract

Tourism industry shows a positive growth and uphold an important role in national economy as the second largest portion of foreign exchange contributor, as well as its role in national employment. In improving tourism industry, it is necessary to develop an effort to balance out the potential demand and supply. Therefore, an accurate forecasting model is needed as the baseline of strategic resource planning as an effort to maximize the utilization and efficiency of the available resources. The objective of this research is to build an accurate forecasting model for Indonesian tourism demand. We use Gross Domestic Product (GDP), Consumer Price Index (CPI) and exchange rate from 5 major visitor countries of Indonesia as independent variable to predict Indonesian tourist arrivals number. Some of the main concerns in forecasting are non-linear relationship and high-fluctuation in data. By nature, tourism is relatively seasonal and fragile. Seasons of boom and lows are frequent, alarming the survival of industry players. We apply artificial neural network backpropagation as it has the ability to adapt to changes in input data. This character makes this method a convenient alternative to the econometrics and time-series forecasting models. We produce a forecasting model for monthly tourist arrivals in Indonesia. We reach an optimum configuration with single hidden layer and 31 hidden neurons.