Comparison of ARIMA, neural networks and hybrid models in time series: tourist arrival forecasting

Abstract

For time series forecasting, different artificial neural network (ANN) and hybrid models are recommended as alternatives to commonly used autoregressive integrated moving average (ARIMA) models. Recently, combined models with both linear and nonlinear models have greater attention. In this article, ARIMA, linear ANN, multilayer perceptron (MLP), and radial basis function network (RBFN) models are considered along with various combinations of these models for forecasting tourist arrivals to Turkey. Comparison of forecasting performances shows that models with nonlinear components give a better performance.

Tourism demand modelling and forecasting using data mining techniques in multivariate time series: a case study in Turkey

“…One of the applied results of Kolmogorov's theorem for neural networks states that 2 hidden layers are enough for a certain approximation of any complex nonlinear function. In fact, usually 1 layer in a network is satisfactory in order to construct an approximation function [10,13,21]. There is no rule that indicates the optimum number of hidden neurons for any given problem.…”

“…(3)) is employed for output neurons, and the second one is used only for hidden neurons (Eq. (4)) [13].…”

“…The ANN method was introduced to tourism forecasting in the late 1990s, and later it was increasingly used to forecast demands for tourism [3,8,11,12]. The ANN seems to be a better method for tourist arrival forecasting than the conventional statistical methods [13].…”

Analysis of international debt problem using artificial neural networks and statistical methods

“…Artificial neural networks (ANN) are widely used in discrimination, sample recognition and estimation problems as an alternative of statistical models [1,7,17,33]. Economics, finance and business have great importance among the problems examined using ANN techniques [1,16,27,29,30].…”

Advanced Computational Approaches for Predicting Tourist Arrivals: the Case of Charter Air-Travel

“…These studies compare advanced computational approaches that have enhanced capabilities in modeling nonlinear characteristics (for example neural networks) with simple linear and stationary approaches such as the ARIMA models. Quite recently, hybrid ARIMA and simple static neural networks, as well as mixtures of static neural network models have also been found to perform better that classical time-series approaches (Aslanargun et al 2007). Regarding modeling of non-scheduled demand, previous work has applied regression models to predict charter international arrivals to major Greek airports and has highlighted that although there is uncertainty and variability in their evolution, historical data can be used to provide good predictions (Karlaftis and Papastavrou 1998).…”