THE COMPARISON OF ROBUST ARIMA MODEL AND ARTIFICIAL NEURAL NETWORK MODEL: AN EXAMPLE OF TOURISM

Artificial neural network is an analysis procedure which became popular after the first half of twentieth century. Having many successful applications in areas such as Pattern Recognition, Sound Analysis and etc., the artificial neural network procedure has also been used in Time Series Analysis. Nevertheless, the factors that effect the success of time series analysis are different than the factors of success in the areas we mentioned previously and factors such as estimation of the time series’ parameters and the existence of breaking points of time series can effect the forecasts of the model. That’s why in our study we seek to develop the artificial neural network procedure. Robust Statistical Methods in time series analysis, have developed effective time series analysis procedures even though there can be a small amount of data. One of those methods is the modified maximum likelihood estimation method. In this method using any time series model, the forecasts are obtained usingmodified maximum likelihood estimates of parameters and using the distribution of residuals we can obtain unbiased and efficient estimators. In this study, we will compare the modified maximum likelihood method and the enhanced artificial neural network procedure on a tourism data