A hybrid GMDH and Box-Jenkins models in time series forecasting
Ani Shabri, Ruhaidah Samsudin · Applied Mathematical Sciences · 2014
The group method of data handling technique (GMDH) and Box-Jenkins methods are two wellknown time series forecasting of mathematical modeling. In this paper, we introduce a hybrid modeling which combines the GMDH method with the Box-Jenkins method to model time series data. The Box-Jenkins method was used to determine the useful input variables of GMDH method and then the GMDH method which works as time series forecasting. The lynx series contains the number of lynx trapped per year is used in this study to demonstrate the effectiveness of the forecasting model. The results found by the proposed GMDH were compared with the results of Box-Jenkins and artificial neural network (ANN) models. The comparison of modeling results shows that the GMDH model perform better than two other models based on terms of mean absolute error (MAE) and root mean square error (RMSE). It also indicates that GMDH provides a promising technique in time series forecasting.