MODELLING AND FORECASTING GOLD PRICE USING GMDH NEURAL NETWORK
Hamideh Moradi, Iman Jokar, Ahmad Forouzantabar · 2015
Forecasting the price of gold and its changes as an economic event has long been within the interest area of investors and financial analysts. This study aims to make gold -price forecast modelling using GMDH neural network and Multilayer Perceptron neural network (MLP), as well as determination of top model using performance evaluation criteria. Necessary data of the research was collected through Internet from Central Bank website (www.cbi.ir) and in particular from Novin Rah Avard [modern approach] software from 1386 [2007] till1391 [2012]. In this research, weekly means of eight effective indexes on gold price were selected for the mentioned interval. Good and acceptable results of performance evaluation criteria obtained from GMDH neural network indicate high capability of this network for identification of ruling patterns on data as well as unique characteristics of fast convergence, high accuracy, and approximation ability of strong function of the network confirming suitability of GMDH neural network for prediction of gold price.