LME LEAD PRICE MODELING AND FORECASTING BY USING NEURAL NETWORK GMDH
Hossein Ostadi, Sayed Ehsan Jebelli · 2015
The objective of this study is to provide a model for prediction of Global price of lead in LME using GMDH neural network. GMDH is considered as an instrument with high capabilities to model complex dynamic non-linear systems. Identification of effective variables and elimination of unwanted variables from among independent variables affecting lead price is also another objective of the present research. In the present research, GMDH neural network method and technical analysis approach with emphasis on past behavior of effective variables were used for forecasting lead price. The data included 177 historical data resulted from a weekly average of lead price in LME during 1.1.2011 to 1.6.2014. The results of the research showed that GMDH model was able to forecast lead price with lower prediction error and higher tracing power based upon mean square error (MSE) and root mean square error (RMSE) and measurement of power of tracing (PT) criterion with 9 input variables and 5 latent effective layers from among 10 input technical variables of the model.