Application of neural network for forecasting gas price in America

This paper presents a neuro-base approach for gas price forecasting of American consumers. in order to forming a neural network structure, effecting parameters on gas price are analyzed and gas production and consumption, import and export gas ,natural gas supplies held in storage,oil price are selected as inputs. this approach is structured as multi-level artificial neural net work(ANN)base on supervised muliti-layer perceptron (MLP),train with the levbergenberg-marquard algorithm .actual data from 1949-2010 is extracted from American energy information administration (EIA) .samples from 19492005 are used to train the multi-level ANN and the rest from 2005 to 2010 are used for network test. Result shows multi-level ANN is train well.