Forecasting Coke' Price In Short-term By Semi-parametric Regression

Because there is existed some factor influenceing coke' price in short-term,so it is diffcult to forecast coke' price in short-term.The target of dissertation is to discover the effective way to forecast coke' price in short-term.Based on the characteristics of consumption of coke, this dissertation establishes the modeling of semi-parametric linear regression and non-linear regression of coke' price estimation by giving the data of coke' consumption market from January 2001 to April 2009 and utilizing the method of regression analysis so as to raise estimation precision of short-term coke price. And then, this dissertation suggested utilizing error correction operator (ECO) to replace the parametric sectors based on the analysis of cointegration theory.The semi-parametric regression combined with error correction operator was constructed.At last, the dissertation suggested utilizing artificial neural network to find relationship between coke' price and iron' price. The semi-parametric regression combined with BP artificial neural network was constructed.The selection of optimum bandwidth will be using cross validation method, the Parabola kernel will be selected in Kernel function. Finally, the least squares estimation will be selected in semi-parametric modeling estimation.The estimation results of real case is demonstrated that the semi-parametric regression model is effective method to forecast coke' price in short-term.