Predicting short-term coke price by neural network—semiparametric model

There are many factors that influence the short-term prices of coke. And the relations between those factors are difficult to be analyzed quantitatively. So, there are big errors in predicting short-term coke prices by the Common Industrial Prediction method and Parametric Regression Method. In order to improve the accuracy of the prediction of the short-term coke prices, an innovative Semiparametric Regression Method was applied in this Article. The neural network—semiparametric model was built by taking the functional relation, which was obtained through the neural network training, as the parametric part and the price of cast iron as the nonparametric part of the semiparametric model, thus to create a neural network—semiparametric regression model. Sample estimates demonstrates that neural network—semiparametric model is not only reduced the boundary estimation error,but also increased accuracy. It is an effective tool for prediction of the short-term coke price.