Research on the forecasting model of China’s rare earth export prices based on BP neural network

Qing Guo1, Lu Zhang2*, Jing Sh

Abstract

Rare earth products have an extensive use in the realm of cutting-edge technology and military engineering. China keeps the largest proportion of rare earth reserve and production all over the world. Whereas, it fails in obtaining the pricing power in spite of its monopolistic status in terms of rare earth volume and even has been prosecuted several times for the reason of export price. Using related export data of China’s rare earth products in the U.S.A market from Jan. 2000 to Dec. 2011, based on different number of hidden cells, this paper establishes several BP natural network models aiming at forecast the export price. In addition, selecting five predicting error indices, involving Mean Absolute Error, Mean Absolute Percent Error, Root Mean Squared Predict Error, Normalized Mean Square Error and Mean Square Percentage Error, this paper compares the forecasting precision of established models. The research indicates that the predicting error is turned out to be lowest and forecasting precision is relatively higher when the number of hidden cells is five. Therefore, this model could be used to forecast the change of export price accurately and to provide scientific basis for decision makers.