Forecasting The Export Of Chongqing Based On Artificial Neural Network

Because of advantaged domestic condition and global environment, the import and export trades in Chongqing have rapidly grown since recent years. Government cannot only obtain foreign source income by export trade to import advanced technology and lacking resources, but also receive profit by more opportunities of jobs offered by export corporations. The prediction of gross export is nonlinear and complex as the impacts of domestic and global factors upon export trade. This dissertation intends to make analysis of the structure of export products and markets and the characteristics of Chongqing's export trade, and then build two neural network models to forecast Chongqing's exports, in order to provide decision support for foreign trade to manage it scientifically and organize exporters efficiently.In the first, this dissertation analyses the structure of export products and markets. Exports can quicken the development of domestic economy because of the contacts between Chongqing and the world being more and more closed. Second, BP neural network models are built, as the system of export trade is non-linear and complex. BP neural networks can resolve non-linear problems effectively by using a sigmoid activation transfer function and a linear activation transfer function. At present, BP networks is widely used in the field of management and economy, example for stock, foreign exchange, GDP forecasting, decision-making of product costs and so on. Based on the historical data during 1987 and 2004, two valid and practical models are proposed for accurate forecast of the total exports of Chongqing. One is based on five common factors and the other is based on single time series. The predicted results agree with the actual data within reasonable experimental error. Finally, the prediction is analyzed in order to discover the trends of exports in Chongqing and be useful to government.