Study Of Artifical Neural Network-Based Economic Forecasting

Based on existing studies of economic forecasting methods, an Artificial NeuralNetwork(ANN)-based algorithm is developed to make economic forecasting morerealistic in this thesis. Through selection of the appropriate ANN algorithms aneconomic forecasting model is first established in this thesis. Then, the model is usedto carry out a case study based on real world data acquired from ShenZhen.First, this thesis summarizes and studies characteristics, classification, methods,steps and error-checked in the economic forecasting. It is revealed that the traditionalmethods are becoming unfit for the modern management need, because of theeconomic management is going to be even more complicated nowadays. For thecomplexity of the economic system and the good performance of the ANN, theANN-based economic forecasting has been found encouraging in this thesis.Secondly, in this thesis an ANN algorithm is worked out, and the support vectormachine algorithm based on an idea of structural risk minimization is presented indetails. Then three types of commonly used forecasting models made from ANN areinvestigated and compared to choose the best one for forecast.Thirdly, this thesis has made a close look into three economic increment modelscommonly used in economics and their features, making sure that some internalunknown relations in an economic system can be simulated by ANN. As a result, thehypotheses for the economic models can be lessened and the forecasting work can bedone with even better agreement with the fact.Finally, an ANN-based economic forecasting model is established and a casestudy is carried out using acquired data from ShenZhen. The result from the casestudy shows that the ANN-based economic forecast method is useful and effectiveand can be a reference for decision-makers of economics.