Research On Economic Modeling And Predictive Method

The reform and opening in Guangdong Province, especially in the Pearl River Delta region, has made great achievements in economic development in the past of thirty years. The environmental, industrial upgrading and other issues occur with the development of economy. How to maintain sustained economic growth is one of most important issues to be studied and resolved for the state and local governments. Under the great support of the project "the Pearl River Delta Economic Development Strategy" from Guangdong Province Development and Reform Commission, the thesis focuses on the establishment of the economic development model for the forecast planning in Guangdong and, and applies the intelligent control and prediction techniques to predict the future economic development of Guangdong Province. The results might provide a scientific basis for decision making for the development and Reform Commission in the formulating economic development plans. This thesis firstly discusses the history of the macroeconomic development, and then systematically analyzes the classical economic theory, the key elements of the Keynes’s state intervention theory, and the market regulation neo-classical economic theory. After analyzing the advantages, disadvantages and limitations of the existing methods, and the key elements of the economic growth model, we study the regional economic development from the point view of the control theory. After sufficiently studying the history and current status of the modeling of the domestic and international macroeconomic and predictive control, we present intelligent modeling and forecasting method for the modern economic nonlinear model with large-scale, globalization, time-varying uncertainty. Combining the structure of self-organizing neural networks with the reinforcement learning method, a novel economic growth model is presented to forecast the economy of Guangdong Province and to provide reference value for Guangdong Province in the future economic development planning for the Pearl River Delta region. The main results of the thesis are listed as follows:(1) Considering the large export-oriented economic components in Guangdong, it is fact that the economic operation is a nonlinear complex and time-varying uncertain system. Based on a self-organized growth neural network model, we present an economic growth prediction method. We also assess the predict effect of neural network model. The method could provide a helpful idea for the establishment and the improvement of the economic model. Based the proposed model, we predict the future level of consumption and the trend of economic development in Guangdong Province. The prediction provides a scientific basis for government regulation on the economic growth.(2) An improved network algorithm is presented for to overcome the problem of local extreme points for the BP artificial neural growth model in the economic growth prediction. The adaptive learning rate of the additional momentum in the neural network training is used to reduce network training time and number, and to improve the efficiency of the network training. Using the chain data reorganization method, the training sample set is expanded to solve the problem of few year data in the macroeconomic forecast. Usually, the trained neural network model has poor generalization ability in the economic data predict for the growth macro-economic indicators in the context of China’s economic development. We introduce the growth rate data of the economic indicators and the series data of the time window to improve the generalization ability of the proposed neural network model. The simulation results show that the proposed algorithm could enhance the generalization ability of neural network and improve the prediction accuracy of the system model.(3) An improved immune particle swarm optimization neural network algorithm is presented to overcome the problem of slow convergence, low accuracy and local convergence. An immune particle swarm optimization algorithm is used to optimize the neural network. A linked data reorganization method is introduced to expand the sample set. This method could satisfy the requirements of the economic system modeling, and greatly strengthen the network generalization capabilities. By combining with the economic prediction algorithm, the method improves the prediction accuracy of the system model. The simulation results show the prediction error is dropped from15%obtained by the original BP neural network to5%. The model is applied to adjust the total social investment and per capita consumption level. The simulation results show that the consumption increasing is the intrinsic motivation of the economic development.(4) We propose an improved optimization algorithm for the SVM economic forecasting. LS-SVM model is used to mode and predict the GDP in the macroeconomic model. The LS-SVM model parameters is optimized by using multi-scale chaotic genetic optimization. The precise forecast for the GDP and other economic indicators is great meaningful for the economic development planning. A time function of the economic forecast indicators is introduced as the model input. Simulation results show that the proposed algorithm could improve the prediction accuracy, with fast convergence and great generalization ability. The average error rate is small than2%while the error rate yielded by the BP neural network is25%. (5) Considering the uncertainties which may raise from the population flow, management innovation, and government regulation, we develop a dynamic optimization structure model to forecast the economic development of Guangdong Province. This model introduces a competitive mechanism to automatically adjust the structure of the neural network and further optimizes the weight allocation adjustments. This method can overcomes the larger approximation errors in the model structure of the fixed networks and improves the prediction control performance. Simulation on the economic forecasts for Guangdong Province is carried out in the theise. The simulation results show that the proposed dynamic optimization algorithm with neural network structure could achieve a satisfying prediction in the economic forecast.Finally, with the great development of the market economy, the volatility characteristics of effects China’s economic operation. We studies the problm of macroeconomic control and prediction of the Pearl River Delta region. Based on the neural network intelligent algorithm, we propose improved control performance and forecasting methods. This methods is a meaningful attempt for the predictive control of the economy and have strong theoretical values and practical significance.