链式数据重组与神经网络在经济预测中的应用  

Application of the chain style data recombination methodand neural networks in macroeconomic forecast

作　　者：郅跃茹[1] 朱维彰[2] 诸静[1] 

机构地区：[1]浙江大学电气工程学院,浙江杭州310027 [2]杭州电子工业学院自动化系,浙江杭州310037

出　　处：《控制理论与应用》2004年第4期643-645,共3页Control Theory & Applications

摘　　要：建立经济模型和基于模型对宏观经济进行预测,是经济运行质量评价、仿真、制定发展规划等所必不可少的.针对宏观经济预测的特殊性:样本少、时变性,提出了反向传播(BP)神经网络的链式数据重组训练方法,并用于实际经济预测.和原数据用于预测的结果相比,达到了较高的预测精度.同时,解决了BP神经网络难以确定隐结点数的问题.It is necessary to model predict and model the economic system and for evalusating economic circulation,simulation,and making development programs.Based on the characteristics of the macroeconomic forecast,such as small data sets and time_varying,a chain_style data recombination method is presented,which is used for the back_propagation (BP) neural networks training.This method was applied to real macroeconomic forecast,and achieved the higher forecasting precision than the original data sets.It also solves the problem how to decide the hidden node numbers for the BP neural networks.

关 键 词：宏观经济预测 数据重组 神经网络训练 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]