基于PCA-SVM的区域经济预测研究

Research on Regional Economy Prediction Based on Support Vector Machines

针对区域经济存在高度的非线性、各指标间存在数据冗余等特征,使得传统的经济预测方法精度较低,为解决上述问题,提出了一个基于主成份分析的支持向量机的区域经济预测模型(PCA-SVM)。选择对影响长株潭区域经济发展的各因子进行主成份分析,消除各因子之间的冗余性,从而减少了支持向量机的输入维数,增加了支持向量机预测速度,利用PCA-SVM模型对长株潭2003-2007年经济数据进行了验证性测试和分析,结果表明,相对于各参比模型,PCA-SVM模型预测精度显著提高,是一种高效的区域经济预测模型,为预测研究提供依据。

There are multiple targets in regional economic and indictors in high degree of nonlinear and redundancy.The traditional modes have low accuracy.In order to resolve the problem,a hybrid forecasting method of principal component analysis based support vector machines（PCA-SVM） is presented for regional economic tendency forecasting.In this hybrid approach,PCA is used for variable selection in order to reduce the model complexity of SVM,eliminate redundancy in the various factors,reduce the input data,and improve the speed of the support vector machines.Using the hybrid forecasting model to test and analysis the regional economic data in Changsha,Zhuzhou and Xiangtan,the empirical results reveal that PCA-SVM model has outstanding forecasting ability and great superiority in predicting accuracy compared with the reference models.