基于支持向量机的福建省GDP预测研究

采用1981～2002年的福建省GDP数据作为支持向量机(SVM)的训练目标,以各期前三年的GDP作为输入向量构成训练样本。首先利用格子搜索法获得支持向量机模型中的参数(C,γ,ε)对样本进行训练。然后用训练所得模型对2003、2004、2005三年的福建省GDP进行测试,平均测试精度达98.12%。可以认为支持向量机具有较强的泛化能力,在宏观经济预测中具有较高的精度,从而可用于未来实际GDP的预测。

Research on Forecasting of GDP in Fujian Province Based on Support Vector Machine

Constitute the training sample by using the GDP data from 1981 to 2002of Fijian province as training target, and using the GDP data of the three years before each period as input vector. First, using the Parameters(C,γ,ε)of SVM model which obtained by grid-search to train the sample, then, using the model obtained from training to test the GDP data of Fijian province from 2003 to 2005, and the average accuracy can be as high as 98.12%. It is considered that support vector machine has good Generalization Ability, and has high accuracy in Macroeconomic prediction. Therefore, it can be used to forecast future trend of GDP data.