组合预测模型在地方财政收入预测中的应用

Application of a combination forecasting model in local financial revenue forecasting

根据地方财政收入预测受到多因素影响和经济系统具有非线性本质的特点,针对现有预测方法的不足,提出了一种组合预测方法。该方法首先通过灰色关联分析确定影响地方财政收入的主要指标,然后用灰色预测模型分别对各指标进行预测,最后将各指标的预测值作为输入,相应的地方财政收入实际值作为输出,训练并建立神经网络模型。实例分析表明灰色关联分析排除了非主要指标的干扰,灰色预测模型提供了较完善的输入数据,神经网络模型考虑了各主要指标的关联关系。实验结果证实该方法在地方财政收入预测中是有效可行的。

Local financial revenue indices are affected by many factors and their economic systems thus are characterized by nonlinear properties. Moreover, current forecasting methods have a few disadvantages. A combination forecasting model therefore was proposed. The methods include the followings steps. First, the main factors for local financial revenue were confirmed via gray correlation analysis. Second, the gray forecasting model, GM（1,1）, was applied to predict each index. Finally, the results of GM（1,1） were used as inputs and the actual data of relevant local financial revenue was used as outputs, then, a neural network was built. The results suggest that the gray correlation analysis can filter the accidental indices, the gray forecasting model can provide good input data sequences, and the neural network can process the relationships of indices. Experimental results demonstrated the availability and feasibility of the model in local financial revenue forecasting.