对数据进行建模预测分析时，较多采用单个模型，而单个模型难以全面反映数据的变化规律。为发挥单个模型自身优势，利用组合原理将单模型组合形成组合预测模型，以提高预测精度。组合模型中常采用线性组合方法，然而被组合模型拟合值与原始数据不具有线性关系时采用该方法效果较差。利用神经网络的高度非线性拟合能力，构建BP神经网络的非线性组合模型，并应用到我国节水灌溉面积年度数据预测上。实证表明，非线性组合预测模型精度优于单模型及基于最优加权的线性组合预测模型。%Single model is usually used for modeling projections of data analysis,but the single model is difficult to fully reflect the variety regulation of the data. To make full use of single model itself ,single models are cimbined by using the combination principle to form combination forecast model,in order to improve the prediction precision. When the value of a portfolio model fitting with the original data has no linear relationship ,the effect of using the linear combination method is poorer. Making use of the neural network nonlinear fitting capability ,we built nonlinear combination model of BP neural network,and applied to the water saving irrigation area in our country on the annual data to predict. The empirical shows that the precision of nonlinear combination forecasting model is superior to the single model and the optimal weighted linear combination forecast model.
展开▼
