人工神经网络处理多变量关系的应用分析

分析了常规方法处理多变量关系时存在人为误差较大、基函数选择及系数求解较困难的问题，通过实例对人工神经网络模型处理多变量关系的模拟及预测结果进行分析，并与常规方法的结果进行对比，指出：运用人工神经网络模型建立降雨径流关系，能较好地反映流域的水文过程，且参数具有明显的物理意义；模型能够应用于流域降雨径流预报，且预报效果良好；人工神经网络能够有效地处理多变量之间复杂的非线性关系，当因变量与自变量间确实存在较密切的联系时。所建立的多变量间的关系具有较高的精度。

Application and Analysis on Multivariate Relation Handled by Artificial Neural Network

The paper analyzes the existing bigger human error and more difficult in primary function selection and coefficient resolution in handling multivariate relations by a conventional method.It analyzes the results of simulation and prediction of multivariate relations handled by artificial neural network through cases and compares with the results of using a conventional method.It points out that the rainfall-runoff relationship established by artificial neural network model can fairly reflect hydrologic process and parameters have remarkable physical significance;the model can be applied into rainfall-runoff forecasting of a river basin and the results are fairly good and;the artificial neural network can effectively handle the complicated nonlinear relations between multivariate and the established relations between multivariate have higher precision when a closer relationship is really existed between variable and independent variable.