改进证据理论融合灰色神经网络的水质预测

摘要

针对水质预测过程中样本数据少的特点，引入了改进证据理论和灰色神经网络相结合的组合预测方法。首先利用灰色神经网络作为单一模型对水质进行初步预测，再用神经网络对预测结果进行分析建模，得到每个单一预测模型的可信度，最后采用改进证据理论进行融合决策，以获得各单一预测模型的权重，从而实现了水质的组合预测。实例分析结果表明，该方法拟合误差小、预测精度高。%Aiming to the characteristics of few sample data in water quality prediction process,this paper introduced a combination prediction meth-od based on improved evidence theory combining with gray neural network. Firstly,the gray neural network was used as single model to preliminary predict the water quality;then the neural network was used to analyze the predictive results and establish models to get the credibility of every single prediction model;finally,the evidence theory was employed to fuse them and to obtain the single prediction mode1 weight. Hence,the water quali-ty prediction was realized. The results show that the method has small fitting error and high prediction precision.