基于熵权的年降雨量预报优化组合模型研究

鉴于单一预测模型在建模时预测值比实际值存在较大偏差问题,为了提高预测精度,在此首先采用自回归综合移动平均ARIMA模型（简称A模型）、Elman神经网络模型（简称B模型）、小波网络分析模型（简称C模型）、灰色系统GM（1,1）模型（简称D模型）,利用广西田东县1990～ 2007年的年降雨量分别进行了模拟计算,然后在各单一模型预测（拟合）的年降雨量偏差值基础上,应用熵权法对4种模型的偏差值进行客观赋权后优化组合,并根据最优组合结果,选用A、B、C单一模型和最优选的A-B-C优化组合模型对广西田东县2008～ 2010年的年降雨量进行预测对比.结果表明,A、B、C和A-B-C模型得到的均方根误差RMSE和模型效率EF分别为0.018、0.015、0.017、0.013和0.817、0.877、0.843、0.897,优化组合模型的预测精度和拟合度比单一模型的结果得到了提高和改善,该组合方法提高了年降水量的预测精度,为诸如广西田东县以雨养农业为主的区域农业干旱预报提供了新的方法和依据.

Optimized Combined Model Prediction of Annual Precipitation Based on Entropy Weight

In view of the fact that a single prediction model＇ s predictive value exists the greater deviation than the actual value,in order to improve the accuracy of the predicted value,this study first adopt the autoregressive integrated moving average ARIMA model （referred to as the A model）,Elman neural network model （referred to as the B model）,the wavelet network analysis model （referred to as the C model）,gray system GM （1,1） model （referred to as the D model） to analyze and forecast respectively utilize 1990 ～ 2010 annual rainfall in Tiandong County,and then on the basis of the annual rainfall deviation value in every single model forecast （fitting）,adopt entropy theory to four model prediction deviation value carrying on objective weighting,and thus to optimize the combination.And in accordance with the optimal combination results,then adopt A model,B model,C model and the most preferred A-B-C optimization combination model respectively to forecast the 2008 ～2010 annual rainfall of Tiandong County,The results show that the root mean square error（RMSE） and model efficiency （EF） of A,B,C and A-B-C model are 0.018,0.015,0.017,0.013 and 0.817,0.877,0.843,0.897 respectively,it can see that the prediction accuracy and goodness of fit of optimization combination model have been upgraded and improved than the single model results,the combination method improve the accuracy of the annual precipitation forecast,and provide a new method and basis for regional agricultural drought prediction such as Guangxi Tiandong County based on mainly rain-fed agriculture.