山东省花生年产量的组合预测模型研究  

Combined predictive model research of annual production of peanut in Shandong province

作　　者：张永强[1] 才正[1] 王刚毅[1] 

机构地区：[1]东北农业大学经济管理学院,黑龙江哈尔滨150030

出　　处：《广东农业科学》2014年第21期11-15,21,共6页Guangdong Agricultural Sciences

基　　金：黑龙江省自然科学基金(G201112);国家自然科学基金(71173035);中国博士后科学基金(20100480973);黑龙江省博士后科学基金(LBH-Z10211)

摘　　要：以山东省花生年产量为研究对象。针对花生年产量的强烈波动性而导致的预测难、准确率低等难题,提出了一种基于GM(1,1)和RBF神经网络的组合预测模型,利用GM(1,1)来捕捉花生年产量的总体趋势,RBF神经网络来预测带有强烈非线性的残差项;同时为了提高RBF神经网络的训练速度和精度,针对标准遗传算法存在的早熟现象和收敛速度慢的缺点,提出了一种改进的自适应遗传算法,对RBF神经网络的初始参数进行优化。试验结果表明,组合预测模型可以较准确预测花生年产量,说明了组合预测模型的可行性。This paper studies annual production of peanut in Shandong province. Considering the problem of difficult prediction and low accuracy due to strong volatility in peanut annual production, this paper proposes a novel combined model on the basis of GM （13） model and RBF neural network. GM （1,1） is to capture the global trend of peanut annual production, and RBF neural network is to predict the strong nonlinear residual item. To improve the training velocity and accuracy, considering the precocious phenomenon and slow convergence rate of standard genetic algorithm, a new self- adaptive genetic algorithm is proposed to optimize initial parameters of RBF neural network. Experimental results demonstrate the new combined model can accurately predict the peanut annual production, which shows the feasibility of this combined model.

关 键 词：花生产量预测 RBF神经网络 遗传算法 组合模型 

分 类 号：S565.2[农业科学—作物学] F326.12[经济管理—产业经济]