作　　者： (张兆同）; (余潜）;

机构地区： 南京农业大学工学院,江苏南京210031

出　　处： 《价格月刊》 2017年第9期31-36,共6页

摘　　要： 运用灰色关联分析法对影响棉花价格波动的诸多因素进行分析,筛选出了4个主要的影响因素:国际市场因素、替代品因素、居民消费价格指数和棉花进口量。并以此优化RBF神经网络模型的输入节点,验证了模型对棉花价格预测的精确性。采用2010年1月~2016年4月的月度数据作为网络训练集,将4个主要影响因素作为输入向量,经训练后网络拟合效果较好;以2016年5月~2017年4月共12期数据作为网络测试集,结果表明:模型预测误差为3.11%,预测精度较为理想,泛化能力强,模型能够较好地把握棉花价格变化的本质规律,为准确预测棉花市场价格提供参考。 This paper uses GILA method to analyze factors influencing the fluctuation of cotton price, and screens out 4 major impact factors： the international market factors,alternative factors, consumer price index and the ＇import of cotton. Based on the above, the RBF neural network model was optimized with this as input spot, and this paper verifies the accuracy of the model on cotton price prediction. The model is trained with monthly data from January 2010 to April 2016 as the training set, the four main factors as the input vector to improve the imitative effect. Then this paper uses the monthly data from May 2016 to April 2017 as the test set to testify the model, the results show that model prediction error is 3.11%, the prediction accuracy is ideal with strong generalization ability. The model can grasp the essential rules of cotton price variation, which provides reference to predict cotton market price.

关 键 词： 棉花价格 影响因素 神经网络 预测