【作者】 葛慧玲；

【导师】 葛家麒；

【作者基本信息】 东北农业大学 ， 生态学， 2007， 硕士

【摘要】 在加入世贸组织之后，我国的农产品市场逐渐开始融入国际市场。而大豆作为我国最重要的油料作物之一也承受着国际市场的压力。因此，建立大豆市场的预测及预警体系对我国大豆产业的健康发展有着极为重要的意义。本文在经过对我国大豆现状进行深入调查的基础上，采用BP神经网络及改进Logistic模型的方法，对大豆的产量、需求量、进口量进行了预测。对我国大豆市场的两个主要警情指标——大豆进口依存度和大豆价格增长率分别建立了相应的预警模型。最后结合预警的结论给出了相应的意见与建议。具体的研究内容包括以下几个方面：1．对我国的大豆生产状况进行了分析。论述了近年来我国大豆的单产、种植面积及总产量的变化情况，对我国的大豆主产区的大豆种植情况进行了分析。在此基础上，建立了大豆总产量的BP神经网络预测模型，并在神经网络隐层节点的数目选择上采用了灰关联度优化分析的方法，通过计算隐层各节点与输出之间的灰关联度对隐层节点的数量作出判断，最终建立了4-13-1结构的BP神经网络模型。并通过了精度检验。利用此模型对我国大豆2006-2008年的大豆单产、种植面积和总产量进行了预测，预测结果显示，在未来的三年，我国的大豆生产情况基本保持平稳，没有太大的增长。2．对我国大豆的供需情况进行了分析。结合国外大豆的生产情况，对近几年我国的大豆供给量、进口量以及需求量的变化情况进行了分析，讨论了需求量急剧增加的原因。在进口量的预测上，建立了改进Logistic模型，在模型中加入了增长率影响因子，最终建立了大豆市场进口量预测模型并通过检验，效果良好。通过此模型，对我国2006-2008年的大豆进口量进行了预测，并用同样的方法建立了大豆需求量的改进Logistic模型。预测的结果显示，在未来的三年中，我国的大豆进口量将继续增加，大豆需求量则保持平稳的增长。3．结合大豆市场的实际情况，选择了大豆进口依存度和大豆价格增长率作为预警的警情指标，并分别选择了相应的警兆指标，利用时差相关系数方法确定了警兆指标的先行、同步和滞后性，并对指标进行了相应的简化与筛选。4．在进口依存度预警中，分别建立了基于综合模拟法的中国大豆市场预警信号灯系统模型、基于合成指数的大豆市场预警系统模型和大豆市场预警趋势外推模型，在趋势外推模型中建立了基于先行警兆指标的多元线性回归模型，并用两阶段最小二乘法对模型进行了修正。最后比较了三种方法的特点与不足，并求得了最终的警度。结合以上的三种方法，可以预警出2006-2008年我国的大豆进口依存度分别处于重警、重警、重警的警区。5．在大豆价格增长率预警中，尝试了对状态空间法的改进，将状态空间法与景气循环法结合，使得该方法更为合理。利用该方法实现了对2006-2008年我国大豆价格增长率的预警。通过改进状态空间法，可以计算出未来三年我国大豆价格增长率分别处于轻警、中警、中警的警区。6．结合预测及预警的结果，对我国未来的大豆市场构建提出了相应的意见与建议，作为我国大豆市场健康发展的依据。更多还原

【Abstract】 After entering into the World Trade Organization, domestic farm produce market graduallymelt into the international market. Thus as one of the most important oil plants, the soybean alsoreceives the press from the international market. Consequently, it is an important meaning to buildthe forecast-warning system of soybean market for domestic soybean industry. In this paper,undergoing to survey the actuality of domestic soybean product, the output, demand and import ofthe soybean are forecasted using the method of the BPNN and the modified Logistic model. Twoimportant warning alert index of domestic soybean market, that is the dependence degree of thesoybean import and the increasing ratio of the price of the soybean, are built the forecast-warningmodel. In the end, suggestions corresponding to the early warning conclusion are given. Theresearch is consisted of the following aspects:1.In this paper, the production status of domestic soybean is analyzed. The recently changecases such as the unit production, the planting area and the total production of domestic soybean,are illustrated and the planting circs change of the soybean in domestic main soybean produce zoneis analyzed. On this base, the forecast model of the BPNN about the total production of the soybeanis bulit and the optimization analysis method of the degree of grey association is used whenselecting the number of the hidden nodes of the NN. As a result, the model of the NN with thestructure in 4-13-1 is produced and is correctly experimented by counting the degree of greyassociation between the hidden nodes and its output to estimate. Applying this model to forecast theunit production, the planting area and the total production of domestic soybean from 2006 to 2008,its result is revealed that the production status of domestic soybean keeps stable in the subsequentthree years.2.The status of the supply and demand of domestic soybean is analyzed. The recently changecases such as the supply quantum, the import quantum and the demand quantum, are illustratedassociated with external cases and the cause of the rapidly increasing of the demand quantum isdiscussed. About the forecast of the import quantum, the modified Logistic model is developed, inwhich the effect factor is added.Then the forecast model of the import quantum is provided,subsequently experimented and obtained a good effect. Using this model, the import quantum isforecasted from 2006 to 2008 and the modified Logistic model of the demand quantum isdeveloped by the same way. Its result is revealed that the import quantum keep gradually increasing and the demand quantum keeps stably increasing.3.Combined with the actual status of the soybean market,the import dependence degree andthe increasing ratio are referred as the warning alert indexes and the corresponding warning signindexs are selected. The foregoing, synchronous and lag properties of the warning alert indexes aredetermined and the indexes are predigested and selected.4.In the forecast-warning of the import dependence degree, three model are developed, whichare the warning signal lamp system model based on the composition simulation method,theforecast-warning system model based on the composition index and the multi-index tendencyextrapolation model of domestic soybean market. In the multi-index tendency extrapolation model,the multi-variable linear regression model is based on the foregoing warning alert index andmodified using the least square method with two steps. In the end, the warning degree is obtainedby comparing the advantage and disadvantage of three methods. Therefore, the import dependencedegrees highly warning and extensively highly warning fron 2006 to 2008.5.In the forecast-warning of the increasing ratio of the price of the soybean, the state-spacemethod is tried to modify, namely the state-space method is combined with the boom-cycle methodsuch that this way is more suitable.The forecast warning of the increasing ratio of the price from2006 to 2008 is realized by the model. The gentle, medium and medium areas with the increasingratio of the price of the soybean are operated in the subsequent three years.6.The suggestions and the ideas are advanced for constructing domestic future soybean marketwith the result of the forecast and the warning, and referred as the gist of the health development ofthe soybean market.更多还原