水果市场价格预测与波动预警研究
发布时间：2019-05-13 07:55
【摘要】：水果是我国第四大作物类别。2013年,我国水果产量为25093万吨,实现总产值6969亿元。而我国人均鲜果消费仅为37.8kg,远低于发达国家水平,水果消费市场潜力巨大。但水果产业的发展也面临诸多挑战,比较典型的就是水果市场价格波动。这种波动直接影响了水果生产经营者的积极性,更是导致水果经营企业面临较大的经营风险。开展水果价格预测和波动预警对水果企业而言具有重要意义。因此本文以水果价格为研究对象,主要研究内容和结论如下:第一,开展水果价格预测。本文通过BP神经网络、SVM和ARMA模型对水果价格进行年度和月度预测。年度预测中,通过供给和需求指标构建了预测模型,SVM预测误差在5%和10%以内的分布优于BP神经网络;月度预测中,通过水果价格时间序列构建了预测模型,BP神经网络和ARMA对三种水果价格的预测误差基本在5%以内,而SVM均在1%以内。基于预测误差比较,最终确定SVM作为预测预警模型。第二,开展水果价格波动预警。基于预测分析,本文最终采用SVM对水果价格波动进行预警。本文以水果价格波动为警情指标,以均值和标准差的统计方法确定警度和警限,在水果价格预测基础上,通过SVM实现了对水果价格年度和月度波动预警。其中SVM对三种水果的年度预警准确率为64.70%,100%和94.12%,月度预警准确率为100%、91.67%和83.33%。第三,分析水果企业在水果价格波动背景下如何进行价格波动预警应对。基于水果价格预测和波动预警分析,本文认为水果批发企业应该强化自身的信息化建设,针对不同的警情和警度完善水果价格波动的预警预案,根据水果价格波动特征优化企业的价格策略,以此增强价格波动的风险应对能力。本文主要有以下创新点:(1)将计量分析应用到企业所处的宏观产业环境的分析中,增强了管理措施的针对性和有效性。(2)改进BP神经网络,通过多次运算取最优均值,提高了模型的稳定性和预测精度;(3)改进模型选取方法,将BP神经网络、SVM和ARMA模型对比选择,优化了预测预警模型的选择过程;(4)改进预警警限和警度的设置方式,以正向波动和负向波动分别统计,通过均值、标准差和单侧置信区间分别设定警限和警度,增强了预警警限和警度的可行性和科学性。
[Abstract]:Fruit is the fourth largest crop category in China. In 2013, China's fruit output was 250.93 million tons, achieving a total output value of 696.9 billion yuan. However, the per capita consumption of fresh fruit in China is only 37.8 kg, which is much lower than that in developed countries, and the potential of fruit consumption market is huge. However, the development of fruit industry is also facing many challenges, more typical is the price fluctuation of fruit market. This fluctuation directly affects the enthusiasm of fruit producers and operators, but also leads to greater business risks for fruit enterprises. It is of great significance for fruit enterprises to carry out fruit price prediction and fluctuation early warning. Therefore, this paper takes fruit price as the research object, the main research contents and conclusions are as follows: first, carry out fruit price prediction. In this paper, BP neural network, SVM and ARMA models are used to predict the annual and monthly fruit prices. In the annual forecast, the prediction model is constructed by the index of supply and demand. The distribution of SVM prediction error less than 5% and 10% is better than that of BP neural network. In monthly prediction, the prediction model is constructed by fruit price time series. The prediction error of BP neural network and ARMA for three kinds of fruit prices is basically less than 5%, while SVM is less than 1%. Based on the comparison of prediction error, SVM is finally determined as the prediction and early warning model. Second, carry out early warning of fruit price fluctuation. Based on the prediction analysis, this paper finally uses SVM to warn the fruit price fluctuation. In this paper, the fluctuation of fruit price is taken as the warning index, and the alarm degree and warning limit are determined by the statistical method of mean and standard deviation. On the basis of fruit price prediction, the early warning of annual and monthly fluctuation of fruit price is realized by SVM. The annual early warning accuracy of SVM for three kinds of fruits was 64.70%, 100% and 94.12%, and the monthly early warning accuracy was 100%, 91.67% and 83.33% respectively. Third, analyze how to deal with the price fluctuation of fruit enterprises under the background of fruit price fluctuation. Based on the prediction of fruit price and the early warning analysis of fluctuation, this paper holds that fruit wholesale enterprises should strengthen their own information construction and perfect the early warning plan of fruit price fluctuation according to different warning conditions and degrees of warning. According to the characteristics of fruit price fluctuation, the price strategy of enterprises is optimized to enhance the risk response ability of price fluctuation. The main innovations of this paper are as follows: (1) the econometric analysis is applied to the analysis of the macro industrial environment in which the enterprise is located, which enhances the pertinence and effectiveness of the management measures. (2) the BP neural network is improved and the optimal mean value is obtained by multiple operations. The stability and prediction accuracy of the model are improved. (3) the selection method of the model is improved, and the selection process of the prediction and early warning model is optimized by comparing the BP neural network, SVM and ARMA models. (4) the setting mode of early warning limit and alarm degree is improved, and the feasibility and science of early warning limit and alarm degree are enhanced by setting alarm limit and alarm degree by means of mean value, standard deviation and unilateral confidence interval, respectively.
【学位授予单位】：华中农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F323.7
本文编号：2475717
[Abstract]:Fruit is the fourth largest crop category in China. In 2013, China's fruit output was 250.93 million tons, achieving a total output value of 696.9 billion yuan. However, the per capita consumption of fresh fruit in China is only 37.8 kg, which is much lower than that in developed countries, and the potential of fruit consumption market is huge. However, the development of fruit industry is also facing many challenges, more typical is the price fluctuation of fruit market. This fluctuation directly affects the enthusiasm of fruit producers and operators, but also leads to greater business risks for fruit enterprises. It is of great significance for fruit enterprises to carry out fruit price prediction and fluctuation early warning. Therefore, this paper takes fruit price as the research object, the main research contents and conclusions are as follows: first, carry out fruit price prediction. In this paper, BP neural network, SVM and ARMA models are used to predict the annual and monthly fruit prices. In the annual forecast, the prediction model is constructed by the index of supply and demand. The distribution of SVM prediction error less than 5% and 10% is better than that of BP neural network. In monthly prediction, the prediction model is constructed by fruit price time series. The prediction error of BP neural network and ARMA for three kinds of fruit prices is basically less than 5%, while SVM is less than 1%. Based on the comparison of prediction error, SVM is finally determined as the prediction and early warning model. Second, carry out early warning of fruit price fluctuation. Based on the prediction analysis, this paper finally uses SVM to warn the fruit price fluctuation. In this paper, the fluctuation of fruit price is taken as the warning index, and the alarm degree and warning limit are determined by the statistical method of mean and standard deviation. On the basis of fruit price prediction, the early warning of annual and monthly fluctuation of fruit price is realized by SVM. The annual early warning accuracy of SVM for three kinds of fruits was 64.70%, 100% and 94.12%, and the monthly early warning accuracy was 100%, 91.67% and 83.33% respectively. Third, analyze how to deal with the price fluctuation of fruit enterprises under the background of fruit price fluctuation. Based on the prediction of fruit price and the early warning analysis of fluctuation, this paper holds that fruit wholesale enterprises should strengthen their own information construction and perfect the early warning plan of fruit price fluctuation according to different warning conditions and degrees of warning. According to the characteristics of fruit price fluctuation, the price strategy of enterprises is optimized to enhance the risk response ability of price fluctuation. The main innovations of this paper are as follows: (1) the econometric analysis is applied to the analysis of the macro industrial environment in which the enterprise is located, which enhances the pertinence and effectiveness of the management measures. (2) the BP neural network is improved and the optimal mean value is obtained by multiple operations. The stability and prediction accuracy of the model are improved. (3) the selection method of the model is improved, and the selection process of the prediction and early warning model is optimized by comparing the BP neural network, SVM and ARMA models. (4) the setting mode of early warning limit and alarm degree is improved, and the feasibility and science of early warning limit and alarm degree are enhanced by setting alarm limit and alarm degree by means of mean value, standard deviation and unilateral confidence interval, respectively.
【学位授予单位】：华中农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F323.7
【参考文献】
相关期刊论文 前6条
1 唐江桥;雷娜;;中国鸡蛋价格波动预警研究[J];西部论坛;2011年06期
2 李艳;刘军;;农产品价格预测系统设计与实现[J];湖北农业科学;2011年14期
3 王孝松;谢申祥;;国际农产品价格如何影响了中国农产品价格?[J];经济研究;2012年03期
4 杨tb;;基于结构方程模型的生猪价格预警研究[J];中国畜牧杂志;2011年18期
5 王桂红;杨勇;吴华瑞;刘向锋;;一种提高农产品市场价格预测精度的改进算法[J];浙江农业学报;2013年06期
6 张建华;孔繁涛;吴建寨;王盛威;朱孟帅;赵璞;;基于小波神经网络的辣椒价格时间序列预测模型[J];中国食物与营养;2014年12期
相关博士学位论文 前1条
1 汪晓银;中国柑橘市场预警研究[D];华中农业大学;2013年
，本文编号：2475717
本文链接：https://www.wllwen.com/weiguanjingjilunwen/2475717.html
