基于时间序列SVR模型的玉米价格预测研究

doi:10.11924/j.issn.1000-6850.casb20191200938

摘要/Abstract

摘要：

研究旨在运用时间序列和支持向量回归(support vector regression,SVR)理论,建立短期玉米价格预测模型,为玉米市场监测预警提供技术支持与决策依据。本研究根据玉米价格序列波动的非线性特征,以河南省2010年1月1日—2019年6月30日的月度数据为研究对象,结合混沌性时间序列与SVR理论,研究一种短期玉米价格预测模型。通过相空间重构的方法对价格序列进行处理,利用重构后的数据训练模型,运用网格交叉验证(GridSearchCV)对研究模型进行优化。得到一种基于时间序列支持向量回归的玉米价格预测模型。将研究模型的预测结果进行对比分析,结果显示,研究模型的预测值更贴近真实值,其平均相对误差(MRE)和均方根误差(RMSE)分别为0.006和20.57,优于传统支持向量回归模型的预测结果。研究模型可以为玉米价格预测提供方法依据,且相较于传统预测方法在精度方面具有优势。

关键词: 玉米价格, 预测模型, 支持向量回归, 相空间重构, 时间序列

Abstract:

The study aims to establish a short-term corn price forecasting model by using time series and support vector regression (SVR) theory, to provide technical support and decision-making basis for corn monitoring and early warning. Based on the non-linear characteristics of corn price series fluctuations, this study took the monthly data of Henan Province from January 1, 2010 to June 30, 2019 as the research object, and combined a chaotic time series with SVR theory, to study a kind of short-term corn price prediction model. The price sequence was processed by the phase space reconstruction method, the model was trained with the reconstructed data, and the grid cross validation (GridSearchCV) was used to optimize the research model. A corn price prediction model was obtained based on time series support vector regression. By analyzing the prediction results of the research model, it is showed that the prediction value of the research model is closer to the true value, and its average relative error (MRE) and root mean square error (RMSE) is 0.006 and 20.57, respectively, which are superior to the prediction results of the traditional support vector regression model, with more accuracy.

Key words: corn price, prediction model, support vector regression, phase space reconstruction, time series

中图分类号:

张宝文, 王川, 杨春英, 王来刚. 基于时间序列SVR模型的玉米价格预测研究[J]. 中国农学通报, 2020, 36(31): 115-120.

Zhang Baowen, Wang Chuan, Yang Chunying, Wang Laigang. Corn Price Prediction Based on Time Series SVR Model[J]. Chinese Agricultural Science Bulletin, 2020, 36(31): 115-120.

图/表 6

图1. 河南省2010年1月—2019年6月玉米价格走势

名称	表达式
线性核	$κ (x, y) = x T y + c$
多项式核	$κ (x, y) = (a x T y) d$
径向基核函数	$κ (x, y) = exp (- γ x - y 2)$
Sigmoid核	$κ (x, y) = tan h (β x T y + θ)$

名称	表达式
线性核	$κ (x, y) = x T y + c$
多项式核	$κ (x, y) = (a x T y) d$
径向基核函数	$κ (x, y) = exp (- γ x - y 2)$
Sigmoid核	$κ (x, y) = tan h (β x T y + θ)$

表1. 核函数种类

图2. 基于时间序列SVR模型预测结果

图3. 基于时间序列SVR模型与传统SVR模型预测结果对比图

序号	真实值/(元/t)	传统SVR模型		GSCV-SVR模型		基于时间序列SVR模型
序号	真实值/(元/t)	预测值/(元/t)	相对误差(RE)	预测值/(元/t)	相对误差(RE)	预测值/(元/t)	相对误差(RE)
1	1685.000	1663.000	0.013	1682.000	0.002	1688.833	0.002
2	1667.778	1698.000	0.018	1670.000	0.001	1725.291	0.034
3	1620.000	1683.880	0.039	1665.374	0.028	1620.100	0.000
4	1628.182	1634.861	0.004	1619.661	0.005	1627.169	0.001
5	1658.214	1644.497	0.008	1636.758	0.013	1721.461	0.038
6	1726.875	1675.387	0.030	1665.939	0.035	1727.007	0.000
7	1757.500	1746.051	0.007	1736.281	0.012	1755.612	0.001
8	1827.679	1776.178	0.028	1759.526	0.037	1827.049	0.000
9	1870.000	1848.683	0.011	1841.306	0.015	1872.285	0.001
10	1870.000	1890.638	0.011	1882.401	0.007	1872.285	0.001
11	1850.000	1889.026	0.021	1873.450	0.013	1849.547	0.000
12	1850.000	1868.108	0.010	1847.418	0.001	1849.547	0.000
13	1845.789	1868.875	0.013	1851.522	0.003	1846.541	0.000
14	1840.000	1864.470	0.013	1846.071	0.003	1840.310	0.000
15	1868.333	1858.570	0.005	1839.461	0.015	1868.104	0.000
16	1916.667	1888.437	0.015	1877.597	0.020	1935.279	0.010
17	1901.765	1937.917	0.019	1936.164	0.018	1903.316	0.001
18	1884.286	1920.289	0.019	1905.228	0.011	1909.223	0.013
19	1880.000	1902.665	0.012	1885.417	0.003	1900.987	0.011
20	1889.444	1898.911	0.005	1884.237	0.003	1911.900	0.012
21	1890.000	1908.964	0.010	1896.717	0.004	1911.738	0.012
22	1906.667	1909.156	0.001	1895.982	0.006	1907.163	0.000
23	1947.895	1926.560	0.011	1916.983	0.016	1950.488	0.001

表2. 玉米价格预测值与真实值对比

评价模型	MRE	RMSE
传统SVR模型	0.014	29.436
GSCV-SVR模型	0.012	28.085
基于时间序列SVR模型	0.006	20.574

表3. 模型预测效果评价

参考文献 27

[1]	王东杰. 大数据视角下的粮食安全预警研究[D]. 北京:中国农业科学院研究生院, 2017.
[2]	许世卫. 农业大数据与农产品监测预警[J]. 中国农业科技导报, 2014,16(5):14-20.
[3]	许世卫. 中国农业监测预警的研究进展与展望[J]. 农学学报, 2018,8(1):203-208.
[4]	李哲敏, 任育锋, 彭春燕. 大数据时代的澳大利亚农业信息监测预警体系[J]. 世界农业, 2015(10):27-32,245.
[5]	Moore, Henry Ludwell. Forecasting the Yield and the Price of Cotton[M]. New York: The Macaillan Company, 1917: 21.
[6]	Jadhav V, Reddy BVC, Gaddi G M. Application of ARIMA Model for Forecasting Agricultural Prices[J]. Journal of Agricultural Science and Technology, 2017,19(5):981-992.
[7]	Puchalsky Weslly, Ribeiro Gabriel Trierweiler, daVeiga Claudimar Pereira, et al. Agribusiness time series forecasting using Wavelet neural networks and metaheuristic optimization: An analysis of the soybean sack price and perishable products demand[J]. International Journal of Production Economics, 2018,203:174-189.
[8]	Mitra Dipankar, Paul Ranjit Kumar, Paul A K. Statistical modelling for forecasting volatility in potato prices using ARFIMA-FIGARCH model[J]. Indian Journal of Agricultural Sciences, 2018,88(2):268-272.
[9]	程贤禄. 北京市农产品批发市场蔬菜价格预测预报体系研究[J]. 北京农业科学, 2002(2):1-10.
[10]	李哲敏, 许世卫, 崔利国, 等. 基于动态混沌神经网络的预测研究——以马铃薯时间序列价格为例[J]. 系统工程理论与实践, 2015,35(8):2083-2091.
[11]	黄彭, 郝妙, 杜永华, 等基于GM(1,1)模型的四川粮食产量预测研究[J]. 农学学报,2017(10).
[12]	刘峰, 王儒敬, 李传席. ARIMA模型在农产品价格预测中的应用[J]. 计算机工程与应用, 2009(25):242-243,252.
[13]	任青山, 方逵, 朱幸辉. 基于多元回归的BP神经网络生猪价格预测模型[J]. 江苏农业科学, 2019(14).
[14]	Lee Hyungyong, Hong Seungjee. Comparison of forecasting performance of time series models for the wholesale price of dried red peppers: focused on ARX and EGARCH[J]. Korean Journal of Agricultural Science, 2018,45(4):859-870.
[15]	Yashavanth B S, Singh K N, Paul A K, et al. Forecasting prices of coffee seeds using Vector Autoregressive Time Series Model[J]. Indian Journal of Agricultural Sciences, 2017,87(6):754-758.
[16]	李海生. 支持向量机回归算法与应用研究[D]. 广州:华南理工大学, 2005.
[17]	谢永华, 张鸣敏, 杨乐, 等. 基于支持向量机回归的城市PM_2.5浓度预测[J]. 计算机工程与设计, 2015(11):3106-3111.
[18]	Takens F. Detecting strange attractors in turbulence[J]. Lecture Notes in Mathematics, 1981,898(1):366-381.
[19]	张淑清, 贾健, 高敏, 等. 混沌时间序列重构相空间参数选取研究[J]. 物理学报, 2010(3).
[20]	Cheng Anyu, Jiang Xiao, Li Y F, et al. Multiple sources and multiple measures based traffic flow prediction using the chaos theory and support vector regression method[J]. Physica Astatistical Mechanics and Its Applications, 2017,466:422-434.
[21]	徐自励, 王一扬, 周激流. 估计非线性时间序列嵌入延迟时间和延迟时间窗的C-C平均方法[J]. 四川大学学报:工程科学版, 2007,39(1):151-155.
[22]	Delafrouz Hadi, Ghaheri Abbas, Ghorbani M A. A novel hybrid neural network based on phase space reconstruction technique for daily river flow prediction[J]. Soft Computing, 2018,22(7):2205-2215.
[23]	Su Yingxue, Xu Huan, Yan Lijiao. Support vector machine-based open crop model (SBOCM): Case of rice production in China[J]. Saudi Journal of Biological Sciences, 2017,24(3):537-547. doi: 10.1016/j.sjbs.2017.01.024 URL pmid: 28386178
[24]	胡瑜, 陈涛. 基于C-C算法的混沌吸引子的相空间重构技术[J]. 电子测量与仪器学报, 2012(5):53-58.
[25]	Cao L. Practical method for determining the minimum embedding dimension of a scalar time series[J]. Physica D, 1997,110(1-2):43-50.
[26]	汤荣志, 段会川, 孙海涛. SVM训练数据归一化研究[J]. 山东师范大学学报:自然科学版, 2016(4).
[27]	Keerthi S S, Lin C J. Asymptotic Behaviors of Support Vector Machines with Gaussian Kernel[J]. Neural Computation, 2003,15(7):1667-1689. doi: 10.1162/089976603321891855 URL pmid: 12816571