APP下载

经验模态分解-图神经网络算法预测农产品价格

2024-05-21赖玉莲马琳娟张延林

济南大学学报(自然科学版) 2024年3期

赖玉莲 马琳娟 张延林

文章編号:1671-3559(2024)03-0356-06DOI:10.13349/j.cnki.jdxbn.20240312.003

摘要: 为了提高图神经网络算法对农产品价格预测精度, 采用经验模态分解法按时间片轮转抽取农产品历史价格信号, 以便对历史价格信号进行特征提取; 将原始价格信号分解成多个本征模态函数及残余项, 并根据本征模态函数构建样本特征; 根据得到的样本特征构建价格预测图结构, 将图结构输出的特征信号通过图神经网络的过渡函数和预测函数, 通过不断减小损失值输出农产品价格预测结果。 结果表明, 经验模态分解可以对原始农产品价格信号的本征模态函数分量进行有效分解和提取, 从而使经验模态分解-图神经网络算法的农产品价格预测平均绝对误差减小71.4%; 相比于其他类型的预测算法, 经验模态分解-图神经网络算法对4类农产品价格预测的平均绝对误差更小, 最大值仅为2.465。

关键词: 农产品价格预测; 图神经网络; 经验模态分解; 本征模态函数

中图分类号: TP391

文献标志码: A

开放科学识别码(OSID码):

Agricultural Product Price Prediction Based on Empirical

Mode Decomposition and Graph Neural Network Algorithm

LAI Yulian1, MA Linjuan2, ZHANG Yanlin3

(1. School of Business Administration,Guangzhou Institute of Science and Technology,Guangzhou 510540, Guangdong, China;

2. School of Computer Science and Technology, Beijing Institute of Technology, Beijing 100081, China;

3. School of Management, Guangdong University of Technology, Guangzhou 510006, Guangdong, China)

Abstract: To improve the accuracy of agricultural product price prediced by using graph neural network (GNN) algorithm, empirical mode decomposition (EMD) method was used to extract the historical price signals of agricultural pro-ducts in turn according to time slices, so as to extract the characteristics of the historical price signals. The original price signal was decomposed into several instrinsic mode functions and residual terms, and the sample characteristics were constructed according to the instrinsic mode functions. According to the obtained sample characteristics, the graph structure was constructed, and the characteristic signals output from the graph structure pass through the transition function and prediction function of GNN algorithm, and the agricultural product price prediction results were output by continuously reducing the loss value. The results show that EMD can effectively decompose and extract the intrinsic mode function components of the original agricultural product price signals, thus the average absolute error of agricultural product price predicted by using EMD-GNN algorithm is reduced by 71.4%. Compared with other types of forecasting algorithms, the average absolute error of EMD-GNN algorithm for the price prediction of 4 types of agriculture products is smaller, and the maximum value is only 2.465.

Keywords: price prediction of agricultural products; graph neural network; empirical mode decomposition; intrinsic mode function

收稿日期: 2023-01-15          网络首发时间:2024-03-13T10:19:23

基金项目: 国家自然科学基金项目(72272039)

第一作者简介: 赖玉莲(1981—),女,广东梅州人。讲师,硕士,研究方向为数据分析、 农业科技、 产品分析。E-mail: lotuser101@126.com。

通信作者简介: 张延林(1974—),男,河南洛阳人。副教授,博士,研究方向为大数据与人工智能、数字技术。E-mail: forestgdut@163.com。

网络首发地址: https://link.cnki.net/urlid/37.1378.N.20240312.1702.006

农产品价格预测作为大数据技术在农业中的一项重要应用, 有效提升了农产品生产计划的预见性和合理性[1], 也影响了广大居民消费习惯, 更重要的是, 农产品价格的平稳运行关乎民生问题。……

登录APP查看全文