基于奇异谱分解的猪肉价格组合预测  

Combined Prediction of Pork Prices Based on Singular Spectrum Decomposition

作　　者：黄增跃 方永美[1] HUANG Zengyue;FANG Yongmei

机构地区：[1]华南农业大学数学与信息学院,广东广州510642

出　　处：《山西科技》2020年第3期96-101,共6页Shanxi Science and Technology

基　　金：华南农业大学校级大学生创新训练项目(项目编号:201710564183)。

摘　　要：通过奇异谱分析方法分解猪肉价格,采用ARIMA模型、SVM模型和BP神经网络模型对分解后的猪肉价格进行组合预测;同时选择ARI⁃MA,SVM和BP神经网络作为基准模型,把组合模型预测的结果与所选的基准模型预测结果进行对比,得到了组合模型预测结果总体上优于基准模型预测结果的结论。通过DM检验,进一步验证了结果的可靠性。预测结果表明,SSA组合模型的预测能力平均比ARIMA、SVM和BP神经网络3种基准模型的预测能力分别高出7.97%、72.79%、67.64%.The pork price was decomposed by singular spectrum analysis.The combined prediction of the decomposed pork price was made by using ARIMA model,SVM model and BP neural network model.At the same time,ARIMA,SVM and BP neural network were selected as the benchmark model,and the predicted results of the combined model were compared with the predicted results of the selected benchmark model,and it was concluded that the predicted results of the combined model were generally superior to the predicted results of the benchmark model.The reliability of the results was further verified by DM test.The prediction results show that the prediction ability of SSA combined model is on average 7.97%,72.79%and 67.64%higher than that of ARIMA,SVM and BP neural network.

关 键 词：组合模型 奇异谱分解 自回归移动平均模型 支持向量机 猪肉价格 

分 类 号：F222.1[经济管理—国民经济]