基于集成学习的ARIMA-LSTM模型在棉粕价格预测中的应用  

Application of ARIMA-LSTM model based on ensemble learning in cottonseed meal price prediction

作　　者：吴展 王春晓[1] WU Zhan;WANG Chunxiao

机构地区：[1]上海海洋大学经济管理学院,上海201306

出　　处：《饲料研究》2025年第2期227-231,共5页Feed Research

基　　金：国家现代农业产业技术体系(项目编号:CARS-47)。

摘　　要：准确预测棉粕价格对于稳定畜产品供给、促进饲料加工业可持续发展以及保障国家粮食安全至关重要。研究旨在基于长短期记忆神经网络(LSTM)的深度学习机制,构建棉粕价格预测模型。首先利用差分自回归移动平均(ARIMA)模型预测时间序列数据中线性变化,并应用LSTM算法估计棉粕价格序列的非线性效应。运用集成学习极限梯度提升(XGBoost)算法来确定残差序列滞后长度作为LSTM模型中的输入节点。最后,将拟合的线性和非线性变化之和作为ARIMA-LSTM组合模型的最终预测值。研究表明,基于XGBoost的ARIMA-LSTM混合模型优于单一的ARIMA时间序列预测模型,具有良好的预测性能。Accurate prediction of cottonseed meal prices is crucial for stabilizing the supply of livestock products,promoting the sustainable development of the feed processing industry,and ensuring national food security.This study aims to construct a cottonseed meal price prediction model based on the deep learning mechanism of the Long Short-Term Memory(LSTM)neural network.Initially,the study employs the Autoregressive Integrated Moving Average(ARIMA)model to forecast the linear changes in time series data,and then applies the LSTM algorithm to estimate the nonlinear effects in the cottonseed meal price series.The Extreme Gradient Boosting(XGBoost)algorithm,an ensemble learning method,is used to determine the lag length of the residual sequence as the input node in the LSTM model.Finally,the sum of the fitted linear and nonlinear changes is taken as the final prediction value of the ARIMA-LSTM hybrid model.The study indicates that the ARIMA-LSTM hybrid model based on the XGBoost algorithm outperforms the single ARIMA time series prediction model and demonstrates good forecasting performance.

关 键 词：深度学习 棉粕价格预测 集成学习 ARIMA模型 XGBoost算法 

分 类 号：S8-9[农业科学—畜牧兽医]