基于注意力机制的GRU模型的豆粕期货价格预测  

Prediction of Soybean Meal Futures Price Based on GRU Model of Attention Mechanism

作　　者：石榕 SHI Rong(School of Statistics,Lanzhou University of Finance and Economics,Lanzhou 730020,China)

机构地区：[1]兰州财经大学统计学院,甘肃兰州730020

出　　处：《洛阳理工学院学报（自然科学版）》2023年第4期87-91,共5页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

摘　　要：针对豆粕期货价格的波动性和复杂性,提出了一种引入注意力机制的门控循环单位(GRU)的价格预测模型。由于豆粕期货价格具有长期依赖性特点,选择近5年的日度数据作为数据集,将基本行情和技术指标利用互信息、相关系数、随机森林树RF模型3种方法相结合进行筛选排序,选用了15种技术指标进行分析后,输入到基于注意力机制的GRU模型中对豆粕主力合约的未来收盘价进行预测。引入注意力机制的GRU模型相较于GRU基准模型,在加入技术指标的多特征时间序列预测性能上表现更优,提高了预测精度。In view of the fluctuation and complexity of soybean meal futures price,a price forecasting model,GRU with attention mechanism,is put forward.Due to the long-term dependence of soybean meal futures price,the daily data of recent five years are selected as the data set,and the basic market and technical indicators are selected and sorted by combining mutual information,correlation coefficient and random forest tree RF model.Finally,15 technical indicators are selected for analysis and input into GRU model based on attention mechanism to predict the future closing price of soybean meal main contracts.The results show that GRU model with attention mechanism is better than GRU benchmark model in the performance of multi-feature time series prediction with technical indicators,and the prediction accuracy is improved.

关 键 词：豆粕 技术指标 价格预测 注意力机制 GRU 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] F832.5[自动化与计算机技术—控制科学与工程]