基于CEEMD-LSTM-Adaboost模型的白糖期货跨期套利策略
Arbitrage strategies of sugar futures based on CEEMD-LSTM-Adaboost model
甘柳燕 1唐国强 1蒋文希 1覃良文1
作者信息
- 1. 桂林理工大学 数学与统计学院,广西 桂林 541006
- 折叠
摘要
以白糖期货合约SR2201 和SR2109 的 5 min高频数据为研究对象,在验证二者存在长期均衡关系的条件下,构建GARCH模型来刻画残差的ARCH效应,将互补集合经验模态分解(CEEMD)方法与长短期记忆网络(LSTM)、自适应提升算法(Adaboost)相结合,通过预测价差涨跌进行套利操作,设置不同开平仓阈值,在样本区间内进行 4 种神经网络套利策略对比研究.结果表明:基于CEEMD-LSTM-Adaboost模型的神经网络套利策略应用于白糖期货市场可行有效,并且其在模型预测精度和套利效果方面均比BP、LSTM和LSTM-Adaboost神经网络更具优势.
Abstract
The 5-minute high-frequency data of white sugar futures contracts SR2201 and SR2109 are taken as the research object.Under the condition that there is a long-term equilibrium relationship between them,GARCH model is constructed to describe the ARCH effect of the residual.When the complementary set empiri-cal mode decomposition(CEEMD)method is combined with the long-term and short-term memory network(LSTM)and adaptive lifting algorithm(Adaboost),arbitrage operation is carried out by predicting the rise and fall of price difference,setting different opening and closing thresholds,and making a comparative study of four neural network arbitrage strategies in the sample range.Results show that the neural network arbitrage strategy based on CEEMD-LSTM-Adaboost model is feasible and effective in white sugar futures market,and it has more advantages than BP,LSTM and LSTM-Adaboost neural networks in terms of prediction accuracy and arbitrage effect.
关键词
跨期套利/CEEMD-LSTM-Adaboost模型/白糖期货Key words
intertemporal arbitrage/CEEMD-LSTM-Adaboost combination model/sugar futures引用本文复制引用
基金项目
国家自然科学基金(11961013)
国家自然科学基金(61763008)
出版年
2024
国家科技期刊平台
NSTL
万方数据