基于LSTM深度学习的大豆期货价格预测

面对越来越复杂的金融市场环境,以传统统计学和计量学为主的时间序列预测模型在发现序列中的长期依赖关系方面存在一定局限性,而深度学习中的长短期记忆(LSTM)网络有望克服这一问题。通过构造一个多层LSTM网络价格预测模型,使用中国2007—2019年大豆期货价格数据进行了实证研究。结果显示,参数调优对LSTM网络模型预测效果有着较大影响,其中影响较大的主要参数包括迭代次数、学习率、窗口大小和网络层数等;与ARIMA模型、MLP模型、SVR模型相比,LSTM网络模型的预测结果准确性更高,在拟合优度(R-2)上分别提高了1.064%、2.147%、1.674%。LSTM网络模型在价格预测方面的良好表现,为预测大豆期货价格提供了新思路。

Soybean future prices forecasting based on LSTM deep learning

Faced with an increasingly complex financial market environment,time series forecasting models based on traditional statistics and metrology have certain limitations in discovering long-term dependencies in the series,but Long Short-Term Memory(LSTM)in deep learning is expected to overcome this problem.By constructing a multi-layer LSTM prices forecasting model,this paper made an empirical study on the China’s 2007-2019 soybean future prices data.The results show that the parameter optimization has a great impact on the forecasting effect of the LSTM model,and the main parameters that have a greater impact include the number of iterations,learning rate,window size and the number of network layers.Compared with ARIMA model,MLP model and SVR model,the forecasting results of the LSTM model are more accurate,and the goodness of fit(R-2)is improved by 1.064%,2.147%and 1.674%,respectively.