基于多频优化组合模型的我国大豆期货价格预测

针对ARIMA、BPNN、LSTM等单一模型在预测大豆期货价格时因不能同时捕获到原始序列中线性和非线性变化特征而导致的预测精度不高的问题,提出基于完全自适应噪声集合经验模态分解(CEEMDAN)的多频优化组合模型,并利用大豆的日期货收盘价数据对多频优化组合模型的有效性进行了实证分析.结果表明,多频优化组合模型在大豆期货价格预测精度上优于BPNN、LSTM等单一模型,以及EMD - BPNN、CEEMDAN - LSTM(未重构)等组合模型,因此该模型在预测大豆期货价格走势中具有良好的参考价值.

Prediction of soybean futures price in China based on multi - frequency optimal combination model

To address the problem that single models such as ARIMA, BPNN, and LSTM cannot capture both linear and nonlinear variation in the original series of soybean futures prices, a multi - frequency optimal combination model based on complete ensemble empirical modal decomposition with adaptive noise(CEEMDAN)is proposed.The empirical results show that the multi - frequency optimal combination model outperforms single models such as BPNN and LSTM, as well as combined models such as EMD - BPNN and CEEMDAN - LSTM(unreconstructed)in predicting soybean futures price trends.Therefore, the model has good reference value in forecasting soybean futures price movements.