Abstract: To capture the complex characteristics of price fluctuations and further improve prediction accuracy of the agricultural futures market, a hybrid model, combined with variational mode decomposition and extreme learning machine, is constructed based on the decomposition and ensemble theory. The VMD model, which can effectively avoid mode mixing and end effects in signal decomposition, is more robust to complex time series. In addition, by randomly generates the connection weights and implicit between the input layer and the hidden layer, the ELM model solves the problems of slow convergence and overfitting of traditional machine learning model. More importantly, aiming to select the key parameter mode K of the VMD model, a K-value optimization method based on the minimum fuzzy entropy criterion is proposed. Taking the rice, wheat, and soybean meal futures closing prices of CBOT as the objects, the empirical results show that the performance of the proposed VMD-ELM hybrid model is optimal.

Key words: VMD, ELM, decomposition and ensemble, agricultural commodity futures, forecast

摘要： 为了捕捉农产品市场期货价格波动的复杂特征,进一步提高其预测精度,基于分解集成的思想,构建包含变分模态分解(VMD)和极限学习机(ELM)的分解集成预测模型。首先,利用VMD分解的自适应性和非递归性,选择VMD将复杂时间序列分解成多个模态分量(IMF)。其次,针对VMD分解关键参数模态数K的选取难题,提出基于最小模糊熵准则寻找最优K值的方法,有效避免模态混淆和端点效应问题,从而提升VMD的分解能力。最后,利用ELM强大的学习能力和泛化能力,对VMD分解得到的不同尺度子序列进行预测,集成得到最终预测结果。以CBOT交易所稻谷、小麦、豆粕期货价格作为研究对象,实证结果表明,该分解集成预测模型在预测精度和方向性指标上,显著优于单预测模型和其它分解集成预测模型,为农产品期货价格预测提供了一种新途径。

关键词: 变分模态分解, 极限学习机, 分解集成, 农产品期货价格, 预测