为提高猪肉价格预测的准确性, 结合互补集合经验模态分解(CEEMD)的分解能力和基于遗传算法的支持向量回归(GA-SVR)的自适应预测功能, 构建猪肉价格集成预测模型. 首先为解决猪肉价格的复杂波动特征, 通过CEEMD对猪肉价格分解得到本征模态函数(IMF)序列集; 然后使用排序熵(PE)对IMF序列进行复杂度分析, 进一步使用快速傅里叶变换方法(FFT)分解复杂度高的序列; 再利用灰色关联度(GCD)对IMF序列集进行关联性分析, 聚合相似IMF序列; 最后基于各IMF序列的数据特征构建相应的GA-SVR预测模型, 并将子序列的预测结果集成获得最终价格预测值. 以中国集贸市场的猪肉价格为研究对象, 实证结果表明, 该集成预测模型在预测精度和方向性指标上, 显著优于其他单预测模型和分解集成预测模型.

In order to improve the accuracy of pork price forecasting, a pork price ensemble prediction model is constructed, based on the decomposition ability of the complementary ensemble empirical mode decomposition (CEEMD) method and the adaptive prediction property of genetic algorithm-support vector regression (GA-SVR) model. First, for solving the complex wave characteristics of pork price, CEEMD is used to decompose pork price and obtain the Intrinsic Mode Function (IMF) sequence set. Second, permutation entropy (PE) is used to analyze the complexity of IMF sequences so as to further decompose complex sequences by the Fast Fourier Transform method (FFT). Three, grey correlation degree (GCD) is used to analyze the correlation coefficient of IMF sequences and combine the similar IMF sequences. Finally, the GA-SVR prediction model is constructed to predict the IMFs, and the prediction results of each sub-sequence are integrated to obtain the final price prediction. Taking the market pork prices in China as samples, the empirical results show that the ensemble prediction model is significantly superior to other single prediction models and decomposition ensemble prediction models in prediction accuracy and directionality indicators.