论文部分内容阅读
国内猪肉市场价格具有波动大、非线性、非平稳,且样本量少的特点,很难进行预测。为了提高预测精度,并有效解释价格波动的内在经济含义,基于集成预测思想,提出EMD-SVM集成预测模型。首先用经验模态分解方法(EMD)把猪肉市场月度价格分解成若干个不同尺度的,相对平稳的本征模态分量(IMF),按照频率高低,将各IMF分量集成为高频部分、低频部分和残余项三大模块,解决波动大、非平稳问题。在此基础上运用支持向量机(SVM)对3个集成模块分别进行预测,从而解决非线性问题。为了使预测模型最优,SVM的参数用遗传算法进行寻优。最后对3个集成模块的预测结果再次进行集成,重构出猪肉市场价格预测值。为了验证模型的有效性,将EMD-SVM集成预测模型与SVM、EMD-BP、BP的预测结果进行分类比较,其RMSE、MAPE和方向性都明显提高。
The domestic market price of pork with volatility, non-linear, non-stationary, and the characteristics of small samples, it is difficult to predict. In order to improve the prediction accuracy and effectively explain the inherent economic implications of price fluctuations, an integrated forecasting model of EMD-SVM is proposed based on the idea of integrated forecasting. Firstly, EMF is used to decompose the monthly price of pork market into a number of relatively stable IMFs. According to the frequency, each IMF component is integrated into a high-frequency part. The low frequency Part and the remaining three major modules to solve the volatile, non-stationary issues. Based on this, SVM is used to predict the three integrated modules respectively to solve the nonlinear problem. In order to optimize the prediction model, the parameters of SVM are optimized by genetic algorithm. Finally, the forecast results of the three integrated modules are integrated again to reconstruct the predicted market price of pork. In order to verify the validity of the model, the RMSE, MAPE and directionality of the EMD-SVM integrated prediction model were compared with those of SVM, EMD-BP and BP.