摘要
准确预测大豆价格对维护农户基本收益和健全粮食价格市场形成机制至关重要。大豆价格因其非线性、高波动性等特点,传统时间序列模型难以满足精度要求和社会经济需要。本文利用4种经验模态分解方法(EMD、EEMD、CEEMD、ICEEMD)对大豆价格时序数据进行分解,然后用5种群智能优化算法(蜂群ABC、布谷鸟CS、蜻蜓DA、蝗虫GOA、粒子群PSO)优化的支持向量回归(SVR)模型分别进行预测,最后子序列预测值通过线性集成方法得到大豆价格最终预测值。为了验证分解-优化混合模型预测效果,引入单一SVR和未分解-优化模型作为基准比对,从30个模型预测结果对比和模型检验两个维度进行分析,发现混合模型在预测大豆期货价格数据上有更好的预测精度。
Predicting soybean price accurately is crucial to maintaining the basic income of farmers and building wholesome grain market system.The traditional time series model is not accurate enough to meet the accuracy requirements and economic needs due to soybean price’s nonlinearity and high volatility.The soybean futures price data was decomposed by four empirical mode decomposition methods (Empirical Mode Decomposition,Ensemble EMD,Complete Ensemble EMD,Improved complete ensemble EMD).Then,support vector regression models optimized by five Smart Optimization algorithms (Artificial Bee Colony,Cuckoo Search,Dragonfly Algorithm,Grasshopper Optimisation Algorithm,Particle Swarm Optimization) were used to predict respectively.Finally,the soybean daily price forecasting model was established by combining the decomposed subsequence prediction values.In order to verify the Decomposition-Optimized hybrid model,the single SVR and the Undecomposed-Optimized model were introduced as benchmark comparisons.From the comparison of 30 model prediction results and model test,it was found that the hybrid model has the best prediction accuracy in predicting soybean futures price data.
作者
何润奇
He Runqi(School of Statistics and Mathematics,Zhongnan University of Economics and Law,Wuhan 430073,China)
出处
《中南财经政法大学研究生学报》
2019年第2期24-32,共9页
Journal of the Postgraduate of Zhongnan University of Economics and Law
关键词
经验模态分解
群智能优化算法
支持向量回归
大豆期货价格
Empirical Mode Decomposition
Support Vector Regression
Smart Optimization Algorithms
Soybean Futures Price
