江西省生猪价格波动的成因及其预警分析——基于灰色关联和LS-SVM模型

以2000年1月至2015年5月的江西生猪价格数据为研究对象,利用Census-X12和HP滤波分解方法探索生猪价格波动的特征,结合逐步回归法和灰色关联分析识别影响生猪价格波动的显著因素,在此基础上,构建LS-SVM模型对生猪价格进行预测。结果表明,生猪价格波动具有明显的季节性,每年的1月份季节因子最大,6月份降至全年的最低点;2000年以来生猪价格共经历了7个波动周期,平均周期为25.3个月;随机性成分对生猪价格的贡献日益增大,玉米价格、仔猪价格、猪肉价格、生产者预期、牛肉价格和疫情对生猪价格的波动有显著作用,其中玉米价格和仔猪价格的影响较大;LS-SVM模型的预测值和真实值很接近,平均误差仅为1.37%,LS-SVM能较好地反映生猪价格及其影响因素之间的复杂的非线性关系。

Fluctuation and forecast analysis of hog price in Jiangxi Province based on LS-SVM with grey correlation analysis

The paper explored the characteristics of fluctuation for hog price from January 2000 to May 2015 by Census-X12 and HP analysis and identified significant factors for hog price by stepwise regression and grey correlation analysis. On this basis, the LS-SVM model was constructed to predict hog price. The results showed that the fluctuation for hog price was seasonal with the largest seasonal factor in January and the lowest in June. Since 2000, hog price had experienced seven fluctuation cycles and each cycle continued 25.3 months. Contribution of stochastic component to price was increasing.Corn price, piglet price, pork price, producers?? expectation, beef price and epidemic disease were notable factors for hog price, among which corn price and piglet price were the largest. The predicted value and actual value of LS-SVM were close. The average error was only 1.37%. The LS-SVM model can effectively describe the nonlinear relations between hog price and its factors.