基于组合模型的大宗商品价格预测与可视分析——以甲醇价格为例

摘要

Due to the influence of domestic and international factors,the prices of commodities fluctuate more greatly and is difficult to predict accurately.A reasonable screening method of influencing factors is designed,and a combination forecasting method based on factor analysis is put forward.By predicting continuously the methanol price for one year,it is shown that the combination forecasting model based on GARCH (generalized auto-regressive conditional heteroskedasticity) and ARMA (auto-regressive and moving average) has a good effect on the prediction of the long-term trend of methanol price.In order to introduce the experience of experts and make up the defect of existing methods in predicting the inflection point of fluctuation,a dynamic visualization analysis system with the experience prediction of expert is constructed,which also can dynamically compare the prediction effect of all kinds of combination model.%大宗商品价格因受国际和国内众多因素的影响而具有较大的波动性,对其进行准确预测具有较大的挑战.从对大宗商品价格影响因素的筛选出发,提出了基于因素分析的组合预测方法.对一年期的甲醇价格的跟踪预测表明,以广义自回归条件异方差(generalized auto-regressive conditional heteroskedasticity,GARCH)模型和自回归移动平均(auto-regressive and moving average,ARMA)模型相结合的组合预测模型对甲醇价格的中长期趋势预测有较好的效果.为结合专家的经验判断,弥补已有方法对波动拐点预测滞后的不足,并对各类组合模型的预测效果进行动态比较,构建了一个融合专家经验值的动态可视分析系统.