考虑外生变量的基于分数差分的石油价格部分线性自回归预测模型

摘要

WTI spot price sequence was modelled and predicted based on fractional differencing and partially linear autoregression model. Firstly,we obtain a sbort memory sequence which is fractional differencing e-liminating long memory of the WTI spot price sequence, and then introduce a model in which the parameters part considers oil prices and the non-parameters aspect considers exogenous variables that is world oil supply,by using partially linear autoregression model for the short memory. The result of the empirical research show that the partially linear autoregression model based on fractional differencing can solve the oil price forecast better,and the introduction of exogenous variables,further enhance the model's explanatory ability,and make up the model of the external factors neglected defects. It has high accuracy of prediction.%利用分数差分和部分线性自回归模型对WTI现货价格序列进行了建模和预测研究.首先通过分数差分消除了WTI现货价格序列中的长记忆性,得到一条短记忆序列.然后,利用部分线性自回归模型对其进行建模,其中,参数部分考虑石油价格,非参数部分考虑外生变量,即世界供应量,并进行了实证研究.研究结果表明:基于分数差分的部分线性自回归模型能较好地解决石油价格预测这一问题,而引入外生变量后,进一步增强了模型的解释能力,弥补了模型对外界影响因素忽略的缺陷,预测精度较高.