Granger相关性与时间序列预测

Granger causality and time series forecasting

提出一种将Granger相关信息用于时间序列预测的方法,以解决时间序列预测过程中信息利用不完全的问题.首先,通过Granger相关性检验确定时间序列系统中的可利用信息;然后,利用神经网络将可利用信息抽取出来;最后,将抽取的可利用信息融入到时间序列的预测中.实验结果验证了所提出预测方法的有效性和稳定性.

A method of time series forecasting using Granger causality information is presented. Which solves the problem of incomplete information in time series forecasting. Firstly, more available information is determined by correlation test among the system of time series. Then, neural networks are used to abstract the available information. Finally, the obtaining information is integrated in the process of forecasting. Experimental results show the effectiveness and stability of the proposed method.