基于最大信息系数的时延数据相关性分析方法

Method for the correlation analysis of data with time delay based on maximal information coefficient

针对无法有效检测两组时延数据间相关关系的情况,提出以最大信息系数(MIC)为基础的平移搜索法。根据实际应用场景,设置合适的平移搜索窗和平移步长,由搜索窗内取得最大MIC值的位置求得时延估计值。将此方法分别应用到航天器载荷安装表面温度之间的相关性分析和狭义货币供应量(M1)与居民消费价格指数(CPI)的相关性分析中,结果表明针对两组时域上不对应的相关数据,利用此方法可以有效地检测出它们的相关性和时延。

Aiming at the problem that the correlation of two sets of data with time delay can＇t be detected effectively,a parallel moving search method based on Maximal Information Coefficient（MIC）is proposed.According to the practical application of the scene,set the appropriate parallel moving search window and step,obtain the time delay estimation value by the position with the maximum MICvalue.Apply these methods on the correlation analysis of the equipment installation surface temperature data in aerospace and M1-CPI data in economics,results show that if two sets of data have correlation but with a time delay in time domain,using this method can detect the correlation and time delay effectively.