基于互信息的直接因果搜索算法

因果关系挖掘是数据挖掘领域一个新的研究方向,具有很大的实用意义,但理论建模的困难阻碍了它的发展。20世纪90年代,在因果建模研究成果的基础上,国外开始针对此方向进行研究,目前已得到了一些理论算法。该文针对已有算法鲁棒性、实用性差,不适合大规模数据挖掘应用的缺点,提出了一种基于互信息的因果数据挖掘算法———直接因果搜索算法;仿真试验表明,该算法能很好地搜索出目标变量的直接因果,与其他算法相比,对于数据、门限的变化更具鲁棒性。

Direct Causal Relation Searching Algorithm

The Causal Relation Data Mining is a new direction in the Data Mining research field,and it is very useful in the application.The Difficulty of modeling blocks its development.In the1990s,on the basis of achievement in the causal relation modeling,foreign researchers began to do some works on it.Now,some theoretical algorithms have been achieved.In this paper,a new causal relation data mining algorithm,named direct causal relation searching algorithm is proposed.The simulation shows that this algorithm can find the right direct causal variables of the object variable.Compared with other algorithms ,it is more robust to the change of data and threshold.