Redundant feature identification and redundancy analysis for causal feature selection

High dimensional data can lead to low accuracy of classification and take a long time to calculate because it contains irrelevant features and redundant features. To overcome this problem, dimension of data has to be reduced. Causal feature selection is one of methods for feature reduction but it cannot identify redundant features. This paper presents Parent-Children based for Causal Redundant Feature Identification (PCRF) algorithm to identify and remove redundant features. The accuracy of classification and number of feature reduced by PCRF algorithm are compared with correlation feature selection. According to the results, PCRF algorithm can identify redundant feature but has lower accuracy of classification than correlation feature selection.