S-IAMB Algorithm for Markov Blanket Discovery

Using Markov Blanket for feature selection is one of important methods in machine learning. This paper analyses IAMB algorithm for discovering the Markov Blanket of a target variable from training data. Based on its characters and deficiency we introduce an improve algorithm of IAMB: SIAMB. The experimental results show that the S-IAMB algorithm performs better than IAMB by finding Markov Blanket of variables of ALARM dataset and by testing the performance with the classification tasks.