An Information Theoretic Scoring Function in Belief Network

We proposed a novel measure of mutual information known as Integration to Segregation (I2S) explaining the relationship between two features. We investigated its nontrivial characteristics while comparing its performance in terms of class imbalance measures. We have shown that I2S possesses characteristics useful in identifying sink and source (parent) in a conventional directed acyclic graph in structure learning technique such as Bayesian Belief Network. We empirically indicated that identifying sink and its parent using conventional scoring function is not much impressive in maximizing discriminant function because it is unable to identify best topology. However, I2S is capable of significantly maximizing discriminant function with the potential of identifying the network topology in structure learning