Weighted Bayesian Association Rule Mining Algorithm to Construct Bayesian Belief Network

Abstract

Bayesian network is an appropriate tool to work with the unpredictability and causality which arises in Clinical Domain. A Bayesian network can learn from medical datasets without explicit access to the knowledge of human experts. Thus, to built Bayesian network by learning method, strong rules are needed from datasets. To express statistical dependence relationships, association rules can be considered. This proposal expects to incorporate two techniques to improve the shortcoming of single technique, so this proposal put forward a Weighted Bayesian Association rule Mining Algorithm (WBAR) for the generation of strong Bayesian association rules for the construction of Bayesian network which combines the weighted concept with Association Rule Mining (ARM) to generate Weighted Two-attributes association rules, Weighted Multi -attributes association rules and Weighted Class Association rules. Two interesting dimensions of association rules mining: Weighted Bayes confidence (WBC) and Weighted Bayes lift (WBL) that assess the relationship between different attributes using conditional dependence and independence based on the joint probabilities which are symbolized and then interpreted by the Weighted Bayesian networks using association rules. The proposed algorithm WBAR results the most significant rules according to WBC and WBL