Exploring Associative Classification Technique Using Weighted Utility Association Rules for Predictive Analytics

Association rule discovery determine the “inter-dependence” among various items in a transactional database. Data mining researchers have augmented upon the quality of association rule discovery for business development by integrating the influential factors like quantity of items sold (weight), profit (utility), for extracting the association patterns. This paper proposes a new model (associative classifier) based on weightage and utility for useful mining of substantial class association rules. In process of predicting the class lables, all attributes do not have same importance. So our framework considers the different frequencies of individual items as their weights and varied significance can be assigned to different attributes as their utilities according to their predicting capability. Initially, the proposed framework uses the CBA-RG algorithm to produce a set of class association rules from a database and as well as exploits the downward closure property of the apriori algorithm. Subsequently, the set of class association rules mined are subjected to weightage and utility constraints like W-gain, U-gain and a combined Utility Weighted Score (UW-Score) is calculated for the mining of class association rules. We purport a theoretical model to innovate new associative classifier that takes vantage of valuable Class association rules based on the UW-Score.