A Concise Representation of Association Rules Using Minimal Predictive Rules

Association rule mining is an important branch of data mining research that aims to extract important relations from data. In this paper, we develop a new framework for mining association rules based on minimal predictive rules (MPR). Our objective is to minimize the number of rules in order to reduce the information overhead, while preserving and concisely describing the important underlying patterns. We develop an algorithm to efficiently mine these MPRs. Our experiments on several synthetic and UCI datasets demonstrate the advantage of our framework by returning smaller and more concise rule sets than the other existing association rule mining methods.