A new method of mining classification rule using relevant concept cover of concept lattice

摘要

Classification rule mining is one of important areas of data mining, and it is a hot subject at present. In this paper, we introduce some definitions: relevant concept, pseudo concept, and relevant concept cover. Then we presents an algorithm to compute relevant concept cover, which divides the context into subsets, calculate the set of relevant concepts by the value of concept shannon entropy, and then increasingly constructs relevant concept cover by deleting redundant concept orderly. Finally, we give a new method to classify rules using relevant concept cover of concept lattice, and through experiments prove, the computing speed of this method is faster and the number of rules is more.