An Effective Clustering-based Approach for Conceptual Association Rules Mining

Abstract

Association rule mining is a well-known data mining task for discovering association rules between items in a dataset. It has been successfully applied to different domains especially for business applications. However, the mined rules rely heavily on human interpretation in order to infer their semantic meanings. In this paper, we mine a new kind of association rules, called conceptual association rules, which imply the relationships between concepts. Conceptual association rules can convey more semantic meanings than those classical association rules. Conceptual association rules can be mined using Formal Concept Analysis (FCA). However, the FCA-based method for conceptual rule mining suffers from high computational cost when dealing with large datasets. To tackle this problem, we propose a cluster-based approach to mine conceptual association rules regionally, rather than globally. A distance metric is also proposed to ensure that the same rule sets will ultimately be obtained when the dataset is clustered. In this paper, we present the proposed clustering-based approach. In addition, the proposed approach has been evaluated with four benchmarking datasets and promising results have been achieved.