A.K.C. Wong and G.C.L. Li (Canada)
Clustering, association patterns, data interpretation
Statistical research in clustering has mainly focused on numerical data sets. It is difficult for existing clustering methods to apply to data sets involving nominal values. Besides, existing methods are seldom concerned with helping the users to interpret the results obtained. This paper proposes a novel clustering method that uses association patterns to obtain and characterize the clustering results. In many data mining applications such as basket analysis, association patterns have been used to capture relationship among events which can be easily understood by human. Using association patterns to describe the obtained clusters make the tasks of cluster interpretation and understanding easier. Experiments show that useful information can be readily acquired from the clustering outputs.
Important Links:
Go Back
