A two-stage clustering algorithm for the two-attribute-set problem

Cluster analysis has long been a highly active topic in the area of data mining research. Generally, traditional clustering algorithms use the same set of attributes for both partitioning the data space and measuring the similarity between objects when clustering data. There are, however, some practical situations where one should make a distinction between these two attribute sets. For example, a bank needs to cluster its customers to learn about the consumption behaviors of customers with different backgrounds. In other words, customers need to be grouped into clusters with similar backgrounds, such as gender, age and income; at the same time, customers in the same cluster should have similar consumption behaviors. Therefore, two different sets of attributes are required, one for measuring similarity, called the similarity-measuring attribute, and the other for partitioning the data set as well as describing the resulting cluster, called the dataset-partitioning attribute. Traditional algorithms do not distinguish between these two sets of attributes which can lead to low quality clustering results for such two-attribute-set problems. In this work, we propose a Two-stage Clustering Algorithm to generate clusters for the two-attribute-set problem.