New approach for clustering relational data based on relationship and attribute information

Abstract

A wide range of the database systems in use today are based on the relational model. As a consequence, more information used by those systems has been stored in multi relational object types. However, most of the traditional machine learning algorithms have not been originally proposed to handle this type of data. Aiming to propose better ways of handling the relational particularities of the data, this paper proposes a new relational clustering method based on relationship and attribute information. In our method, attributes have weights associated with their importance between the object types. An empirical analysis is performed in order to evaluate the effectiveness of the proposed method, comparing with two traditional methods for relational clustering. Three relational databases were used in the experiments. © 2012 Springer-Verlag.