A New Clustering Algorithm On Nominal Data Sets
Bin Wang · 2010
Abstract—This paper presents a new clustering technique named as the Olary algorithm, which is suitable to cluster nominal data sets. This algorithm uses a new code with the name of the Olary code to transform nominal attributes into integer ones through a process named as the Olary transformation. The number of integer attributes we get through the Olary transformation is usually different from that of the original nominal attributes. Meanwhile, an extension of the Olary algorithm, which we call the ex-Olary algorithm, is introduced. Furthermore, we provide a useful way to estimate the number of underlying clusters by the use of a new kind of diagram, which is called Number of Clusters versus Distance Diagram (NCDD for short). Index Terms-Clustering, nominal, the Olary code, the Olary transformation, the ex-Olary algorithm, NCDD