Clustering Mixed Data Set by Fuzzy Set Partitioning

Abstract

K mean clustering is a very popular clustering algorithm for clustering numerical data. . It is popular due to its simplicity of understanding and linear algorithmic complexity measure. But it has the serious limitation of clustering numerical only data. Therefore several researchers tried to improve the k mean algorithm to cluster not only numerical but also categorical dataset. In this work an effort have been made to put forward a proposed FCV mean algorithm which is a modified version of the traditional k-mean algorithm and is able to cluster objects having mixed type attributes i.e. numerical and categorical. For categorical data fuzzy set similarity is used and for numerical data differences from maximum dissimilarity is used. Experiment shows that the mixed data are highly clustered with high accuracy compared to other approach in literature. General TermsPattern Recognition