Abstract

ABSTRACT

An Efficient Technique for Clustering Data with Mixed Attribute Types

Rahmah Brnawy
Clustering is a technique used to extract useful information and discover patterns from data. Existing clustering techniques have often focused on datasets with attributes that are either numeric or categorical but not both. The problem of clustering mixed numeric and categorical datasets has received increased attention more recently and a number of solutions have been proposed. In this research, we study these solutions and propose two clustering algorithms. The first algorithm that we present is called Cluc+, which extends and improves Cluc, an existing algorithm proposed for clustering pure categorical data. Using Cluc+, we then develop a new algorithm, called k-mixed for clustering data with mixed numeric and categorical attribute types. We conduct numerous experiments to evaluate the performance of our proposed algorithms using real-life benchmark datasets. Our results indicate increased efficiency and accuracy of the proposed solution techniques.