Clustering Categorical Data Using Community Detection Techniques.

Huu Hiep Nguyen

doi:10.1155/2017/8986360

Abstract

With the advent of the k-modes algorithm, the toolbox for clustering categorical data has an efficient tool that scales linearly in the number of data items. However, random initialization of cluster centers in k-modes makes it hard to reach a good clustering without resorting to many trials. Recently proposed methods for better initialization are deterministic and reduce the clustering cost considerably. A variety of initialization methods differ in how the heuristics chooses the set of initial centers. In this paper, we address the clustering problem for categorical data from the perspective of community detection. Instead of initializing k modes and running several iterations, our scheme, CD-Clustering, builds an unweighted graph and detects highly cohesive groups of nodes using a fast community detection technique. The top-k detected communities by size will define the k modes. Evaluation on ten real categorical datasets shows that our method outperforms the existing initialization methods for k-modes in terms of accuracy, precision, and recall in most of the cases.

Highlights

Clustering task is a form of unsupervised learning that aims at finding underlying structures in unlabeled data
While hierarchical clustering produces a hierarchy of partitions over the dataset by applying agglomerative or divisive strategies, partitional clustering usually assumes a fixed number of clusters and tries to maximize the homogeneity within the clusters
We develop a new clustering method for categorical data based on community detection techniques [11]

Summary

Introduction

Clustering task is a form of unsupervised learning that aims at finding underlying structures in unlabeled data. To better initialize cluster centers in k-modes, a lot of methods have been developed [7,8,9,10]. The common point of [7, 8, 10] is to use the density of each data point together with the distance to determine sequentially the k initial cluster centers. We develop a new clustering method for categorical data based on community detection techniques [11]. (i) We propose a novel clustering method called CDClustering for categorical data using community detection techniques.

Related Work

Preliminaries

Algorithm

Evaluation

Conclusion