Ant colony based hybrid optimization for data clustering

Abstract

Purpose

A new algorithm based on ant colony optimization (ACO) for data clustering has been developed.

Design/methodology/approach

ACO technique along with simulated annealing, tournament selection (GA), Tabu search and density distribution are used to solve unsupervised clustering problem for making similar groups from arbitrarily entered large data.

Findings

Distinctive clusters of similar data are formed metaheuritically from arbitrarily entered mixed data based on similar attributes of data.

Research limitations/implications

The authors have run a computer program for a number of cases related to data clustering. So far, there are no problems in convergence of results for formation of distinctive similar groups with given data set quickly and accurately.

Practical implications

ACO‐based method developed here can be applied to practical industrial problems for mobile robotic navigation other than data clustering and travelling salesman.

Originality/value

This paper will enable the solving of problems related to mixed data, which requires the formation of a number of groups of similar data without having a prior knowledge of divisions, which lead to unbiased clustering. The computer code developed in this work is based on a metaheuristic algorithm and presented here to solve a number of cases.

Citation

Nath Sinha, A., Das, N. and Sahoo, G. (2007), "Ant colony based hybrid optimization for data clustering", Kybernetes, Vol. 36 No. 2, pp. 175-191. https://doi.org/10.1108/03684920710741215