A modified tabu search approach for the clustering problem

Adnan Kharrousheh, Salwani Abdullah, Mohd Zakree, Mohd Zakree Ahmad Nazri

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Clustering is a data mining technique used to classify a number of objects into k clusters without advance knowledge such that the distance between objects within a same cluster and its center is minimised. The modelling for such problems is quite complex and searching for optimal solution usually impractical in term of computation times. Therefore, metaheuristic methods are used to find an acceptable solution within reasonable computational time. The problem addressed here provides the motivation for this work to develop a metaheuristic method to solve clustering problems. Most of previous techniques are based on K-means algorithm. However, K-means algorithm is highly depends on the initial state and very fast to get trap in local optimal solution. This work presents a modified tabu search approach to solve clustering problems that consists of two phases i.e., a constructive phase to generate an initial solution using K-means algorithm and an improvement phase, where a modified tabu search is employed with an aim to enhance the solution quality obtained from the first phase. In this study, a double tabu lists is introduced where the first tabu list is used to keep the neighbourhood structures and the second tabu list is used to keep the moves that are involved in that particular neighbourhood structure. The performance of the proposed algorithm is tested on five well-known datasets. Preliminary computational experiments are encouraging and compare positively with both the K-means and a standard tabu search alone.

Original languageEnglish
Pages (from-to)3447-3453
Number of pages7
JournalJournal of Applied Sciences
Volume11
Issue number19
DOIs
Publication statusPublished - 2011

Fingerprint

Tabu search
Data mining
Experiments

Keywords

  • Clustering
  • Double tabu list
  • K-means
  • Modified tabu search

ASJC Scopus subject areas

  • General

Cite this

A modified tabu search approach for the clustering problem. / Kharrousheh, Adnan; Abdullah, Salwani; Zakree, Mohd; Ahmad Nazri, Mohd Zakree.

In: Journal of Applied Sciences, Vol. 11, No. 19, 2011, p. 3447-3453.

Research output: Contribution to journalArticle

@article{9e3c88aba9f4462096bdaa0371afe326,
title = "A modified tabu search approach for the clustering problem",
abstract = "Clustering is a data mining technique used to classify a number of objects into k clusters without advance knowledge such that the distance between objects within a same cluster and its center is minimised. The modelling for such problems is quite complex and searching for optimal solution usually impractical in term of computation times. Therefore, metaheuristic methods are used to find an acceptable solution within reasonable computational time. The problem addressed here provides the motivation for this work to develop a metaheuristic method to solve clustering problems. Most of previous techniques are based on K-means algorithm. However, K-means algorithm is highly depends on the initial state and very fast to get trap in local optimal solution. This work presents a modified tabu search approach to solve clustering problems that consists of two phases i.e., a constructive phase to generate an initial solution using K-means algorithm and an improvement phase, where a modified tabu search is employed with an aim to enhance the solution quality obtained from the first phase. In this study, a double tabu lists is introduced where the first tabu list is used to keep the neighbourhood structures and the second tabu list is used to keep the moves that are involved in that particular neighbourhood structure. The performance of the proposed algorithm is tested on five well-known datasets. Preliminary computational experiments are encouraging and compare positively with both the K-means and a standard tabu search alone.",
keywords = "Clustering, Double tabu list, K-means, Modified tabu search",
author = "Adnan Kharrousheh and Salwani Abdullah and Mohd Zakree and {Ahmad Nazri}, {Mohd Zakree}",
year = "2011",
doi = "10.3923/jas.2011.3447.3453",
language = "English",
volume = "11",
pages = "3447--3453",
journal = "Journal of Applied Sciences",
issn = "1812-5654",
publisher = "Asian Network for Scientific Information",
number = "19",

}

TY - JOUR

T1 - A modified tabu search approach for the clustering problem

AU - Kharrousheh, Adnan

AU - Abdullah, Salwani

AU - Zakree, Mohd

AU - Ahmad Nazri, Mohd Zakree

PY - 2011

Y1 - 2011

N2 - Clustering is a data mining technique used to classify a number of objects into k clusters without advance knowledge such that the distance between objects within a same cluster and its center is minimised. The modelling for such problems is quite complex and searching for optimal solution usually impractical in term of computation times. Therefore, metaheuristic methods are used to find an acceptable solution within reasonable computational time. The problem addressed here provides the motivation for this work to develop a metaheuristic method to solve clustering problems. Most of previous techniques are based on K-means algorithm. However, K-means algorithm is highly depends on the initial state and very fast to get trap in local optimal solution. This work presents a modified tabu search approach to solve clustering problems that consists of two phases i.e., a constructive phase to generate an initial solution using K-means algorithm and an improvement phase, where a modified tabu search is employed with an aim to enhance the solution quality obtained from the first phase. In this study, a double tabu lists is introduced where the first tabu list is used to keep the neighbourhood structures and the second tabu list is used to keep the moves that are involved in that particular neighbourhood structure. The performance of the proposed algorithm is tested on five well-known datasets. Preliminary computational experiments are encouraging and compare positively with both the K-means and a standard tabu search alone.

AB - Clustering is a data mining technique used to classify a number of objects into k clusters without advance knowledge such that the distance between objects within a same cluster and its center is minimised. The modelling for such problems is quite complex and searching for optimal solution usually impractical in term of computation times. Therefore, metaheuristic methods are used to find an acceptable solution within reasonable computational time. The problem addressed here provides the motivation for this work to develop a metaheuristic method to solve clustering problems. Most of previous techniques are based on K-means algorithm. However, K-means algorithm is highly depends on the initial state and very fast to get trap in local optimal solution. This work presents a modified tabu search approach to solve clustering problems that consists of two phases i.e., a constructive phase to generate an initial solution using K-means algorithm and an improvement phase, where a modified tabu search is employed with an aim to enhance the solution quality obtained from the first phase. In this study, a double tabu lists is introduced where the first tabu list is used to keep the neighbourhood structures and the second tabu list is used to keep the moves that are involved in that particular neighbourhood structure. The performance of the proposed algorithm is tested on five well-known datasets. Preliminary computational experiments are encouraging and compare positively with both the K-means and a standard tabu search alone.

KW - Clustering

KW - Double tabu list

KW - K-means

KW - Modified tabu search

UR - http://www.scopus.com/inward/record.url?scp=81455148186&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=81455148186&partnerID=8YFLogxK

U2 - 10.3923/jas.2011.3447.3453

DO - 10.3923/jas.2011.3447.3453

M3 - Article

VL - 11

SP - 3447

EP - 3453

JO - Journal of Applied Sciences

JF - Journal of Applied Sciences

SN - 1812-5654

IS - 19

ER -