Comparison of Four Initialization Techniques for the K-Medians Clustering Algorithm

  • A. Juan
  • E. Vidal
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1876)


Clustering in Metric Spaces can be conveniently performed by the so called k-medians method. It consists of a variant of the popular k-means algorithm in which cluster medians (most centered cluster points) are used instead of the conventional cluster means. Two main aspects of the k-medians algorithm deserve special attention: computing efficiency and initialization. Efficiency issues have been studied in previous works. Here we focus on initialization. Four techniques are studied: Random selection, Supervised selection, the Greedy-Interchange algorithm and the Maxmin algorithm. The capabilities of these techniques are assessed through experiments in two typical applications of Clustering; namely, Exploratory Data Analysis and Unsupervised Prototype Selection. Results clearly show the importance of a good initialization of the k-medians algorithm in all the cases. Random initialization too often leads to bad final partitions, while best results are generally obtained using Supervised selection. The Greedy-Interchange and the Maxmin algorithms generally lead to partitions of high quality, without the manual effort of Supervised selection. From these algorithms, the latter is generally preferred because of its better computational behaviour.

Key words

Clustering Metric Spaces K-Medians algorithm K-Medians initialization Greedy-Interchange algorithm Maxmin algorithm 


  1. 1.
    R.O. Duda and P.E. Hart. Pattern Classification and Scene Analysis. Wiley, 1973.Google Scholar
  2. 2.
    E. Granum and M. G. Thomason. Automatically Inferred Markov Network Models for Classification of Chromosomal Band Pattern Structures. Cytometry, 11, 1990.Google Scholar
  3. 3.
    A. K. Jain and R. C. Dubes. Algorithms for Clustering Data. Prentice Hall, 1988.Google Scholar
  4. 4.
    A. Juan and E. Vidal. Application of the K-Nearest-Neighbour Rule to the Classification of Banded Chromosomes. In ICPR’00 (accepted).Google Scholar
  5. 5.
    A. Juan and E. Vidal. Fast k-means-like clustering in metric spaces. Pattern Recognition Letters, 15(1):19–25, 1994.CrossRefGoogle Scholar
  6. 6.
    A. Juan and E. Vidal. Fast Median Search in Metric Spaces. In A. Amin, D. Dori, P. Pudil, and H. Freeman, editors, Advances in Pattern Recognition, volume 1451, pages 905–912. Springer-Verlag, 1998.Google Scholar
  7. 7.
    O. Kariv and S. L. Hakimi. An Algorithmic Approach to Network Location Problems. II: The p-Medians. SIAM Journal on Applied Math., 37(3):539–560, 1979.zbMATHCrossRefMathSciNetGoogle Scholar
  8. 8.
    I. Katsavounidis. A new initialization technique for generalized Lloyd iteration. IEEE Signal Processing Letters, 1(10):144–146, October 1994.Google Scholar
  9. 9.
    C. Lundsteen, J. Philip, and E. Granum. Quantitative analysis of 6985 digitized trypsin G-banded human metaphase chromosomes. Clinical Genetics, 18, 1980.Google Scholar
  10. 10.
    Mirchandani P.B. and Francis R.L. editors. Discrete Location Theory. Wiley, 1990.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2000

Authors and Affiliations

  • A. Juan
    • 1
  • E. Vidal
    • 1
  1. 1.Institut Tecnològic d’InformàticaUniversitat Politècnica de ValènciaValènciaSpain

Personalised recommendations