Influence of outliers on results of cluster analysis Cover Image

Identyfikacja obserwacji odstających w analizie skupień
Influence of outliers on results of cluster analysis

Author(s): Jacek Batóg
Subject(s): Economy
Published by: Wydawnictwo Uniwersytetu Ekonomicznego we Wrocławiu
Keywords: Cluster analysis; k-means method; outliers

Summary/Abstract: The research concerns comparison of methods that enable identifying spatial outliers. The analysis was based on the statistical data describing income and public debt of gminas of Zachodniopomorskie voivodship. All considerations were applied to partitions made by k-means method. Identification of false positive outliers was provided by means of Wang, Zhang, Li and Song method, author’s modification of this method and additionally methods proposed by Kandogan and Hautamäki. The level of objects’ differentiation within group was used as a measure of homogeneity of partitions. The received results were very similar for all considered methods. Some differences occur only in order of indicated outliers.

  • Issue Year: 2016
  • Issue No: 426
  • Page Range: 13-21
  • Page Count: 9
  • Language: Polish