K-Means vs. Topology Representing Networks: Comparing Ease of Use for Gaining Optimal Results With Reference to Data Input Order

• Ganglmair, Alexandra ^[1] ; Wooliscroft, Ben ^[1]
  1. [1] University of Otago

     University of Otago

     Nueva Zelanda

     
• Localización: Tourism analysis, ISSN 1083-5423, Vol. 5, Nº. 2-4, 2000, págs. 157-162
• Idioma: inglés
• Texto completo no disponible (Saber más ...)
• Resumen

  • K-Means clustering is the most popular statistical method used for market segmentation, and SPSS is the most popular statistics program in the world. It has been suggested that Neural Networks may offer superior segmentation solutions over K-Means and other classical cluster analyses. This research compares the stability of segmentation solutions of both K-Means and TRN, a recent development in Neural Network algorithms, when input order of data is changed. TRN was found to produce considerably more stable segmentation solutions without extensive use of other subjective analysis techniques, offering decision makers a more consistent picture of their market.