作者: A. Vellido , P. J. G. Lisboa , K. Meehan
DOI: 10.1007/10720076_43
关键词:
摘要: The neural network-based Generative Topographic Mapping (GTM) (Bishop et al. 1998a, 1998b) is a statistically sound alternative to the well-known Self Organizing Map (Kohonen 1982, 1995). In this paper we propose GTM as principled model for cluster-based market segmentation and data visualization. It has capability define, using Bayes’ theorem, posterior probability of cluster/segment membership each individual in sample. This, turn, enables be used perform different levels detail or granularity, encompassing aggregate one-to-one micro-segmentation. definition that also makes tool fuzzy clustering/segmentation. capabilities are illustrated by case study real-world Internet users opinions on business-to-consumer electronic commerce.