Free Access
Volume 14, Number 1, 2000
Neural Models in Economy and Management Science
Page(s) 93 - 108
DOI: 10.1051/ejess:2000111

European Journal of Economic and Social Systems 14 N$^\circ$ 1 (2000) 93-108

A bio-mimetic approach to marketing segmentation: Principles and comparative analysis

Jacques-Marie Aurifeille

The FACIREM & GREGEOI labs., University of La Reunion.


A regression algorithm is proposed for partitioning a population into clusters characterised by homogeneous models and predictions. This algorithm, Typren, is based on the hybridisation of a genetic algorithm with linear regression. An empirical illustration is provided, using real marketing data, which compares Typren with the fuzzy clustering approach, Glimmix, based on a regression mixture model. Typren provides better predictivity and better within-cluster homogeneity of predictions. However, the results are slightly less robust compared to Glimmix.

Keywords: Segmentation, clusterwise regression, genetic algorithms

Copyright EDP Sciences 2000