Application of connectionistic models for the evaluation of market structures

An often occurring problem in marketing research is the consumers' segmentation by means of so-called conjoint data. In this article different approaches based on neural networks are presented and evaluated within a procedure comparison. Observations of a Monte-Carlo-Simulation are used as data basis. The approaches will be compared to the traditional k-means method and a mixture regression analysis and parallels between these procedures will be pointed out. Why these methods lead to identical results will be shown and explained.