Kohonen Self-Organizing Maps: A Neural Approach for Studying the Links between Attributes and Overall Satisfaction in a Services Context

Abstract

This research aims at analyzing and understanding the attributes-overall satisfaction links (A-OSL) for a service. To date, marketing managers tend to assume that these links are linear, even though scholars have for at least two decades pointed out that they can often be non-linear as well as asymmetric (Kano, Seraku, Takahashi and Tsuji 1984; Anderson and Mittal 2000). Blindly assuming that these links are linear may lead to serious mistakes in estimating the attribute levels which trigger the highest degree of targeted consumers' overall satisfaction (TCOS). In this article, the A-OSL relationship is explored by using a powerful neural network methodology: Kohonen Self-Organizing Maps (KSOM). KSOM have the ability to infer the functions describing A-OSL from data. This methodology also classifies the input data in relation to prototypes on a topological map by using a Euclidian distance criterion.