Knowledge Discovery by Genetic Fuzzy Systems Applied to Consumer Behavior Modelling

Consumer behaviour discipline has made traditionally use of models to understand consumers. Thus, following the scientific method, marketing academics usually pose theoretical models which are subsequently tested by means of several statistical methods. When such models are complex –i.e. several dependent and independent constructs with multiple relations among them– the method usually used for estimating it is Structural Equation Modelling (SEM). In this sense, the hegemony of SEM for estimating this kind of consumer models has been quite obvious during the last decades. However, we think that this method presents some lacks which constraints its usefulness beyond an academic framework; i.e. it is useful to test models, though results provided by SEM are not good enough for being the kind of support that marketing managers need for guiding their market decisions. Thus, the main motivation of this paper is caused by a strong belief in the necessity that marketing modelling analytical methods have to evolve, considering the application of other tools of analysis more appropriate to aid the marketing managers’ decisional processes. This paper briefly presents a brand new methodology to be applied in marketing (causal) modeling. Specifically, we apply it to a consumer behavior model used for the experimentation. The characteristics of the problem (with uncertain data and available knowledge from a marketing expert) and the multiobjective optimization we propose make genetic fuzzy systems a good tool for tackling it. In sum, by applying this methodology we obtain useful information patterns (fuzzy rules) which help to better understand the relations among the elements of the marketing system (causal model) being analyzed; in our case, a consumer model.