The hedonic prices method has been proposed to determine the price of a complex product, with several separable attributes, or to evaluate its relative price. It is based on the evaluation of a separate price for each component of the bundle with a statistical approach of market equilibrium. This method is used for automobile and for durable goods as well (computers, apartments...). One of the major drawbacks is the insufficient facial validity of estimated prices, due to a multicolinearity problem. In this paper, we analyze and compare the practical interest of two neural multilayer networks versus linear regression. These results lead us to the conclusion that there is no obvious superiority of neural networks in spite of their higher adaptability to cope with the interactions between the characteristics.
DESMET, P. and HENDAOUI, F. (2000). La relation prix-qualité dans l'automobile : comparaison de méthodes d'estimation des prix hédoniques. Revue Française du Marketing, pp. 167-179.