Analyzing purchase data by a neural net extension of the multinomial logit model

Abstract

Brand choice models used to analyze households purchase data as a rule have a linear (deterministic) utility function, i.e. they conceive utility as linear combination of predictors like price, sales promotion variables, brand name and other product attributes. To discover nonlinear effects on brands’ utilities in a flexible way we specify deterministic utility by means of a certain type of neural net. This feedforward multilayer perceptron is able to approximate any continuous multivariate function and its derivatives with the desired level of precision. In an empirical study the neural net based choice model leads to better results with regard to both estimation and test data, and implies different choice elasticities.