Neural networks and the multinomial logit for brand choice modelling: a hybrid approach

Abstract

The study of brand choice decisions with multiple alternatives has been successfully modelled for more than a decade using the Multinomial Logit model. Recently, neural network modelling has received increasing attention and has been applied to an array of marketing problems such as market response or segmentation. We show that a Feedforward Neural Network with Softmax output units and shared weights can be viewed as a generalization of the Multinomial Logit model. The main dierence between the two approaches lies in the ability of neural networks to model non-linear preferences with few (if any) a priori assumptions about the nature of the underlying utility function, while the Multinomial Logit can suer from a speci®cation bias. Being complementary, these approaches are combined into a single framework. The neural network is used as a diagnostic and speci®cation tool for the Logit model, which will provide interpretable coecients and signi®cance statistics. The method is illustrated on an arti®cial dataset where the market is heterogeneous. We then apply the approach to panel scanner data of purchase records, using the Logit to analyse the non-linearities detected by the neural network.

Deep Neural Networks for Choice Analysis: A Statistical Learning Theory Perspective

“…This is generally true; however, in the case of DNN, an accurate estimator of the choice probability function ŝ(x) could satisfy most of our interpretation purposes traditionally achieved through using ŵ. In fact, several studies in the transportation field have visualized or computed the gradient information of the choice probability functions to interpret the ML classifiers, supporting our definition of the interpretation loss based on the choice probability functions [53,9,27]. Moreover, the process of interpretting elasticity ds dx j is the same as the discussion of using input gradients in the ML community [3,43].…”

Deep Neural Networks for Choice Analysis: Extracting Complete Economic Information for Interpretation

“…It is often referred to by different names such as sensitivity analysis, saliency, or attribution maps in computer vision [84,51], or attention mechanism in natural language processing [103]. In the transportation field, some studies used input gradients to describe the relationship between inputs and outputs in DNNs [13,76,37]. Recently, researchers in the ML community are increasingly focusing their attention on the properties of DNNs' input gradients, owing to their importance in DNN interpretation [87,82,90,79].…”

Performance Evaluation of Bagged RBF Classifier for Data Mining Applications

“…Bentz and Merunkay also showed that Copyright © 2013 MECS I.J. Information Engineering and Electronic Business, 2013, 5, 49-56neural networks did better than multinomial logistic regression(Bentz 2000).…”