Analyzing joint brand purchases by conditional restricted Boltzmann machines

Abstract We introduce the conditional restricted Boltzmann machine as method to analyze brand-level market basket data of individual households. The includes marketing variables and household attributes independent variables. To our knowledge this is first study comparing homogeneous heterogeneous multivariate logit models for across several product categories. explain how estimate starting from a without turns out excel all other investigated in terms log pseudo-likelihood holdout data. interpret selected based on coefficients linking purchases hidden variables, interdependences between brand pairs well own cross effects indicates pairwise relationships brands that are more varied than those model are. Based inferred we determine competitive structure by means cluster analysis. Using counterfactual simulations, investigate what three different (independent logit, machine) imply with respect retailer’s revenue if each put display. Finally, mention possibilities further research, such applying areas or retailing.