Finding latent groups of customers via the poisson mixture regression model
Keiji Takai, Katsutoshi Yada · 2011
Due to developments in technology, movement data tracking a customer's movements in a supermarket in addition to conventional POS data are now available. A problem in analyzing such data is that an ordinary statistical model assuming customer homogeneity does not fit well to such data. In this article, we propose a framework for analyzing such data in a collection of supermarket departments. The framework is based on the mixture regression model assuming the customers' heterogeneity. By the model, we find the latent homogenous groups of the customers and explain the number of items by a stationary time based on the regression model in each latent group. The method of the mixture regression model is explained in addition to the estimation method. We found that a small number of the customers buy more items by going to the supermarket departments and are more sensitive to the stationary time, while a large number of the customers buy less and are less sensitive.