Joint Modeling of Customer Purchasing Behavior Using the Marked Hawkes Process

Abstract

Although customer purchasing behavior is of great interest to both practitioners and researchers, the two primary components, purchase interval and purchase amount, have typically been handled separately. However, considering the tendency of actual purchase behavior, it is desirable to model them simultaneously. Thus, this paper proposes the method for stochastically estimating the joint mechanism of purchase time and amount while considering all the past purchasing behavior by applying the marked Hawkes process, a type of stochastic process, to the context of customer purchasing behavior in a super-market. After estimating the parameters of our model incorporating customers’ random e.ects and covariates such as gender and age group through the EM algorithm, we obtained several managerially important implications. In addition, our simulation revealed that the marked Hawkes process proposed in this paper had an extremely higher predictive power than traditional models.