Estimating product-choice probabilities from recency and frequency of page views

A latent-class model for estimating product-choice probabilities from clickstream data

“…This has many applications in statistics, operations research, and image processing [31]. Iwanaga et al [20] recently used the maximum likelihood method to estimate product-choice probabilities subject to the monotonicity, convexity, and concavity constraints with respect to recency and frequency. Their shape-restricted model was a new effective application of shape-restricted regression to the analysis of clickstream data.…”

“…This section introduces the optimization models developed by Iwanaga et al [20] for estimating productchoice probabilities from the recency and frequency of each customer's previous PVs.…”

“…The recency and frequency of a customer's previous purchases have been shown to be key indicators for forecasting repeat purchases [10,11,21,33,34]. In light of this fact, Iwanaga et al [20] recently devised optimization models for estimating product-choice probabilities from the recency and frequency of each customer's previous PVs. These models exploit the properties of recency and frequency of PVs to enhance their predictive performance.…”

“…In particular, the monotonicity-convexity-concavity (MCC) model estimates product-choice probabilities that satisfy monotonicity, convexity, and concavity constraints with respect to recency and frequency. Iwanaga et al [20] demonstrated that higher predictive performance is achieved with MCC model than with the standard general-purpose methods for binary classification, namely, logistic regression and kernel-based support vector machines.…”

“…Section 2 gives a brief review of related work. Section 3 presents the MCC model [20] for estimating product-choice probabilities from clickstream data. Section 4 develops our latent-class MCC model and EM algorithm.…”