Online customer journeys are at the core of e-commerce systems and it is therefore important to model and understand this online customer behaviour. Clickstream data from online journeys can be modelled using Markov Chains. This study investigates two different approaches to handle missing transition probabilities in constructing Markov Chain models for purchase prediction. Imputing the transition probabilities by using Chapman-Kolmogorov (CK) equation addresses this issue and achieves high prediction accuracy by approximating them with one step ahead probability. However, it comes with the problem of a high computational burden and some probabilities remaining zero after imputation. An alternative approach is to smooth the transition probabilities using Bayesian techniques. This ensures non-zero probabilities but this approach has been criticized for not being as accurate as the CK method, though this has not been fully evaluated in the literature using realistic, commercial data. We compare the accuracy of the purchase prediction of the CK and Bayesian methods, and evaluate them based on commercial web server data from a major European airline.
Lin, Jihao; Holland, Christopher P.; and Argyris, Nikolaos, "IMPUTING OR SMOOTHING? MODELLING THE MISSING ONLINE CUSTOMER JOURNEY TRANSITIONS FOR PURCHASE PREDICTION" (2022). ECIS 2022 Research-in-Progress Papers. 52.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.