Session modeling to predict online buyer behavior
Andrii Cherniak, Jesse S. A. Bridgewater · 2013
In this paper, we explore the feasibility of long-term prediction of buyer behavior using context-based approach based on variable-length markov chains. We discuss different strategies on event log pre-processing and the impact of this operation on the accuracy of the result. We report our results on the accuracy of the prediction of buyer / seller behavior on eBay marketplace for one week in the future1.