Memory biased random walk approach to synthetic clickstream generation

Personalized recommender systems rely on personal usage data of each user in the system. However, privacy policies protecting users' rights prevent this data of being publicly available to a wider researcher audience. In this work, we propose a memory biased random walk model (MBRW) based on real clickstream graphs, as a generator of synthetic clickstreams that conform to statistical properties of the real clickstream data, while, at the same time, adhering to the privacy protection policies. We show that synthetic clickstreams can be used to learn recommender system models which achieve high recommender performance on real data and at the same time assuring that strong de-minimization guarantees are provided.