Probabilistic sequence modeling for recommender systems

Probabilistic topic models are widely used in different contexts to uncover the hidden structure in large text corpora. One of the main features of these models is that generative process follows a bag-of-words assumption, i.e each token is independent from the previous one. We extend the popular Latent Dirichlet Allocation model by exploiting a conditional Markovian assumptions, where the token generation depends on the current topic and on the previous token. The resulting model is capable of accommodating temporal correlations among tokens, which better model user behavior. This is particularly significant in a collaborative filtering context, where the choice of a user can be exploited for recommendation purposes, and hence a more realistic and accurate modeling enables better recommendations. For the mentioned model we present a fast Gibbs Sampling procedure for the parameters estimation. A thorough experimental evaluation over real-word data shows the performance advantages, in terms of recall and precision, of the proposed sequence-modeling approach. Copyright © 2012 SciTePress - Science and Technology Publications.