A personalized product recommendation algorithm based on preference and intention learning

We propose a hybrid learning approach to provide automated assistance for personalized product recommendation. The novel feature of this work is that the system learns and uses models of both user preferences and the user's intentional context. Both learning types are based on the same user input, but elicit different aspects of the user model. User preference is learned via support vector machine (SVM) with user ratings on the products, whereas the user's intentional context is inferred using a hidden Markov model (HMM) from given product access sequences. We propose a product recommendation scheme based on an analysis on both the preference and intentional context model. An empirical analysis shows that the hybrid approach is able to support users with different preference structures and intentional contexts.