In CARS, these long-term preferences and user interests are usually expressed as ratings and are modeled as the function of not only items and users, but also of context. In other words, ratings are defined with the rating function: R : User ×Item ×Context → Rating . In this approach, there is a pre-defined finite set of contextual types in a given application and each of these types has a well-defined struc- ture.