A New Method of Context-based Multidimensional Collaborative Filtering Recommendation

摘要

Most existing recommender systems offer calculation of recommendation in user-item dimensions, omitting the context information.Context is the dynamic information describing the situation of items and users which affects the users' decisions.This paper proposed a novel collaborative filtering method to take contextual information into consideration.We presented a context-based multidimensional collaborative filtering recommendation model.Using the multidimensional approaches, we introduced how to build the context-based multidimensional users' profile and discussed the parts of the model in details.Furthermore, we elaborated on the contextbased multidimensional collaborative filtering algorithms.We described a new algorithm to figure out the similarities between the previous contexts and the current context of the active user, and designed a new equation to get the context-based ratings of the nearest neighbors of the active user.And then we worked out the ultimate rating of the active user as the aggregate of the ratings of the nearest neighbors.This study hopes that by establishing context-based multidimensional users' profile and using the new algorithms, the recommendation quality will be largely increased.