Collaborative Filtering Recommendation Algorithm Incorporating Multi-Criteria And Context Information

The personalized recommender system as a important information filtration mean is a potential method to solve the problem of information overload currently. And collaborative recommendation is the most widespread and successful recommendation technology in personalized recommender systems. With scale expansion of electronic commerce system and sharp growing of user number and item number, the traditional collaborative filtration recommendation algorithms, unabling to synthetic using multi-criterion ratings and context information to generating recommendation, causes the lower precision of recommendation information. This paper has further conducted deep research of collaborative filtration recommendation algorithm incorporating multi-criteria ratings and context information.Firstly, aim at the problem that the traditional collaborative filtering recommendation algorithm can not recommend with multiple criteria, by introducing the concept of multi-criteria rating for extending the standard collaborative filtering algorithm, a multi-criteria recommendation algorithm based on Widrow-Hoff neural network is proposed. Based on the outstanding characteristics of Widrow-Hoff least mean square (LMS) algorithm on the process of system identification to match with highly accuracy, an approach to compute user preferences eigenvector based on Widrow-Hoff LMS algorithm is proposed. Measuring users'similarity by adopting the user preferences eigenvector and spatial distance matrix so as to locate a neighbor set for the best recommendations, finally improving efficiently the precision of recommendation.Secondly, aim at defects and deficiencies of existing context-aware recommendation algorithms, a context-aware recommendation algorithm based on Fuzzy C-Means clustering is proposed. First, the history context information are clustered by history context discovery algorithm in order to produce clusters and membership matrix; then the aware contexts are matched with clustering centers of history context information, and non-membership data is mapped to membership data by adopting membership degree as mapping coefficient; finally choose rating matrix which conforms to the activity context information.Finally, we have compared the performance between the proposed algorithms and other algorithms, and the validity of the proposed algorithm is confirmed.