Rating Matrix Prefilling Algorithm Based on Users' Social Strength
Xiujin Shi, Chunli Zhang · 2014
In order to solve the problem of personalized recommendation in social network, a collaborative filtering algorithm based on users' social relationship mining was proposed with the social network analysis method. Mobile devices and location-based-services have generated rich datasets of people's location information at a very high fidelity. In particular, we employed an entropy-based model (EBM) that not only infers social connections but also estimated the strength of social connections. And obtaining the most similar set of users based on the degree of similar relationship between the users and then calculating unrated items to prefill the original rating matrix. Then producing recommendation used user-based collaborative filtering on the basis of filled rating matrix. Experimental results showed that this algorithm could effectively alleviate data sparsity problem in collaborative filtering and had higher recommendation efficiency.