Merging into social trust personalized friend recommendation algorithm

Online social networks have provided the high quality of friend recommendation to assist users in making effective selection decisions from plenty of choices in recent years. In these applications, collaborative filtering is a widely used technique to generate recommendations based on the ratings of like-minded people. However, these recommendation methods are low accuracy for the sparsity of data and cold start. To address these problems, we propose a novel personalized recommendation algorithm to merge social trust information in providing recommendations. Firstly, the trusted neighbors of the active user are identified and aggregated, and trust information can be applied for getting more trusted neighbors by trust propagation. Secondly, a new rating of user's neighbor can be formed to represent the preferences of the active user through merging into trust factor and using the improved collaborative filtering algorithm. In the end, the recommendation set of active user can be acquired from the highest predicted rating. Experimental results based on Epinions data set demonstrate that our method can enhance the effectiveness and accuracy of recommendation compared with the traditional friend recommendation algorithm. Copyright © 2014 Binary Information Press.