This paper proposes an approach for integrating social networks into intelligent tutoring systems (ITS), especially for predicting student performance (PSP). Firstly, we review the main concepts of the ITS as well as matrix factorization technique and social matrix factorization in Recommender Systems (RS). Next, this work introduces how to map the ITS data into the context of user, item and rating in the RS, and shows how to convert social networks to relationships matrix. Then, the prediction model, which is integrated social networks, is proposed for building the ITS. Experimental results on two real datasets show that integrating social relationships of students can be able to increase the accuracy of the prediction models