One of the fundamental methods used in collaborative filtering systems is Correlation based on K-nearest neighborhood. These systems rely on historical rating data and preferences of users and items in order to propose appropriate recommenders for active users. These systems most of the times do not have a complete matrix of input data. Exact matrix completion technique tries to predict unknown values in data matrixes. As the correlation method deals with sparse data matrixes, and this challenge leads to a decrease in the accuracy of recommendation for new users, as a result using the matrix completion technique as a preprocessing has many advantages. The main advantages of proposed method in this paper are the higher prediction accuracy and an explicit model representation. The result of experiments shows that significant improvement in prediction accuracy can be achieved over other substantial methods
