The final step is to duplicate the book name in Borrow Table in Book Table. On the basis of all above data preparations, it is time to establish reader books’ two-dimensional matrix model now. Following the initialization process and the filling in the two-dimensional matrix in the borrowing records, the algorithm continues to seek nearest neighbor set. As for the user-based collaborative filtering calculation, the prime objective is to derive a user set similar to target users. The thesis hereby defines more than 20 neighbor users (or sets up the number of other neighbors according to requirements). User similarity is defined in line with the number of borrowed books (Fig. 3).