Recommendation Method Study Based on User's Page Interest Degree

Mining users' interest information and recommending corresponding products to them is a goal that many websites pursue. This paper focuses on the degree of interest in a page by the user. The proposed technique generates a user's page interest degree matrix. It is assumed that this matrix will be a sparse matrix with null entries in a column. These null entries are updated with the average value of the column to create a not-sparse matrix against which a Singular Value Decomposition (SVD) is run. The Slope One Algorithm (SOA) is run against this generated matrix to improve predicting accuracy. Finally, encity.com website data was used to test the results by comparing the sparse generator matrix with the not-sparse generator matrix after SVD under Slope One. It was found that the latter is better than the former in predicting user's page interest degree's accuracy.