A recommendation algorithm based on the hybrid model

With great growth of information resources on the internet, how to recommend interesting items from mass data to users with different interests and hobbies according to their information characteristics has become an urgent problem to be solved. In this paper, we improved the recommendation algorithms of SVD (singular value decomposition) and KNN (K-nearest neighbor algorithm) by different characterization factors, and then proposed a hybrid algorithm based on improved algorithms. The test results on MovieLens dataset show that the root mean squared error (RMSE) of scoring prediction by using the hybrid recommendation algorithm reduces greatly and prediction accuracy of the recommendation algorithm also increases significantly. © 2019, Politechnica University of Bucharest. All rights reserved.