Collaborative Filtering Recommendation Algorithm Based on Matrix Factorization and User Nearest Neighbors
Zhongjie Wang, Nana Yu, Jiaxian Wang · Communications in computer and information science · 2016
The disadvantage of the traditional CFAbMD algorithm is no consideration of impact of local users’ neighbor on item rating. Aiming at this problem, a new CFAbMD algorithm is proposed considering both ALS matrix factorization and user nearest neighbor (CFAbMD-UNN), which integrates the similarity information among users into the matrix factorization of model. Furthermore, the CFAbMD-UNN algorithm was implemented in parallel on Spark. Experiments on Movielens shows that the propsosed CFAbMD-UNN algorithm outperforms the traditional CFAbMD algorithm.