Matrix Factorization Recommendation Algorithm Based On User Clustering And Mobile Context

With the development of mobile communication network technology,mobile devices gradually become the main platform of network services and information sources.It is difficult for users to find the valuable information which attracts them in With the development of mobile communication network technology,mobile devices gradually become the main platform of network services and information sources.While mobile users enjoy the convenience of mobile devices,they also are faced with the information overload problem.Recommendation system is a "tool" that solves information overload and helps users make decisions.In mobile environment,the context factor has become an important influencing factor of mobile user decision-making,so we can't apply directly traditional internet recommendation algorithm.In such a scenario,how to combine mobile context for improving the accuracy of the recommendation system and user satisfaction that has become a hot spot and difficult thing.In order to more accurately and quickly find the service of users interested and aim at the puzzle resulted from data sparsity,ignoring context similarity and context similarity measurement and fully considers mobile context information,this paper proposes the method of improved matrix factorization recommendation algorithm based on user clustering and mobile context: Context MF and UCMC MF.Experimental results on real datasets demonstrates that the proposed algorithm can effectively improve the accuracy of prediction.Based on the algorithm,a place recommendation prototype mobile APP is designed.The main work of this paper is as follows:(1)This dissertation firstly describes the background and situation of this topic,and then puts forward the solution for the primary issue of study on context-based recommendation.At last,this paper introduces the working content and purpose.(2)Aiming at the difficulty of data sparsity and ignoring the context similarity in personalized recommendation systems,a new matrix factorization recommendation algorithm(Context MF)using users clustering and mobile context was presented.(3)Aiming at the difficulty of data sparsity in personalized recommendation systems,we present a new matrix factorization recommendation algorithm based on users clustering and mobile context factors(UCMC MF).UCMC MF firstly uses user scores for clustering,and uses matrix factorization recommendation algorithm based on context set limit the target users score by the context of similar users in each cluster,and then modify the loss functions.Therefore the objective function was established based on context,which was solved by gradient mehod.The matrix factorization recommendation algorithm based on user clustering and context first clusters user according to user preference first at all,and then finds context similar users in similar cluseters,using the way of context MF predicting user ratings.Finally the experiment indicate that the recommendation model improves the precision and coverage of recommendation effectively.(4)Using UCMC MF algorithm desiges a context travel APP for users who need.This paper introduces the design of APP background demand,system design and final prototype implementation.