Research And Application Of Mobile Personalized Recommendation Based On Collaborative Filtering

With the rapid development of mobile Internet technology,more and more users through mobile devices to get internet services.The mobile e-commerce recommender system with features of mobile and personalization is the new direction of electronic commerce.The number of E-commerce's items and users are in a large number grow,how to quickly find items to meet the interest of users is the focus of attention.Compared to the traditional e-commerce,mobile e-commerce user context is extremely diverse and rich,the users' requirement become more diverse,personalized and mobile,so people expect to get more personalized service.So how to support recommend personalized service to users to meet their interests and context is a challenging research topic.In this thesis we design a hybrid personalized recommendation algorithm based on item and context similarity for the characteristics of mobile e-commerce,to design a mobile e-commerce personalized recommendation.Firstly,after the analysis of the current mobile e-commerce personalized recommendations theories and methods,combined with user interests and mobile context scenario,considering the changes of user interest,a new model of user interest has been proposed.Compared with the traditional recommendation model,this model has greatly reflect the interest with different context scenario.Secondly,on focus of the current user rating data sparseness problem,this paper uses a new missing values filling algorithm based on the scenario category.This method considered the users interest under different scenarios which can effectively improve the recommendation accuracy.Thirdly,The values data of mobile e-commerce recorder system includes the context scenario which is different with the traditional e-commerce,so we adopt an improved user model.With the base of the new model we proposed a correction of scenario similarity and item similarity.The neighbors are generated by the similarity.we uses a hybrid collaborative filtering recommendation UC-CF algorithm,it can improve the recommend accuracy.Finally,we verify the improvement of the new algorithm's performance and new model.Experimental results show that the precision of recommend based on the data with missing data filling has been improved and the algorithm is more accurate compared with traditional algorithms.This study promotes the development of related personalized recommendation technology in the mobile e-commerce applications,and provides a strong practical application of theoretical and methodological support.