Research On Collaborative Filtering Of Personalized Recommendation Based On Users' Interest

With the rapid development of the Internet,the results which are presented in explosive style and the advertising bombardment makes people drown in massive information repository and become confused in the online world.With the continuous development of information filtering technology,such ways of presentation of content are gradually being replaced by personalized recommendation.Personalized recommendation tracks and analysis people's Internet consumption,social networking activity,and then dig out the user's potential interests and needs.In online shopping site,personalized recommendation can effectively help users find and selection the goods.For the manufacturer,people who apply personalized recommendation will surely have the priority right to speak to customers and hold the market.In all technology of personalized recommendation,collaborative filtering technology has the most satisfactory and reliable recommendation.But along with the rapid popularization of mobile Internet,almost everybody has a smart terminal,which makes that people use the network in rich scenes of life,resulting in continuous influx and rapid growth of network users and the amount of data resources.This makes the traditional collaborative filtering algorithm not only faced with the classic data sparse,cold start and other problems,also faces the challenge of mining user interest.To deal with the problem that traditional collaborative filtering algorithm is not well integrated into the user's interest,we propose the degree of entropy difference about the interest among users,designed to put the user's own interest into the personalized recommendation algorithm.According to the new keywords,we proposed an improved collaborative filtering algorithm based on the difference of the interest between the users;in addition,we improved the traditional calculation method of the similarity by drawing 0-1 knapsack problem,proposing another improved collaborative filtering algorithm.Then,the results of the experiment that we conducted show that the two improved algorithm proposed in this paper can accurately portray the interest of users themselves,but also effectively alleviating the data sparseness problem,thereby improving the accuracy of recommendation algorithm.Finally,we conducted experiments with the Movielens data,we compares the new collaborative filtering algorithm proposed by us with traditional collaborative filtering algorithm,the experimental proves that:the new collaborative filtering algorithms proposed by us have a better accuracy.