The Research Of Improved Film Recommendation Algorithm Based On Collaborative Filtering

With the rapid development of Internet technology,information overloading is increasingly serious.The research of personalized recommendation system has become an inevitable trend.Personalized movie recommendation system is mainly composed of user modeling module,recommending object modeling module and recommendation algorithm module.Recommendation algorithm is the core of personalized recommendation system.The most widely used recommendation algorithm is the collaborative filtering recommendation algorithm.But there are some inevitable problems like data sparsity,cold start in the collaborative filtering recommendation algorithm.On account of the data sparsity,our passage proposes an algorithm called NUBCF,which considers the rareness of user common score projects and the absence of user demand information.In this paper,the main research work and innovation are as follows:Firstly,based on the research of the existing literature,this paper cardings the research status of personalized recommendation system and expounds the related concepts,theory,technology of personalized recommendation and common personalized recommendation algorithms,thus forms a more comprehensive understanding of the field.Secondly,this passage proposes an improvement method to compute the similarity on account of the rareness of user common score projects and introduces the punishment function into the similarity computation of UBCF recommendation algorithm.This method decreases the similarity of users whose common score projects are little,so that it makes the calculation of similarity more reasonable.Thirdly,this paper put forwards the improved method of similarity calculation according to the absence of user demand information and considers the demographic information into the similarity computation of UBCF recommendation algorithm.This method fuses the population information to the similarity of the traditional similarity calculation,finding nearest neighbor users that are more similar with the target users,letting higher quality of the recommendation;Finally,this paper put forwards NUBCF in view of the film personalized recommendation system.NUBCF recommendation algorithm takes into account the rareness of user common score projects and the absence of user demand information.By bringing in punishment function and demographic similarity when calculating the similarity of users,this algorithm comes up with the new user similarity computing method.Experiments show that mean absolute error and F1 of NUBCF recommendation algorithm are lower than UBCF algorithm's.On the same computational complexity of UBCF recommendation algorithm and NUBCF recommendation algorithm,the latter behaves better.