Research And Application Of Movie Recommender System Based On Taxonomy Driven Recommendation Algorithm

With the development of e-commerce platform the recommender system has got more and more attention and results.From the early Amazon using artificial recommender system to prepare the recommended list,to the present use of data mining,machine learning and other technologies to form the intelligent recommender system,both of them have constantly improved the recommender system and bring users a revolutionary user experience.With the rapid development of the recommender system,it is no longer confined to the field of e-commerce,but also to the other platforms like Douban reading,Headlines today,Douban movies and so leisure platforms to promote user experience and satisfaction.As for the film recommendation,the Netflix company in America has achieved good results its own online video platform through the competition of getting an efficient score prediction algorithm.There are no strict personalized recommender system for the users in current domestic online video platforms.Generally they just use the relevance among videos to generate the recommended lists.On one hand,this thesis improved the taxonomy-driven recommender system to improve the performance for film recommendation and to generate a recommended list with users' satisfaction.On the other hand,by using the Topic Diversification filter algorithm to handle the recommended lists generated by traditional cooperative filtering we enhance the recommended results of the degree of surprise.On the basis of this,we build the individual recommender system for users.The main work of this thesis has the following points:1)Based on the existing taxonomy-driven recommendation algorithm,the composition of the movie taxonomy tree is improved for the particularity of the film recommendation field and the content of the movie is no longer described by the genres of the movie.Use the various attributes of the film(mainly including the genre of film,the artist,the language,the year of the release,the award and the country)to construct the taxonomy tree to describe the content of the film.Each node in the tree is saved with a movie information,called a Topic.2)According to the movie content of the taxonomy of the tree and the user's viewing situation to calculate the user's interest matrix,no longer used the movie content of all related Topic as an equal consideration in the past,but according to Topic different Type to calculate the user's similarity.3)Based on the improved movie content taxonomy tree,the Topic Diversification algorithm is used to filter the recommended results of traditional collaborative filtering to generate a more recommended list of different degrees of difference4)Using the above improved algorithm,we establish the film recommender system,to combine the algorithm with the practical application...