Research And Application Of Hybrid Recommendation Algorithm Based On Content And Collaborative Filtering

In the era of big data,people on the one hand enjoy the dividend brought by big data,and various platforms provide users with a variety of choices to enrich people's lives.However,on the other hand,users also have to face the challenge of information explosion.How to accurately select personalized information in line with their own needs from a wide range of data has become an urgent problem to be solved,and personalized recommendation system is born.In the development course of personalized recommendation system,the most important is its core content--the research and application of recommendation algorithm.Content-based recommendation algorithm and collaborative filtering algorithm are the mainstream recommendation algorithms.The content-based recommendation algorithm can make full use of the attributes and characteristics of the recommendation object to make recommendations,but its recommendation results are limited and can not find out the potential interests of users.Collaborative filtering recommendation algorithm has relatively stable performance,but there are also problems such as data sparsity and "cold start" caused by difficulty in recommending new items without user rating.In addition,the core content of traditional collaborative filtering algorithm is the calculation of similarity,but some commonly used methods of similarity calculation usually do not consider the overlap degree between two user rating vectors and the attenuation of user interest.Based on the advantages and disadvantages of the above mainstream recommendation algorithms,this paper proposes a hybrid recommendation algorithm(CB-TUB)based on content and collaborative filtering to improve the performance of the algorithm.The specific work of this paper includes: first,for the problem of "cold start" of the project,describe the project content with the feature "label",and make content-based recommendation.Secondly,when the user-based collaborative filtering recommendation algorithm is used for recommendation,the user-rating matrix is constructed through the implicit rating of the project by the user,so as to alleviate the problem of data sparsity.In the calculation of similarity,on the one hand,the time weight of users' scoring is introduced,on the other hand,the distance of jackard unified operator is introduced to measure the proportion of common scoring items among users.Finally,the paper creatively puts forward the "user interest index" and gives the calculation method.When the algorithm is mixed,the weighted user interest index of the recommended items is sorted from high to low to generate top-n recommendation for users.This paper uses the TV program data set in question B of 2018 Teddy-Cup to verify the performance of the hybrid algorithm.It is found that after weight mixing,the hybrid algorithm has improved in accuracy,recall rate and score compared with the traditional single algorithm,and has the advantage of recommendation.This paper applies the hybrid algorithm to the personalized recommendation of TV programs.The performance of the hybrid algorithm was experimentally simulated using the TV program data set in question B of the 2018 teddy cup.It is found that after weight mixing,the hybrid algorithm has improved in accuracy,recall rate and score compared with the traditional single algorithm,and has the advantage of recommendation.