Research On Video Recommendation Method Based On Collaborative Filtering Technology

The information revolution brought by intelligent Internet technology reshapes the traditional video service industry.Internet is an important part of human life,with the continuous development of its technology,massive video resources continue to fill human life.People cannot get information that they want quickly from massive video resources,which leads to the information explosion.In order to solve the problem of information overload,video service platform introduces recommendation system module.The emergence of recommendation technology greatly promotes the interaction between users and information service platform,and improves user stickiness.Among the basic recommendation algorithms,collaborative filtering algorithm is based on the idea of group wisdom,which can achieve personalized needs on the basis of users.However,traditional collaborative filtering still has its defects,such as extremely sparse data and user interest deviation.In addition,with the rapid increase of users and video resources,the problem will be more serious.Combined with the above background,the work of this thesis mainly includes the following points:(1)To solve the problem of sparse user ratings,this thesis introduces user preference information for video categories.According to the user’s original rating data and the corresponding relationship between videos and categories,a category preference matrix is constructed to effectively reduce the dimension of the data.On this basis,considering the user’s explicit rating information and the implicit meaning between the user and the category,the user similarity between the rating matrix and the category preference is weighted.At the same time,in order to improve the searching efficiency of the nearest neighbor in the calculation process,the k-means algorithm based on the minimum spanning tree is introduced into the category preference matrix to obtain the target user’s cluster class,which is used as the search space of the target user’s nearest neighbor set.(2)Aiming at the problem of user preference deviation,the time weight function is introduced under the inspiration of forgetting curve rule.The longer it is from now,the smaller the weight will be given.Under this idea,the original scoring prediction formula is improved.Finally,a new hybrid collaborative filtering algorithm based on time weight function is obtained by combining the time weight algorithm with the collaborative filtering algorithm based on category preference.(3)In order to verify the performance of the proposed algorithm,several experiments are carried out on Movie Lens data set.Experimental results show that,compared with the traditional collaborative filtering,the hybrid collaborative filtering algorithm based on the time weight function proposed in this thesis can reduce the sparsity problem and improve the recommended accuracy to some extent.(4)According to the previous research and experiments,a video recommendation system is built to verify the effectiveness of the improved algorithm in practical application.