Research On Personalized Film Recommendation Method Based On Deep Learning

Watching online movies has gradually become a part of entertainment life.How to provide users with videos that satisfy their preferences has become the focus of research on various film and television websites.Currently,the most widely used recommendation algorithm is collaborative filtering algorithm.In recent years,with the development of deep learning technology,applying the deep learning method to the recommendation system has become one of the current research hotspots.Traditional collaborative filtering algorithms have data sparseness problems.In the current research of recommendation system,the hybrid recommendation method is generally adopted.The main method is to extract features from text content and then populate the collaborative filtering algorithm to alleviate the data sparsity problem.However,when extracting text content,the traditional method of extracting text features is often used.This method can only analyze and calculate a single word,and can not effectively combine the context in the sentence.This paper proposes to use the deep learning model Doc2vec model to extract text features,which can deeply explore the semantic relevance of sentences.In the process of text content feature filling,since the text information does not exist or the text content does not match,the accuracy of the recommendation will be affected.Therefore,the fusion image,text and scoring data information are also proposed.The VGG network model is used to extract the features of the image.The implementation process of the multimodal movie recommendation algorithm is implemented by fusion similarity algorithm which is used to fuse the the similarity between movies obtained by the Doc2vec model training movie introduction information,the similarity between movies obtained by the VGG network model training movie poster information,and the similarity between the movies obtained by the Item-CF algorithm training user rating information.And get the best recommendation results through different fusion ratios.The data used in this experiment is the movieLens dataset,movie introduction information and movie posters.In this paper,a variety of fusion ratios are used to calculate the recommendation effect of the hybrid recommendation algorithm.Experiments show that the recommendation effect of the hybrid recommendation algorithm is improved compared with the traditional collaborative filtering algorithm.