| In recent years,the film industry has been paid more and more attention,more and more prosperity.There are hundreds of movie theaters each year.Because there are too many movie programs to choose,it is difficult for users to find out what they are interested in.How to find their favorite film has become a problem in the massive movies.So,the movie recommender system came into being.It is to solve how to find the key point in the movie a lot of information,recommended to satisfy the user's request to the user for a movie recommender system,how to predict whether a given user will love a particular project(prediction problem)or to identify a given user interest set(top-N recommendation)is the key recommender system.In all kinds of movie recommendation system,the content of the movie recommendation algorithm used as input for movie content based on the new film release,can from the "cold start" problem.At the same time,item-based collaborative filtering according to the evaluation of the film all users,found that the similarity between items and items,and then similar items will be recommended to the user according to the user the historical preference information,personalized degree is better.Although the content based and collaborative filtering recommendation has been widely used,but due to some inherent characteristics lead to a series of problems,such as poor timeliness and accuracy is not enough.Aiming at the above problems,this paper combines the advantages of content-based and item-based collaborative filtering,neutralizes the disadvantages of both,and uses neural networks to improve the performance of recommendation.In this paper,we put forward a hybrid movie recommender system based on neural network content extraction and collaborative filtering,which recommend movie to user,in order to obtain a higher matching accuracy,better timeliness and user satisfaction.Firstly,training the neural network model,based on Word2Vec CBOW,the content of information(e.g.director,actor,etc.)as the training data acquisition of each feature vector form elements,and then use the linear relationship to calculate the similarity between each learning feature film;secondly use item-based collaborative filtering method to get the similarity of film;combine the similarities which are trained by neural network and collaborative filtering;finally,according to the similarity between the movies and the historical preference information of the user,the prediction score of the items is obtained,and then the recommendation list is generated for the user.The experiment is based on the MovieLens-hetre data set,and the last experimental results show the validity of the proposed method. |
