Research On Movie Recommendation Method Based On Convolution Neural Network And Long Short Term Memory Network

Due to the rapid development of big data technology,the quantity of information on Internet enlarges day by day and the problem of information overload poses a great challenge to us.For domestic user,it may take them too much time to find the effective information that meets the requirements from the huge amount of data.Therefore,the development of the recommendation system is imminent.Actually,the recommendation system can analyze user's characteristics and historical behavior records to get user's interest preference and then recommend to user.There are many common recommendation models,such as content-based(CB),collabora-tive filtering(CF)and hybrid recommendation model.However,there are many shortcomings about traditional recommendation models,such as long running time,insufficient feature extraction,no consideration about user's dynaumic interest.Therefore,based on the dataset of MovieLens,this thesis designs a movie recommendation method based on neural network.Firstly,basic features of users and movies are extracted by simple neural network.Then,Text-CNN is used to extract potential text features of movies which are combined with the basic characteristics of movie to obtain movie feature vector.Next,user feature and movie feature are input into the neural network to predict user's rating on movie and network is trained to save the optimal parameters.Then,the optimal model is used to extract user interest and movie feature which are taken as the input ofusers'interest CF recommendation model to calculate users'similarity,movies'similarity,on movies.Finally,the system sorts based on user's interests on movies and selects topN movies with the highest interest for recommendation.However,it does not consider the time dynamics of user interest in the above method.Therefore,this thesis proposes another model based on user dynamic interest.On the basis of the assumption that user's behavior reflects user's interest,user's recent rating sequence of movie is input into the Long-Short-Term-Memory(LSTM)network to extract the short-term interest of users,which is then combined with user's long-term interest to obtain user's dynamic interest feature.Then,user's dynamic interest feature and movie features obtained in the previous model are input into the neural network for prediction.Finally,user interest and movie feature are extracted by using the optimal model and taken as the input of CF recommendation model to generate topN movies with the highest interest for recommendation.After comparing with the traditional recommendation model,it is found that the methods proposed in this thesis both have a certain improvement in accuracy,especially the second one.In addition,this thesis designs a movie recommendation system using Django technology where users can log into the system through the web page,browse movie information and the system will recommend movies to users based on user's login information.