Design Of User Personalized Recommendation Method Based On Deep Learning

With the development of the Internet,hundreds of millions of new information appear every day.In this era of information overload,it has become a huge difficulty for people to find useful information or to present it to others,and the emergence of recommendation systems has solved this problem.The current recommendation system integrates deep learning technology,integrates massive multi-source data.It makes full use of potential user and item characteristics to improve the performance of the recommendation system and makes the results more relevant to user preferences.In this paper,the corresponding recommended methods are studied.For the recommendation of similar projects,improved collaborative filtering based on text convolution is used.For the recommendation based on users' historical behaviors,the Top-N recommendation method based on convolutional neural network is used.Based on these two methods,this paper takes personalized movie recommendation as the application scene and realizes the personalized movie recommendation platform.The main tasks are as follow:1)A recommendation method based on text convolution to improve collaborative filtering.When calculating project similarity,this method incorporates multi-source auxiliary information such as user characteristics and project characteristics.When calculating the end user's interest in the project,the user's interest in the project obtained by feature fitting is incorporated.Experimental results show that the proposed method is more effective than the traditional project-based collaborative filtering algorithm.2)Top-N recommendation based on convolutional network.Firstly,according to the interaction record between users and movies,LFM is used to extract the characteristics of users and movies.Secondly,two kinds of convolutional neural networks are used to extract the features of the film to obtain the final features.Finally,user and movie features are input into the fully connected network layer to predict the probability value that users are interested in each movie.And the top N items with high probability values are recommended to the user.The experimental model data were adjusted during the experiment,and the experimental results showed that the improved experiment recommendation was more effective.3)A personalized movie recommendation system was built.Using personalized movie recommendation as an application scenario,a user-specific movie recommendation system has been systematically built based on the improved method in this paper.The first step is to analyze the overall system requirements,then design the overall system architecture and functional modules based on the requirements analysis,and finally use the Python language and Django framework to implement a personalized movie recommendation system.Through the study of text convolution based improved collaborative filtering for recommendation and the Top-N recommendation method based on convolutional networks.Taking movie recommendation as the application scene,the two methods are applied to real life to provide technical support for movie recommendation.