Recommendation Method And Application Based On Embedded Representation Learning

Due to the rapid development of information technology and the gradual formation of world economic integration,the Internet has also been able to achieve rapid development on a global scale.With the number of users worldwide has increased year by year,a large amount of user data are created.According to statistics,from 2018 to 2025,the global data volume will increase rapidly from 33 ZB to 175 ZB.On the one hand,since Industry 3.0,it has affected the daily life of more and more people by providing users with a huge amount of information,but on the other hand,the information age has brought a lot of opportunities and challenges to the industrial and research fields.Faced with a large amount of data,how to effectively and rapidly obtain information for users to meet their demand and effectively alleviate the problem of information overload is a major problem solved by researcher today.In the face of information overload,the earlier solution was search engine technology represented by Google,Baidu,etc.However,due to the versatility of search engines,it is cannot to meet the individual needs of each user.The recommendation system provides a personalized recommendation service for individual users by analyzing the difference of individual information of users,which is a good effective solution for relieving information overload.The tag-based recommendation system utilizes the tag information which is used by user to tag items to recommend items for users,but the recommended methods in the existing tag system only uses the item-tag-user relationship and does not pay attention to the characteristics of users and items.When measuring the similarity between users and items,the existing methods do not combine user similarity and item similarity.In addition,due to the problem of data sparsity,the final result will be affected when recommending the items for users.How to alleviate the problem of data sparsity and improve the accuracy of recommendation results is an important and practical topic in the current recommendation field.The content of this paper for this topic is as follows:1.Aiming at the single similarity problem,an algorithm which obtaining similarity through the tag information of users and items is proposed.The algorithm linearly combines the results which comes from tag information of users and items.The method is based on the user's personalized recommendation,and a collaborative filtering recommendation algorithm based on the similarity of tag information features is constructed.The experimental results on the Last.fm dataset shows that the algorithm can improve accuracy of recommendation results.2.For the problem of data sparsity,a method of calculating similarity by representation learning is proposed.This method embedding the tag to calculate the similarity between users and items,and applies it to the scene of recommending the movie for the user by uses the tag usage record of users and items.The experimental results on the Last.fm and Movielens datasets shows that this method can effectively alleviate the problem of data sparsity,which is helpful for improving the recommendation effect.3.Based on the above algorithm,this paper designs and implements a movie recommendation system.Based on Spring,Spring MVC and My Batis framework,Python is used to calculate recommendation results,and interact with users by Java and Java Script.Then generate a Top-5 movie recommendation list for users based on user viewing records.