Research On Recommender System Based On User Interest

With the popularity of the Internet,the number of articles on the Internet has grown at an alarming rate.Now we have entered an era of information explosion.Faced with massive amounts of data and information,it is difficult for users to filter out the content they are interested in to read.Therefore,reducing the difficulty of accessing and distributing articles,improving article reading efficiency,and providing a user-friendly reading platform have become the most important issues that need to be resolved.Article recommendation is an important way to solve this problem by prompting users to read articles they are interested in.Collaborative filtering algorithms are widely used in recommendation systems due to their simplicity and practicality.However,the algorithm only uses user behavior as a credential of interest,and ignores the content of the article text,so the accuracy of the recommendation is low.In addition,the collaborative filtering algorithm needs to find the user's similar neighbors through the search of all users.When the number of users of the recommendation system increases,the efficiency of the algorithm will be greatly reduced.Therefore,based on the user's collaborative filtering,this paper combines the LDA model and Word2vec to identify the topics that users are interested in,grouping users,and proposes an article recommendation method based on user interests.First of all,LDA defines the global hierarchical relationship from words to topics and from topics to documents.Word2vec is a word embedding model that predicts the target word from the context around the target word.This paper combines the LDA model with the Word2vec model,and uses the distance package to extract semantic spatial features from the document.The distance package not only establishes the relationship between the document and the topic distribution,but also integrates the context relationship of the words,making the document classification based on the topic feature more meaningful.At the same time,this article uses a combination of topic model and word vector model to analyze the user's article content,and describes the user's interest in the form of the distance distribution between the article and the topic.Secondly,in order to reduce the search cost of similarity calculation and improve the efficiency of article recommendation,this paper uses the optimized K-means algorithm to perform clustering operations on users,and then performs clustering search on similar users.Finally,the design and implementation of the article recommendation system is completed,and experiments have verified that this system has a certain improvement in the accuracy of recommended articles.