Research On News Recommendation System Based On User Logs

Since the 1990s,as information technology has become more and more advanced,various resources have been generated on the World Wide Web.In the face of a large number of redundant network data,it is difficult for users to discover the data resources that users really need.To solve this problem,some search engines^[1]and portals filter the information to some extent.However,the traditional search engine is better at directed exploration.It seems that the information that users can not describe clearly is not enough.In addition,for different users,the traditional information filtering technology provides the same resource list,and can not recommend personalized resources to users according to the potential needs of the users according to their potential interests.With the Recommender system^[2]put forward,this problem has been solved.Recommended system as a link to effectively connect users with data information.The recommendation system can not only replace the user to filter unnecessary resources,but also provide the target user with information that the user has not paid attention to but may like,which is a process known from the user to the user is unknown.The proposal of the recommendation system has greatly alleviated the problem of"information overload"^[3]and"information fog"in modern society.This article focuses on the research of news recommendation system,analyzes the user's behavior through the user's browsing news log,and mines user interest information,so as to make the personalized recommendation results for the user.The main research contents are as follows:First,we study and improve the topic distillation method.This method is improved on the basis of traditional LDA algorithm.Since the word frequency distribution of the news of the document conforms to the power-law distribution,through the study of the characteristics of words weighted method,combination of Gaussian distribution and the TF-IDF method for feature weighting,so as to improve the weights of the low-frequency words,reduce the weight of high frequency words,achieving the purpose of improving the accuracy of topic distillation method.Second,we study the news recommendation algorithm based on user log.The algorithm combines the collaborative filtering recommendation algorithm and the association rule-based recommendation algorithm.The collaborative filtering algorithm includes the user-based collaborative filtering algorithm and the item-based collaborative filtering algorithm.In addition,according to the characteristics of the news document,the similarity calculation method is improved.This algorithm reduces the time cost of calculation by clustering method,and combines these two algorithms to make news recommendations based on user logs.Thirdly,design and implement the functions required by the news recommendation system based on user logs.The system is calculated based on the recommendation algorithm discussed in this paper.It completes the different functional parts for the needs of the news recommendation system.The system provides the basic functions of a news recommendation system and presents the recommended system results through the web interface.