Design And Implementation Of Personalized News Recommendation System

With the rapid development of the Internet,people are facing the problem of information overload when they get huge amounts of information.Recommendation system emerges as the times require in such an environment,which can recommend items of interest to people.As a part of the recommendation field,news recommendation has become a research hotspot.A good personalized news recommendation system can attract a large number of users for news websites and bring large user traffic,which can be realized through these user traffic.The process of personalized news recommendation includes two parts:candidate generation and candidate ranking.The effect of candidate generation and candidate ranking stage will greatly affect users interest and clicks in personalized news recommendation,which will have a great impact on the traffic and revenue of news websites.Based on the above analysis,this thesis designs and implements a personalized news recommendation system,and studies the candidate generation model and candidate ranking model.Specifically,the main work of this paper is as follows:(1)The requirements of personalized news recommendation system are analyzed in depth,and then the overall architecture of the system is designed based on the needs analysis and related technologies.This paper divides personalized news recommendation system into six modules:Web display module,business logic module,candidate generation module,candidate sorting module,log collection and processing module and data processing module.Then based on this design scheme,the realization of each module is elaborated in detail.Finally,the function and performance of the system are tested to verify the effectiveness and availability of personalized news recommendation system.(2)For the news candidate generation module and candidate ranking module,the deep candidate generation model based on text information and the improved neural factorization ranking model are proposed respectively.The deep candidate generation model based on text information uses long-term and short-term memory network(LSTM)to mine user log information and convolutional neural network(CNN)to mine text information.Compared with other candidate generation strategies,it can better mine important information of news recommendation scenarios.The improved neural factorization ranking model also incorporates user browsing news information and text information into the model.The comparison experiments show that the proposed depth candidate generation model based on text information and the improved neural factorization ranking model have better recommendation effect in news recommendation scenarios.