Research And Implementation Of Personalized News Recommendation Algorithms Based On User Behavior Log

With the popularity of wireless Internet and mobile devices,the convenient way to obtain information makes people more dependent on information.News acquisition has changed from active acquisition to news recommendation.The original news recommendation is to help users find out what they are interested in and recommend it to them,but now it is to actively explore users' potential interests and recommend them.With the rapid increase of information,the problems in recommendation are becoming more and more prominent,such as slow update speed of user model,excessive recommendation of repeated content,too narrow news coverage,cold start,etc.The change of life style makes the users obtain news not purposefully but randomly,which is contrary to the traditional news recommendation method.The traditional model establishment does not well realize the non-purpose process of users browsing news,so this paper introduces hidden Markov model on the basis of the traditional collaborative filtering recommendation algorithm,simulates users' non-purpose browsing behavior through the randomness of Markov process,and studies the application of Markov model in news recommendation algorithm.The main research content is divided into three parts.1)Research the advantages and disadvantages of current mainstream recommendation algorithms and Markov model,apply Markov model to news recommendation and implement a collaborative filtering algorithm based on hidden Markov model.The experimental comparison proves that the algorithm is better than the traditional recommendation algorithm.In order to further improve the efficiency of the algorithm,the algorithm has been optimized and improved.The improved algorithm has improved the accuracy and recall rate,and its efficiency is significantly higher than the previous algorithm.2)On the basis of hidden Markov model,a three-dimensional model is proposed and designed.The hidden Markov model is taken as the main dimension,and the three-dimensional model is composed of user features and environmental features.The application of the three-dimensional model in recommendation algorithm is studied.By mixing the content-based recommendation algorithm and Top N algorithm,a three-dimensional hybrid recommendation algorithm based on hidden Markov model is realized.Experiments show that hybrid recommendation of threedimensional models has higher accuracy than single recommendation algorithm in recommendation effect,and has obvious effect in solving cold start.Further optimization and improvement of recommendation algorithm through clustering algorithm have better effect than before.3)Build a small cluster platform and a small Spark ecological environment,and design a three-dimensional hybrid recommendation algorithm based on hidden Markov model as the main algorithm of the recommendation system.Through the design of the recommendation system,personalized recommendation based on the user behavior log is realized,and the recommendation effect is displayed through the Web page.