Design And Implementation Of Personalized News Recommendation System Based On Geographic Location

News,as an important means of people getting information,is a hot topic in the field of recommendation system.A good news recommendation system will attract a large number of users,and the reading needs of different users in different places are different.Adding geographic location into the recommendation system can make the news recommendation system more valuable.Based on this,this thesis designs a personalized news recommendation system based on geographical location.The main work of this thesis is as follows:(1)Based on Flink distributed flow computing engine and geographic location context information,this thesis designs and implements a personalized news recommendation system based on geographic location.The system has three functions:Web service,personalized news recommendation based on geographical location and data processing and storage.Data processing and storage functions use redis,elasticsearch,HDFS and other storage media to store the intermediate products and results generated in the recommendation process;personalized news recommendation based on geographic location provides relevant news recommendation service,personalized recommendation service and popular recommendation service based on geographical location for users;Web service is used for use The user provides basic registration and login function,and displays the news recommendation result list after rendering,so that users can read and click to meet the basic needs of users.The distributed storage and distributed computing engine involved in the system have good cluster expansion ability,and the distributed cluster is set as the master and slave mode,which makes the system have good stability.(2)Due to the introduction of the geographical location factors,this system mainly solves two problems,namely cold start of geographical location and popular recall based on geographical location.Location cold start draws on collaborative filtering algorithm.If the location is similar,the reading prefernces of the users in two locations should be similar.Clustering the places where users have been before to get the label list of each cluster.Then the user's current location and the center point of the cluster are calculated to determine which cluster the user is in,and the label in the cluster is matched with news.Location-based Hot Recalls count the top news stories in real time in that geographic location.By testing the two functions,it is proved that both of them can increase the click volume of the recommendation system,and cold start could also attract more new users for the system.