Research And Implement Of Recommender System Based On Social Network Potential Hotspot

With the rapid development of mobile Internet,social networks have been deep into people's daily life.Internet users can express their own views on the surrounding life,hot spots and son on through the social network whenever and wherever.The social networks provide massive information to users,while they also brought the information overload and other issues.The users is difficult to find quickly and accurately information that they are interested in.The search engine is passive and non-intelligent.And it is increasingly difficult to meet the needs of the people.So the recommender system came into being.The amount of data processed is often huge in the recommender system.In order to respond quickly to user needs,this paper provides running environments for high reliability and high performance.It includes distributed storage,parallel computing and fulltext retrieval through the integration of large data technologies such as HDFS,Spark and SolrCloud.And this can improve the operational efficiency and stability of the recommender systemThe recommender system based on social networking site is mainly concerned with the public opinion analysts of the relevant departments and recommend some interested people or content for them.It can help them timely access to potential hotspots and public opinion tendencies.The paper first introduced the relevant background and then in-depth discussion of the large data technology framework,and introduced the common recommended algorithm.Finally,combined with large data processing technology,the paper designed and implemented a social network hybrid recommender system.The main works of this paper are as follows:1.Achieve the long-term and stable information extraction program base on Micro-blog,Twitter and Facebook.It provide data sources for public opinion analysis and recommendation in social networking site.2.Implement a hybrid recommended engine based on social networking site.In order to solve the problems such as cold start and non-popular items,design and achieve the recommendation engines based on user interest words,the recommendation engines based on blog content and the recommendation engines based on social networking relationship.The paper produces the personalized recommended results through the use of weighted parallel.And then filter,sort and explain the results for showing to the users.3.Build a stable large data running platform,provide a good recommended Interactively web UI.Finally,complete the overall system running test.