Research On Technology Of Personalized Tweet Recommendation On Twitter

With the rapid expansion of the Internet,it is increasingly highlighted the necessity of the recommendation system which can filter out useless information to a great extent.At the same time,researchers in various fields,including social and computing science,physics and many interdisciplinary,have deeply studied the recommendation system.In many areas of industry,there are many mature recommendation systems,such as e-commerce Amazon,Taobao,Jingdong,eBay,etc.,movie video Youku potatoes,Youtube,etc.,music Netease music,watercress and other social networks the Facebook,Twitter,Weibo.In these areas,the recommendation system has a very sophisticated architecture.This paper mainly studies the information recommendation technology for Twitter.In recent years,Twitter has become a very important way for information-sharing.Users can diffuse information through tweets.However,with the explosive growth of information in Twitter,social network users must face a flood of information.To dig out useful information in the mass of information is very difficult.So the purpose of this paper is to solve the problems to recommend them interested users.Social network tweets recommendation differs from the electricity supplier in the field of product recommendations.Products differs from tweets.Tweets have a propagation effects.A tweet stream can be configured to propagate the tree in theory.But the purchase of goods is simply bipartite graph.Therefore,we need to consider some tweets recommend specific factors when applying the relevant recommendation algorithm.Based on the relevant study of the information recommended,we propose our main research contents and algorithm models,which contain the following main aspects:(1)Personalized Recommendation Model Based on Retweet NetworkAfter the research and summary of existed information recommendation algorithms,we proposes the algorithm based on traditional collaborative filtering.Based on the retweets relationship matrix of users and tweets,we then calculate the similarity and trust value among users according to the social network tweets forward network.Finally we propose similarity-based recommended model,trust-based recommended model and recommended model based on both similarity and trust.(2)Personalized Recommendation Model Based on the Content LabelTaking into account that personalized recommendation model is mainly based on the relationship Retweet network,this paper presents an improved model based on collaborative filtering.But collaborative filtering algorithm face problem of cold start and data sparseness.To solve this problem,this paper introduce the personalized recommendation model based on content labels.By building labels of interest,emotional tendencies,the temporal behavior of users,we quantified the users.Finally,we calculate the interest of users of the tweets.(3)Design of Recommendation Prototype SystemThrough the above two models,the paper proposed the hybrid model: recommendation model based on retweet-network and content-label.At the same time,Based on the theory of B/S and the framework of J2 EE.finally we design and implement the information recommendation prototype system.In summary,this paper take the information recommendation in social network as background.Using text analysis methods in data mining and the theory of natural language for Twitter social platforms,this paper research personalized recommendation model based on retweeted network,personalized recommendation model based on the content label,and hybrid model algorithm for accurate information recommendation.Then we verify the effectiveness of the proposed algorithms through comparing experiments.Finally we design and implement an information recommendation prototype system.