Key User Selection Oriented To Outbreak Information Detection In Large-Scale Social Networks

In recent years,with the rapid development of the Internet and the widespread adoption of mobile smart terminal equipment,social networks have penetrated into the lives and work of users.Because of the extensive participation of users in social networks and the rapid spread of information,the impact on the country and society become more and more deeply.To select key users with great information dissemination capability efficiently and effectively from large-scale social network and historical user massages is important to control the spread of bad news and monitor public opinions.For the detection of sudden-information in large-scale social networks,we proposed an efficient selection method for the key users in large-scale social networks.In order to deal with today's large-scale social network and its massive user release of the message data efficiently,we use today's popular memory-based Spark distributed computing framework.In this thesis,all data processing and algorithms are based on Spark calculation model.First,we use the social network structure and its users publish the historical data of the message to construct a weighted social network model.Then,on the basis of the weighted social network model,we define the information dissemination ability of the user nodes and give the quantitative measurement method.And the measurement for user's information dissemination capacity is established in terms of PageRank.Further,the selection of the key users is based on the information dissemination capability of the user nodes.Here we propose the d-distance algorithm for node selection to find key users,which makes the overlap of information dissemination ranges of different key users be as less as possible by multiple iterations.Experimental results based on Sina Weibo datasets show that the approach proposed in this thesis is efficient,feasible and scalable.Upon on the theoretical research of this thesis,we implement the prototype system for burst information detection based on Web Service architecture,which further reflects the significance of this thesis.Specifically,the main work of this thesis is summarized as follows:(1)Upon on the Spark framework,the constructive data model of social network is constructed by using the structure information of social network and the historical data of user publishing message.(2)Upon on the constructed social network model,a quantitative measurement method based on Spark's node information propagation capability is given to get the information dissemination ability of each node in the social network.(3)Aiming at the high efficiency and effectiveness of the key users in applications such as burst information detection and public opinion monitoring,a Spark-based d-distance selection algorithm is proposed in this thesis.(4)Based on the experiment of Sina Weibo data sets and Spark distributed cluster,the efficiency,feasibility and scalability of this method are tested.(5)Upon on the research of the key user selection in this thesis,we implement the prototype system for burst information detection with Web Service architecture.