Research On Modeling And Application Of Temporal Social Attributes In Mobile Social Networks

With the popularization of smart mobile devices and the rapid growth of mobile users,Mobile Social Networks have attracted wide attentions in many research fields.Mobile social networks are mainly composed of a large number of users carrying mobile devices,so the mobile characteristics of nodes in the network have a close relationship with human behavior patterns.People's daily activities usually have a high regularity,by mining the social relationship among people,we can better find the evolution of the community law.At present,there are a lot of studies on Mobile Social Networks,which mainly use the static aggregation method to model the network.These models ignore the time-varying dynamic nature of Mobile Social Networks,and may lead to the loss of some valuable information.Therefore,this dissertation models the network based on the temporal evolution model,and studies the changes of the structure from the dynamic perspective,so as to excavate its change rules and apply them.Centrality and community are two important metrics of Mobile Social Networks.Therefore,this dissertation focuses on investigating centrality metric and community detection in Mobile Social Networks from the temporal perspective.The main research work includes:The centrality of nodes in Mobile Social Networks is investigated from the temporal perspective,and a time-ordered aggregation model is proposed.The dynamic network is simplified into a series of time-ordered networks,and then a new centrality metric: cumulative neighbor relationship(CNR)is proposed.Based on the time-ordered aggregation model,the average time-ordered aggregation method,the linear time-ordered aggregation method and the exponential time-ordered aggregation method are proposed.By combining CNR and the time-ordered aggregation model,we propose a new centrality metric: Temporal Cumulative Neighbor Relationship(TCNR).In order to evaluate the performance of TCNR,extensive experimental simulations are carried out.The results show that TCNR based on the exponential time-ordered aggregation method can measure the importance of node more accurately not only in the MIT Reality trace,but also in the Infocom 06 trace.Therefore,this dissertation uses the exponential time-ordered aggregation method to measure TCNR of nodes in a period.Furthermore,the results show that TCNR is better than other existing centrality metrics.This dissertation also studies the problem of community detection in Mobile Social Networks from the temporal perspective.By combining the time-ordered aggregation model with community detection,a temporal community detection algorithm based on hierarchical clustering is proposed.Extensive experimental simulations are carried out by using the MIT Reality and Infocom 06 traces,and the similarity of communities under each time window is compared.The evolution law of communities is found.The results show that with time goes by,the similarity between communities gradually decreases,but increases again in weeks,which conforms to the social law of human beings.Finally,this dissertation designs an efficient routing protocol Routing Protocol based on Temporal Community and Degree Centrality(RPTCDC)for MSNs by combining the temporal community and degree centrality of nodes based on time-ordered aggregation model.The RPTCDC algorithm is compared with the Epidemic algorithm and the Prophet algorithm on the ONE simulation platform.The simulation results show that the routing protocol proposed in this dissertation perform best in terms of transmission rate,network overhead,and average delay.