Analysis Of Human Mobility Features In Dynamic Social Networks

Dynamic Social Networks DSNs have emerged as a novel communication paradigm in pervasive environments in which the time-evolving social characteristics and behavior of mobile nodes(i.e.,users and their devices)are employed to design effective and efficient networking protocols.In DSNs,the nodes have non-random mobility and contact patterns,which are strongly dependent to their personal and social behavior as well as environmental parameters.Thus,social network analysis techniques such as centrality and community can be applied to explore their movement patterns and extract meaningful patterns and relationships between them.Identifying node centrality in networks with greater accuracy may have unprecedented benefits,as it may allow one to design better routing strategies or to select key players in a more appropriate way.It is also useful in the context of criminal and terrorist network.Various metrics and algorithms have been proposed to quantify node centrality in DSNs.However,most of these methods have been formulated to study static networks in which the nodes and links do not change dynamically.However,DSNs evolves dynamically that pose several challenges.Such changes in dynamic networks may occur due to link removal or link appearance between nodes due to the mobility of the nodes themselves.Furthermore,identification of the most effective nodes,called Key Players Problem(KPP),is one of the lines of research in social network studies,in which one attempts to identify a set of nodes with the most positive or negative influences in the networks.By positive effects one usually addresses the influence a node has in the spread of information through the network.As for negative effects,one would examine the criticality of the node in rendering the whole network damaged or fractured.We state three major research problems in DSNs and present our proposed solutions.First,we address detection of overlapping communities in DSNs.In particular,we present a set of algorithms according to various factors concerning this issue.First,we exploit a system which gives weight to the nodes we create a slight difference in the nodes connecting friends who are very close and friends who are only acquaintances.Then,we look at groups which are shared in clusters and take a look at their roots in individual fashion.These Communities of connection determine whether friends are close in real life or whether the pattern of connection is caused by cluster knots.Then this information also shows us to what extent each person is likely to accept influence from a connection or friend according to their level of closeness.This trait of openness to information dependant on the source is stored as a tag in our designated memory space.Our conducted experiments show that in the three factors F-Score,Nicosia,NMI there has been presented a certain algorithm,which has become more useful and effective for this process.Second,we design a novel method to predict the future centrality of mobile users in DSNs.The concept of centrality plays an important role in analyzing the social behaviors and employing desirable actions such as data routing and dissemination in opportunistic mobile networks.In other words,predicting the future centrality of the mobile users help we design efficient routing protocols in energy-constrained DSNs.We formulate our approach on the basis of calculating correlations between current and past centrality measures of the users.The states of centralities in the network are represented using a vector.In order to check the behavior of the network nodes,we apply correlation and distance vector metrics in the experiments.Furthermore,we analyze the performance of our approach using the variation of correlation and distance MSE on three real-world human mobility datasets(Intel,Cambridge,and Infocom).Our experimental results show that our scheme performs better in comparison to some existing benchmark methods in terms of data dissemination success ratio and average data delivery delay.Finally,we address key player identification problem in DSNs and develop an efficient method to identify the most influential individuals that help us better understand and control the behavior of network nodes.Since most of existing real-world social networks have social and time-varying features,it is non-trivial to study the importance of social attributes of individuals on the key player identification problem.Therefore,in this thesis we investigate the inter-relationship between some important social features,such as flexibility and sociability of the individuals regarding their roles in DSNs.In addition,we use particular applications for information diffusion through the network using our proposed scheme.To this end,we design a dynamic network model to provide definitions for the individuals' social attributes and explore opinion transmission among them.In the experiments,we evaluate the effectiveness of various sets of individuals in information diffusion using our proposed method.By selecting sets of players from different regions,we analyze the relations among these social features and the importance of the individual's roles.Extensive experiments are carried out to evaluate and compare our approach against some existing schemes in terms of execution time,delivery ratio,and overhead ratio.