Influential Nodes Selection For Mobile Content Distribution

With the rapid development of information technology, computer network has penetrated into all aspects of people’s life and work, and mobile-connected device is gradually becoming the main tool for obtaining information and communication. Under this background, making dynamic network and data transmission by using the communication between mobile devices has become a promising content distribution method. Direct communication between terminals can not only reduce the backbone network traffic load, but also make up for the lack of infrastructure network coverage ability. But at the same time, highly variable network topology also impacts real-time connectivity and network transmission performance. Researchers have carried out a lot of research on DTN, opportunistic network and D2D.Different from most of existing studies that focus on network transmission and routing, based on mobile content distribution in this dynamic network, this thesis discusses how to select influential node set effectively to make possible coverage maximum within a certain duration. This demand has often been seen as influence maximization problem, solving this problem can help realize content distribution in mobile scene, and be used in information propagation in online social network. The main work of this thesis include:(1) Research on node centrality in dynamic network. Study timing relationship between nodes in dynamic network, define temporal degree, betweenness and closeness correspond to traditional centrality by means of constructing reachability graph and calculating foremost journey in dynamic network, then make empirical research and analysis on correlation between temporal centrality and static centrality based on openly released third-party dataset and users online data of the campus network.(2) Influential node set selection based on centrality. In Epidemic model, put forward two methods of selecting influential node set by using centrality, the former selects influential node directly based on sorting of node centrality in reachability graph; the latter makes use of submodular characteristics of the covering problem, chooses one node which gains the greatest contribution as influential node orderly based on greedy algorithm. This paper studied performance of this node selection algorithm through experiment.(3) Influential node set selection based on probability propagation model. In the more realistic model of Independent Cascade model, this thesis studies influential node set selection method based on calculation of influence path, this algorithm is called path probability-based greedy algorithm. We put forward method of estimating influence propagation probability between any pair of nodes more precisely on the basis of calculating non-overlapping path, and then design a greedy algorithm to select influential nodes. This research compares the information coverage result of this method with other methods.(4) The application of node set selection algorithm in actual scenarios. Different from the general research which neglects network dynamic evolution, this thesis pays more attention to content coverage effect within some time, which is more consistent with the need of practical application. Therefore, the proposed algorithms can be used in mobile opportunistic network and online social network respectively to show the effect of the algorithms, based on real campus network user data and Sina micro-blog trace.