Research On Structural Analysis And Dynamical Processes On Complex Networks

Motivated by the availability of massive data and significant improvement of the computational power, scientists are able to study and begin to understand the structural patterns of complex systems. Specifically, more and more researchers have been keen on exploring the relationships between topological properties and the systematic behavior. Inspired by the previous studies in this field, this dissertation is dedicated to understand the subject from the perspective of topology and trying to figure out the possible influence brought by the structural patterns to the dynamical processes. The main content of the dissertation is summarized as following.(1) Studies of real world network systems are provided to demonstrate the power of network modeling and structural analysis. The first case is from the industrial system. A task-resource network is modeled based on processes data from a refinery. Based on the knowledge of production process principles, we analyze the structure properties of the system and study the structure-activity relationships. The results prove that the structural analysis could provide insightful information for process system analysis.(2) The second case is from the economy system. A national wide trading network is built based on data from an online shopping website. Using the globe connectivity patterns, we perform the backbone analysis and the dollar experiment to fully understand the trading relationships and the different roles of cities in the online business. Interesting results provide new perspective of data mining and prove the power of network analysis.(3) To understand the relationships between topologies properties and systematic behavior, we first discuss the possible influence brought by static structural pattern to the dynamical processes happening on networks. By applying the logistic model on the networks with different clustering structure, we could simulate the logistic processes and evaluate its performance. Simulation results demonstrate that the logistic dynamics of supply networks are highly related to the clustering structure, which could help to provide more efficient logistic performance.(4) Further studies focus on the networks whose connectivity is changing over time. We discuss the mechanisms of dynamical processes on top of time-varying networks. Activity-driven network model is used to build the time-varying networks and the contagion processes are studied using metapopulation framework. Analytical analysis and simulation results show that the threshold of the contagion could be specifically described by the dynamical properties of the structure. (5) A more sophisticate problem is to study the control strategies of dynamical processes on time-varying network. Using the dynamical structure information, we propose three immunization strategies to control the SIS contagion processes on activity-driven networks. The critical immunization thresholds are derived analytically and the effectiveness of the three control strategies is assessed. The results show that the target strategy is the most effective way to suppress the spreading while the random strategy can only provide slight interference. Ego-centric strategy could overcome the lack of global information and provide decent performance on hampering the contagion processes.In the end, promising studies and applications of structural analysis are discusses in the conclusion.