Heterogeneous Agent-based Modeling And Computational Experiments Of Infectious Disease Transmission

The spread of infectious diseases threaten the health and life of human being. The large scale outbreak of epidemics can pose panic and disturbance in human societies, and bring about challenges for public health and social management. In recent years, the emergence and diffusion of novel pathogens are unconventional. It stems from the difficulty in detecting and predicting the initial intrusion of novel pathogens. The traditional methods of studying epidemic spreading, such as epidemic field reconnaissance and statistical analysis, can not support the evaluation of disease control policies and social management. However, computational modeling and simulation becomes a promising approach that could meet the higher requirement of social management and disease control. The main idea of this approach is to firstly build artificial social scenes to simulate the emergence, development and evolution of disease spreading events, and then execute computational experiments to support decision making and evaluation on preparative policies.The dissertation aims to build the models of artificial societies, execute the computational experiments of epidemic spreading, and analyze disease control policies. After analyzing the limitation of existing epidemic spreading models, we study the heterogeneous and stochastic nature of epidemic spreading and the representation of human behavior related to epidemic diffusion. At the same time, we proposed Heterogeneous Agent-Based Modeling for epidemic spreading(HABM). HABM is a bottom-up modeling method that emphasizes the representation of heterogeneous and stochastic micro process of epidemic spreading, such as epidemic progress based on heterogeneous and stochastic timescale, the heterogeneous evolving infectivity of infectious individuals, heterogeneous links between individuals in social networks, and heterogeneous and stochastic human contact patterns. Our contributions are represented as follows.(1) We propose the method of modeling agents’ temporal-spatial contact behavior based on weighted two mode networks. We integrate weighted spatial mobility networks and weighted contact networks to build weighted tow mode networks that are used to build agents’ spatial mobility models and contact behavior models, where Markov Chain and Task Schedule are employed. At the same time, we design an algorithm of space moving path planning and the mechanism of selecting contact objects for agents. Then we build the models of agents’ behavior changes based on weighted networks at the perspective of disease information, including agents’ behavior changes based on adaptive weights and the adaptive mechanism of selecting contact objects.(2) We propose the method of building individual-based heterogeneous models based on statistical distributions. To resolve the problem that there is no real data to support heterogeneous agent-based modeling for epidemic spreading, we use random variables subject to statistical dsitributions to represent the heterogenesous attributes and states of individuals. This method is a feasible resolution of building the heterogeneous models of epidemic spreading and the large scale agent-based models of simulating epidemic spreading in cities.(3) We propose the framework of representing epidemic spreading models based on matrice and vectors. To resolve the limitation of mathematical models and complex network based system dynamics models of epidemics to represent individual heterogeneity, and the difficulty in using mathematical language to represent agent-based models of epidemics, we use matrice and vectors to represent the attributes and states of entity objects. Then we build experimental statistical variables based on matrice and vectors, and design algorithms to drive the computation of epidemic models. This framework not only realizes mathematical representation of heterogeneous agent-based epidemic models, but also supports theoretical analysis of epidemic spreading patterns.(4) We study epidemic spreading velocity in weighted networks. To provide a real understanding of epidemic spreading velocity in weighted networks, we study epidemic spreading velocity in weighted evolving scale-free networks, weighted Barabási-Albert scale-free networks, weighted Newman-Watts small-world networks, and weighted Erd?s-Rényi random networks. Results indicate that a higher heterogeneity of edge weights leads to a faster spreading velocity in weighted evolving scale-free networks. Epidemic spreading velocity is faster in weighted evolving scale-free networks than in unweighted Barabási-Albert scale-free networks. In addition, when edge weights are correlated with node degree, a higher heterogeneity of edge weights leads to a faster spreading velocity in weighted networks. When edge weights are anti-correlated or uncorrelated with node degree, a higher heterogeneity of edge weights leads to a slower spreading velocity in weighted networks.(5) We study the mechanism of super spreading events aroused by the SARS epidemic in 2003. We explore the mechanism of super spreading events and analyze the characteristics of super spreaders by using node degree, contact patterns, pathogen load and shedding rate, and delayed admission time. Results indicate that individuals who have a large node degree, a long delayed admission time, an active contact pattern, and a higher pathogen load and faster shedding rate are potential super spreaders. If the imported case is not admitted and isolated earlier, and infects many other individuals, super spreading events and large scale epidemic outbreaks may emerge with a high probability.(6) We study the largest collective outbreak of H1N1 influenza at a Chinese university in Hebei province in 2009. We build the models of No.7 dormitory building in the campus, including geographic space, population distribution, social networks, and individual behaviors and interactions. Then we reproduce the outbreak of H1N1 influenza in the building, and evaluate the control policies of H1N1 influenza that are used in the real outbreak of H1N1 influenza.In this dissertation we focus on epidemic modeling and simulation. Our contributions enrich and advance epidemic models, and play an important role in explaining the phenomenon and discovering the patterns of epidemic spreading, and support decision making on disease control. Our works can facilitate the development of epidemic models and simulation systems.