Research Of Collaboration Based On Complex Network

Nowadays, problems we are facing can no longer be solved by someone alone. Thus, to conquer greater challenges, collaboration is built up between one and others. Different thinking, resources and skills gather with collaboration. By taking part in collaboration, agents come up to a huge network. In this network, nodes stand for the collaborators, and edges stand for the collaboration among the nodes. This kind of complex network is called collaboration network. Robert K. Merton thinks that nodes in the collaboration network a large connected cluster comprising most of the agents, which is so-called "invisible college".In fact, it is not only building up a team. A team with better performance is needed to complete tasks. At the same time, how to minimize the cost shall be considered, too. Therefore the collaboration is concerned by many researchers.Guimera et al. have provided a groundbreaking research for the collaboration networks. In their model, agents collaborate by organizing teams. After it, many researches of collaboration networks are carried out, such as the scientist collaboration networks, the actor collaboration networks, the author collaboration networks etc. However, most of them only consider the topological structure and the mechanism of making team, with point at the different relationship between agents, but neglect that agents are different. In fact, each agent also has its different specialties, abilities and properties. Agents of different levels may play a different part in his team.To take each agent’s different properties into account, we will investigate empirically and theoretically a new kind of mechanisms by which teams of creative agents is assembled. We analyze the player collaboration network of NBA firstly. With the features of NBA player collaboration network, the new model is given out. By simulation with computer, we investigated the macro evolution of the collaboration network by testing the local team assembly mechanism. Moreover, we compare simulation results of our model with the real data to test the rationality.The main contributions of this thesis could be summarized as follows:Firstly, we studied the substance, models and characterizations of complex networks, and collaboration networks. Secondly, we analyzed data from the NBA. We considered all3200teamxyears in the period of60years, and found the ts%normal distribution, the ts%and the degrees assortative and the high-value club phenomenon。With the findings we give out a model of team assemble mechanism based on Guimera’s model, not only considering the diversity of incumbents and newcomers, but also the different capacity of each agent.Finally, we compare our results with the real data and make a simulation of10000agents to study how different parameters control the production of invisible college.