Native Multi-agent Simulation Of The Evolution Of Industrial Clusters

From 70s to 80s in the 20th century, under the depression backgroud of World economy, outstanding and excellent performance in the economy of High-tech Enterprise cluster in the Silycon Vally and traditional industrial cluster in the Third-italy has attracted the attetion of scholars and policy-makers. After that, there is a following new wave of research and study on the phenomenon of spatial economy like industrial cluster, while some countris' local governments has turned from industry policy to cluster policy, and consider "Industrial Cluster" as a tool to boost the development of local economy.In the theoretic field, originally, Many study focus on the analysis of static efficiency of certain industrial cluster from their individually theoretic viewpoint. However, as more knowledge gained from industrial cluster, people begin to learn that: for one thing, some question could be explained more effectively only we put it on the dynamic background of cluster development;For another thing, explanations from one theoretic viewpoint don't make much sense on certain questions in the industrial cluster. In the practice field, orginally, many local governments has consider cluster as a cornucopia for the development of local economy. But many case study analysis of industrial cluster demonstrate that cluster still has a risk of declining, some clusters may fall down after a period of development. Therefore, it seems people have to conduct more research on the the mechanism of industrial cluster's evolution.However, the evolution of industrial cluster has been influenced by many different factors. What's worse, people are met with some theoretic and methodological difficulties when we want to study the evolution of industrial cluster, such as how to consider the influence of multi-factors at the same time, how to extend the validity of conclusion gained from our analysis of case study of certain cluster, how to obtain more reliable longitudinal data for our analysis. Seen from above, in order to study the evolution of industrial clsuter better, this paper tends to use new theory and new approach, namely complex adaptive theory and multi-agent simulation approach in our analysis to explore the mechanism of the evolution of industrial cluster.We first make a discussion on the defination of industrial cluster, by innovating a defination that reflect the self-organized trait and spontaneous trait, we define the research object of this paper as "Generative" industrial cluster. Next, we summarize the factors that influence the evolution of industrial cluster from economic view, social view and technical view. In order to review the literature more logically, we explore the principle of "From pattern to evolution". We find that the present studies on the cluster's evolution basically have 2 different patterns, one is based on the time dimension while the other is based on system dimension. Through the literature review, we find some drawbacks of present studies and obtain some hint as well, namly the study of micro-behavior of firm agent in the industrial cluster could be a new join-point for the analysis of cluster's evolution. After that, this paper tries to explore the idea of "Adaptiveness create complexity" in the analysis of cluster'sevolution.we construct a dynamic analysis framework of cluster evolution based on the behavior of micro-agent, and this framework has three parts: external Environment, interaction and linkage between agents, serf-behavior of agents. In order to describle the evolution path of Industrial cluster, we use the concept of "adaptive cycle" by Holling(2001) for reference. Then, we try to explore multi-agent simulation approach to specify the possible path of cluster evolution based on the previously presented dynamic analysis framwork. When we put all the factors together into the synthesized model, the model seems to evolve in a path of adaptive cycle.New ideas of this thesis are mainly as follows: In the aspect of new theoretic ideas, this thesis study the evolution of industrial cluster based on the different behaviors of micro-firm agent in the light of the idea of "Adpativenss creats complexity" in complex adaptive system theory. In the aspect of new methodological ideas, this thesis construct a multi-agent simulation model of industrial cluster evolution based on the KISS principle. What's more, we innovate some new approaches during the model construction. For example, by combining the single-layer perceiption with enforced learning algorithem, agents in our model can connect their learning behavior based on experience with information decisioin process well. Besides In order to explore more thing from the model, we put forward a procedure of constructing model and make analysis, namely "De-synthesis, synthesis, Re-de-synthesis."Multi-agent simulation approach is a new tool for the study of industrial cluster. Nowadays, Those study which explore MAS approach in the industrial cluster literature is increasing aborad, while related study in china has just began. Therefore, this thesis can be regarded as something reference for those researchers who want to explore MAS approach in their study. With the limitation of time, energy and knowledge preparation, there is still some defects in this research, the future study can be targeted at the aspects such as, enrich the behavior of agent, optimize some details of the model, and introduce new elements into the model. What is more, because of the modularization of our MAS model, researchers can study other questions through the revision of original code, such as, the mechanism of innovation diffusion in the cluster, or routine-searching mechanism which is a main topic in the evolutionary econimics. These questions are not concerned in this thesis, but our model is quite flexbile, we hope it will be helpful for other researchs.