Research On Financial Market Modeling Based On Complex Networks

With the rapid development of global economy in recent years, the banking has become the leading role during the process of development and has obtained increasing attention of research scholars. The examination of practice has found that the descriptions of price fluctuation by the traditional theory do not conform to that of the real market. Therefore, the research scholars begin their further study. They have found that the traditional financial theory does not explain many phenomena in the real market, so that they have raised doubts about the traditional financial theory of efficient market hypothesis based on the linear equalization.In recent years, the scholars have begun their further intensive study so as to look for the evolution mechanism of the real market and remedy the defect that the traditional financial theory can not explain the various phenomena in real market. As a result, the nonlinear method has emerged as required and has become the best means to study the law of capital market. At the same time, the success that the science of complexity has achieved on the settlement of nonlinear problems has created opportunities for the further development of financial theory.Firstly, this dissertation has applied the science theory of complexity during the research of financial complexity problems. Through the whole research process, this dissertaion has conducted the modeling by the percolation theory based on the scale-free network and the structure of dynamic foreign investment group. In this dissertaion, the financial market has been considered as a nonlinear dynamics system while the self-organization evolution market model for the financial market has been established through the clustering behaviors of various intelligent agents under the restriction of transaction rules. Through the comparison of that model with the real market, this dissertaion has concentrated on the research of the market price fluctuations and the dynamic evolution mechanism of the whole market.Secondly,a mechanism is proposed to extract weighted networks from a financial markert based on the correlations of retruns of stock price.Analysising the model to find the hierarchial organization and disassortative property.Consequently, in view of the above illustrations, this dissertation has conducted the main tasks and innovations from the following aspects:(1) the investment group structure in the model has owned the dynamic and foreign nature. As the connected relation among all traders has formed a scale-free network, each individual in the transaction under the model has had different positions and each investment individual can join or leave the network at any time, so that the topological structure of network may change at any time.(2) this dissertation has put forwarded a new statistical method that can weigh the position of profit distribution and the dynamic change of time. Based on the time series analysis of market index, the price profit obtained has owned the obvious aggregation nature, which shows that the price fluctuation series has owned the successive huge fluctuation and the process relevance. Due to the application of the statistical method, the scholars can better comprehend the dynamics evolution behaviors of market price fluctuation and better understand the relevance of price fluctuation. At the same time, the statistical method can reproduce the profit distribution of real market well.(3) this dissertation has found that the so many statistical characteristic quantities have conformed to those of the real market through the modeling method that simulates the price fluctuation relevance, the aggregation nature of price profit, the spike tail fat characteristics of profit distribution, and other representative statistical features. As a result, the above statistical characteristic quantities have shown that the model has been able to reproduce the real market and simulated the behaviors of market price fluctuations. Moreover, it has found the production mechanism of phenomena that have been existed in the real market but the traditional financial theory cannot explain-the herd effect and the self-organization nonlinear evolution of market network topological structure.(4) correlation-based weigheted financial networks are analyzed to present cumulative distribution of streghth,clusteing,connectivity,average nearest-neighbor degree,coreness of the financial network.The results suggest the financial nertwork is hierarchial organization and disasortative mixing.These research results have shown that the financial market model established in this dissertation will be able to conduct the relatively precise simulation of most features of price time series in real financial market, which will help us to understand the internal operation mechanism of financial system and will also provide some basis and reference for the establishment of forecasting model for financial market.