Multi-Agent Models Of Sentiment Contagion In Stock Market Based On Complex Network

Behavioral Finance believe that investor sentment contagion have influence on the stock price behavior and financial market stability.This is a challenge to Traditional Finance and also broke the frame of Traditional Finance assumptions, such as "Efficient Market " and"investors completely rational",which caused a wave of researches on investor sentiment and investor sentiment contagion. Thought contagion of the investors, evolution of financial ideologies and their effects on markets, is one of popular issues in modern finance (including behavioral finance).Computational Experiment Finance is a common tool to research investor sentiment contagion. It is a branch of finance which reveals the the cause of macro characteristic in stock market by studying the actions of microlevel "Agent (investor)" in established market structure.There are a lot of researches on investor sentiment contagion model based on complex network technology, which can depict many phenomenons in stock market; and the multi-agent technology has also been used on the research of market microlevel "Agent (investor)" behavior. Therefore, based on the complex network and multi-agent, this thesis will build a sentiment contagion model in artificially stock market. In this model, multi-agent is used to reflect the interaction between individuals and the complex network is used to represent the social relationship between multi-agents. To compare the effects of different market investor structure, this thesis will ordinally establish three sentiment contagion models based on regular networks, small-world networks and actual network. The models in this paper have three characteristics:1) Adding distribution of investors to investor sentiment contagion mechanism, when establishing network, consider regional differences to make the model closer to reality; 2) Establish three different investor relation networks for comparative analysis; 3) Use call auction mechanism (rather than market maker pricing mechanism), more accord with the operation mechanism of Chinese stock market.Through simulation, main conclusions are:1) Relative to regular network, sentiment contagion in small-world network can more truely reflect market operation. Under the same conditions, leptokurtic distributions of Stock Returnrate in regular network are not obvious, while obvious in the small-world network model. Based on investigation data, leptokurtic distributions of Stock Returnrate are more obvious in the actual network; 2) The interactions of different types of investors are different, information-based manipulation of institution will affect stock market stability; 3) More larger the investors are influenced by market information, more unstable the stock market is.This thesis is committed to the research on microscopic mechanism of investor sentiment contagion in stock market, and offers beneficial theory and model references for supervisor and administrative department to quantitative explore microscopic mechanism of investor sentiment contagion and its influence to stock price. Conclusions and suggestions in this thesis can assist administrative departments to make regulatory decision, so as to promote the healthy and stability of stock market. Meanwhile, It can also help the investors make more rational investment decisions.