考虑交易对手网络的多级羊群行为股票多元价格时间序列模型研究

股票价格集体特征事实机理和股票市场的异常波预警一直是金融学领域富有挑战性的课题。目前较多的研究使用随机过程方法来理解单一价格形成机理。对多元股票价格时间序列特征事实微观机理和股票市场异常波动预警尚缺乏深入研究。本项目拟结合多主体模拟和双层网络理论，研究多元股票价格时间序列特征事实的形成机理，并预警股市集体异常波动。首先，考虑投资者在个股、板块和市场整体的多级羊群行为构建多主体模型，再现多元价格时间序列的特征事实。其次，依据已有的交易对手网络实证分析结果，研究交易对手网络对多元价格时间序列特征事实的影响。最后，通过数值模拟和实证分析，从投资者羊群行为和交易对手网络结构两方面预警股市集体异常波动并提出监管建议。预期研究成果能够作为投融资机构决策的实证基础，也为监管机构采取有效的市场风险防控措施提供理论依据。

The stylized facts of the stock prices together with the abnormal fluctuations of the stock market have been some long lasting hotspots and challenging topics which drawn broad attention from the academia community. The traditional research paradigm of the financial economics resides in the widely adopted stochastic process modlling to analyze the stylized facts of single stock price. However, the micro-mechanism of stylized facts for the multivariate stock prices and the formation mechanism of the abnormal fluctuation of the stock market are still lacking. Thus we utilize the multi-agent modelling together with the bi-layer network theory to investigate the price formation mechanism of the multivariate stock prices. Firstly, we incorporate the multilevel herding behavior into the multi-agent modeling in order to better realize the multivariate stock prices and their stylized facts. Secondly, based on the empirical findings about the trading network, we reveal the impact of the trading network on the stylized facts of multivariate stock prices. Lastly, with the help of the previous theoretical model, we thoroughly study the impact of the herding behavior and trading network on the abnormal collective fluctuations of the stock market. The expected achievements should promote the understanding about the price formation for the investment and financing institutions and become the corner stone of their decision making. Meanwhile, the results from this study should also be taken as solid theoretical foundation for the policy maker to perform risk prevention and control.