大数据视角下外部信息环境因素对股价崩盘风险影响机理与作用机制研究

With forty years’ opening and reform, China stock market experiences a dramatic growth. However, in China stock market, the phenomenon of stock price steep fall occurs frequently. Multiples of studies and practices show big data analysis techniques can effectively measure investor behavior and discover financial risk. In this view, we employ the classification algorithm, such as random forest, to commit the textual and semantic analysis of investors’ intraday postings in the financial interaction community and measure the dynamic evolution of investor sentiment. Also, we employ the directed acyclic and network topology algorithm to conduct the network analysis of investors’ daily online searches behavior and measure the dynamic evolution of investor attention. Furthermore, we explore the impact mechanism between external information environment factors and stock price crash risk under mid and high frequency level. Finally, we propose the agent-based modelling approach to explore the inside mechanism between external information environment factors and stock price crash risk. The aim of this project is through big data analysis of intraday investor interaction and online searches data, we measure the dynamic evolution of external information environment factors and further explore the impact mechanism to stock price crash risk, under mid and high frequency level, which effectively overcome the timeliness problem caused by the traditional financial data and research scope deficiency resulting from the internal information centered studies.. Our findings would greatly facilitate comprehensive, timely and accurate understandings of stock price crash risk, improve the price discovery efficiency of risky asset, and promote the steady and healthy development of stock market.

改革开放四十年来，我国股票市场蓬勃发展。然而，我国股票市场中股价暴跌现象层出不穷。大量研究与实践表明，大数据分析技术能够有效地测度投资者行为，揭示金融风险。因此，本项目拟采用随机森林等分类算法对于投资者在财经互动社区中日内发帖信息进行文本与语义分析，测度投资者情绪动态演化过程；拟采用有向无环和网络拓扑算法对投资者的互联网搜索行为进行网络化分析，测度投资者关注的动态演化过程；并进一步地在中高频维度下探究以投资者情绪与关注为代表的外部信息环境因素对股价崩盘风险的影响机理。最后，本项目采用计算实验金融的方法探究外部信息环境因素演化对股价崩盘风险的内在作用机制。本项目有效地克服了传统财务数据研究因时效性不足所导致的信息缺失以及主要以公司内部信息研究为中心的研究范围上的不足。研究成果将极大地促进全面、及时且准确地认识股价崩盘风险，提高风险资产的价格发现效率，促进股票市场平稳健康发展。

我国证券市场起步较晚，股价暴涨暴跌的现象屡见不鲜。股价的暴跌往往引发投资者的巨额损失，因而受到监管者和学术界的广泛关注。在信息技术飞速发展的今天，本项目结合中国股票市场散户较多的客观实际，从大数据的角度出发，探究外部信息环境因素对股价崩盘风险的影响机理与作用机制。项目研究内容主要从媒体关注度、投资者乐观情绪以及宏观经济不确定性等当前受到较高关注的宏微观外部信息环境因素展开。考虑到风险资产的联动性，本项目还对研究内容进行了适当扩展。研究结果表明，媒体关注度、投资者乐观情绪以及宏观经济不确定性对于股价崩盘风险、外汇以及加密货币市场具有显著地影响。此外，基于实证研究的结果，本项目构建了基于多资产订单簿的计算实验金融模型，研究结果发现市场制度的变化会带来股票波动率的显著改变。通过上述研究，项目负责人以第一作者身份发表SCI/SSCI检索论文7篇，此外，还获得地厅级科研成果奖2项。研究成果基本达到了项目研究的预期要求，为投资者的风险规避以及监管机构的宏观调控提供了有益的参考。