【作者】 邓兰松；

【导师】 郑丕谔；

【作者基本信息】 天津大学 ， 管理科学与工程， 2004， 博士

【摘要】 论文归纳了风险收益与风险管理的基本理论,对证券公司经营业务的风险分析与风险管理进行了研究,并运用极值理论与GARCH模型对金融市场风险价值VaR与CVaR进行了实证计算,最后设计了基于多智能体技术的客户数据挖掘系统作为证券公司客户关系管理的基础。各章节的主要研究内容简述如下。第一章首先介绍了论文的研究背景,然后对国内外相关研究领域的研究现状进行了综述,针对目前研究现状中的主要问题,概括了本文论文的主要研究内容与创新之处。在第二章,根据证券公司在其业务经营过程中面临着各种各样的风险,分别对证券公司承销业务、经纪业务、自营业务与并购业务的风险来源与风险因素进行分析。第三章介绍了风险与收益理论中的有效市场假设、资产组合理论、资本资产定价模型(CAPM),套利定价理论。对市场风险管理的方法进行研究,并对各种有关的风险度量理论的优缺点进行了较为深入的分析,介绍了风险管理的VaR体系。第四章介绍了极值理论的定义与表达式、极值理论的统计推断以及极值理论的超阈值模型,并对风险度量的新方法——CVaR进行了介绍,最后基于极值理论对证券指数的VaR与CVaR进行了实证研究。在第五章,将极值模型推广为平稳收益率序列,引入极值指标消除极值数据间的局部相关性,并应用改进的极值模型对证券指数进行了实证计算。第六章运用GARCH模型分析收益率序列的波动聚集现象,对其中的随机项分别采用正态分布、t分布和广义Pareto分布进行拟合,通过对证券指数的VaR实证计算发现GARCH-GPD模型在VaR计算研究中能够得到更好的结果。在第七章针对现有数据挖掘技术在证券公司客户关系管理应用中存在的困难,设计了基于多智能体技术的客户数据挖掘系统作为证券公司客户关系管理的基础,并对该系统的结构和运行进行了分析。论文最后是全文的总结并提出了进一步研究兴趣。更多还原

【Abstract】 This dissertation introduces the basic theory of risk income and risk management,and studys the risk analysis and risk management of the business of the securitiescompany. Extreme value theory and GARCH model are applied to the demonstrationof VaR and CVaR of financial market risk. At last a CRM system is designed basedon the data mining technology of multi-agent. The main content is as follows:1. Introduction: the research background is presented, survey of recentdevelopment is made, and the main content of the dissertation is as follows.2. Risk analysis and management of securities company: the risk source andfactors of the business of consignment business, brokerage business, self-employmentbusiness and assets management business are in detail analyzed.3. Risk income and risk management theory: the modern risk managementtheory and models, specially the VaR-at-Risk (VaR) syetem are introduced.4. Study of VaR and CVaR based on extreme value theory: the extreme valuetheory (EVT) and the new method of risk measurement CVaR, and demonstrationwith the securities index are introduced.5. Study of VaR estimation of stationary return time series: the EVT model isextended to stationary time series, and extreme index is employed to eliminate thelocal correlation of the extreme value.6. Study of VaR estimation based on GARCH model and EVT: GARCH modelis used to analyze the volatility clustering phenomenon in the return ratio series. Asfor the random component in the model, normal distribution, t-distribution andgeneralized Pareto distribution assumptions are applied, respectively. Best result isobtained based on the generalized Pareto distribution assumption for the randomcomponent in the GARCH model.7. Customer relation management of securities company based on data miningtechnology: the CRM system based on data mining technology of multi-agent isdesigned, and the structure and operation of the system is detailed. Finally, the dissertation is fully summarized and some futher intersets arepresented.更多还原