项目名称: 基于互联网海量金融情感信息的多方位金融市场智能关联研究及在线决策支持系统
项目编号: No.71271211
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 管理科学
项目作者: 梁循
作者单位: 中国人民大学
项目金额: 50万元
中文摘要: 通过互联网金融信息研究股市波动属于典型的计算机和金融交叉领域。本项目利用互联网采集机,收集积累互联网包括论坛、博客和微博上的海量互联网金融信息数据,构造出含有褒贬的新的计算机-金融信息时间序列W,深层次多方位地展开对W的挖掘:利用支持向量机智能地、自动地找出W与金融市场波动之间的复杂非线性关系,并对W完成自动EGARCH建模;改进我们现有的基于数据椭圆分布的支持向量机,并应用于W建模,来进一步减小回归误差;在W中改进我们现有的对未定义语义模板的时间序列形态挖掘算法;挖掘在Web 2.0下股民提供和传播信息的时间序列动态的因果网络结构关系。在理论上升华为智能计算金融学;在应用上讨论网络海量金融信息在线监管决策支持系统。本申请探索度量和挖掘互联网金融信息的新的基础原理,可以从一个不同的侧面探索股市微观结构,为同行积累互联网金融数据,从网络金融角度上发展金融学体系。
中文关键词: 金融市场;互联网;智能建模;金融风险;情感分析
英文摘要: The research on stock volatility based on internet financial information is a multidisciplinary cutting-edge subject. In this project, the web harvester is built up to collect massive financial information on the web including BBS, blog and mirco-blog, and obtain a new web information sentiment-based financial time series W, which is mined from varous aspects with depth and full spectrum. Utilizing the capability that support vector machines are able to automatically recognize the complex nonlinear relationship, this novel computer-finance time series W is then associated with the stock market volatilities using EGARCH and support vector machines. The W is also modeled with our improved hyperellipsoidal statistical method in a reproducing kernel Hilbert space to further reduce regression errors. Moreover, W is further analyzed with our shape detecting technique permitting unknown shapes. In the Web 2.0 environment, the information provided by the investors is investigated based on the causal time series. The study is then structured into the intelligent computational finance in theory, and is applied onto the decision support systems for online management on massive web financial information. The research studies the fundamental principles from measuring, recognizing and monitoring the web financial information,
英文关键词: financial market;internet;intelligent modeling;financial risk;sentiment analysis