面向金融市场的多源异构数据知识表示与应用研究

This project intends to focus on three types of data related to financial market, including market data, financial news data and social media data. Market data belongs to time series in the structure, which are main data source of traditional financial analysis. Both of the news and social media information are unstructured text data, but the former is long text, whose grammar is relatively standard but content is broad; the latter is short text, whose grammar is loose and semantic expression is not accurate. How to use new methods to find effective information of financial market from the three categories of data, and apply them to the practical problems of the financial markets, is exactly what this project will study. We will introduce new methods to excavate and forecast the historical market data. For financial news, we will identify the topics from the perspective of content and analyze the relationship between the news topic and the market, further perform market forecast based on multi-level news system. Based on the identification of investor sentiment through social media information, we will study the investor sentiment contagion and diffusion mechanism in financial market. These researches will provide new perspectives and tools for financial analysis, and foundation and guidance to market forecast, risk control and trading strategy design in big data environment.

本课题拟围绕与金融市场密切相关的三类数据——市场行情数据、财经新闻数据和社交媒体数据展开研究，其来源不同，结构相异。市场行情数据在结构上主要表现为时间序列数据，是传统金融研究的主要数据来源。新闻报道和社交媒体信息虽同为非结构化的文本数据，但前者主要是长文本，语法结构相对标准，但内容较为宽泛；而后者多为短文本，语法结构松散，语义表达不准确。如何运用新的方法从这三大类多源异构数据中挖掘金融市场所需的有效信息进行知识表示，并结合金融市场的实际问题进行应用，是本课题要深入研究的内容。本课题拟引入新的方法对市场行情数据的历史模式进行挖掘与预测；对于财经新闻，拟从内容角度进行主题分析，并基于多级新闻体系进行市场预测；还将通过社交媒体信息研究金融市场上的投资者情绪传染及扩散机制。本课题将为研究金融市场的具体问题提供新的视角和方法，也为大数据环境下的交易策略构建、市场预测、风险监控提供依据与指导。

本课题围绕与金融市场密切相关的多源异构数据，结合金融数据的特征，引入深度学习、物理、生物等跨学科领域的方法对其进行信息提取和知识表示，研究了其中蕴含的信息对金融市场的影响及机制。课题重点研究了四个方面的问题：（1）基于社交媒体数据研究投资者行为及与金融市场的互动关系；（2）基于新闻文本数据研究股市的媒体效应及财经新闻对股票市场的影响；（3）基于高频和超高频的市场行情数据，开发新的算法进行市场预测并构建交易策略；（4）将跨领域的方法与计量方法相结合，对金融市场的一些重要问题展开应用研究。本课题的研究方法和研究结果为研究金融市场的具体问题提供了新的视角和工具，也为大数据环境下的交易策略构建、市场预测、风险监控提供了依据与指导。