基于知识关联的金融大数据价值分析、发现及协同创造机制

Financial big data have great value which needs to be efficiently discovered, utilized and produced. New theoretical framework and value production mechanism are required to promote the efficiency of value production and enlarge the application area. In this project, we would design a new approach to organize and represent the knowledge information with its association structures from various levels and perspectives. Based on this approach we would build a knowledge large graph on financial data, and the studies and the solutions of classical financial problems are combined to the construction and evolution mechanisms of the financial knowledge large graph through the knowledge associations. The knowledge large graph would provide abundant semantic information from various perspectives for the value analysis and discovery of financial big data to promote its efficiency and quality. On the other hand, with the financial knowledge large graph, the financial system can be treated as a self-organization system composed of humans and financial knowledge. With such a system we can explore the collaboration rules of financial activities from the agents’ behaviors and establish the new collaboration work model and the framework for financial value production. A new full-featured knowledge computation framework and the backbone services would be built based on this model. Furthermore, a collaborative innovation platform for financial big data from the security area would be built to satisfy the business and research requirements from various kinds of users by providing various application services, including business portraits, intelligent investments, risk management and control, business information recommendation and business crowdsourcing.

金融大数据蕴含着巨大的商业价值，高效地发现利用这些价值并提升价值创造效率、扩大其应用范围，需要突破传统方法、引入新的价值分析发现的理论框架并建立相应的价值创造平台。本课题拟基于大数据中知识的多层次、多角度关联结构设计知识组织和表示方法，以此为基础构建金融大数据的知识大图，通过知识关联将金融领域的典型研究问题与方法和知识大图的构建与演化机制紧密结合，为金融大数据的价值分析、发现提供更加丰富的语义信息以提高价值创造的效率与质量。同时利用金融知识大图将金融系统作为一个人与金融知识的自组织系统，从主体行为角度探讨金融活动的协同规律并建立新的金融价值创造的协同工作模式和基本框架。基于该理论框架设计实现功能完整的知识计算支撑服务，并基于证券行业提供的各类金融大数据建设协同创新平台，提供包括企业画像、智能投资、风险管控、信息推荐以及业务众包等在内的多种应用服务，满足各类用户的金融业务与研究需求。

金融大数据蕴含着巨大的商业价值，高效地发现利用这些价值并提升价值创造效率、扩大其应用范围，需要突破传统方法、引入新的价值分析发现的理论框架并建立相应的价值创造平台。本项目对金融大数据中的知识关联特点进行深入分析，并以企业基本面和股权结构信息为核心，获取、整合典型金融数据源中蕴含的丰富知识，构建了具有原创性的中国金融知识大图。基于金融知识大图，本项目对全国银行业股权网络结构及与之相关的潜在系统性风险进行了整体分析，完成了银行股权网络的基本特征分析、银行股权穿透式查询及分类持股比例分析、银行隐含控股股东及实际控制人揭示、银行股权网络拓扑结构及隐藏资本系发现，以及银行股权网络关联及环路风险剖析等多个核心问题的研究。同时，基于金融知识大图，本项目设计了基于交易成本的量化投资组合模型、基于基本面和技术面的量化投资策略模型与算法、以及基于机器学习的量化投资方法。在此基础上，本项目开发了基于金融知识大图和金融领域多个数据源数据交换API、金融股权网络的完全穿透构建与拓扑结构性分析技术等核心计算服务，以及基本面信息的查询、统计、分析等基础设施服务，形成了功能完整的金融股权知识大图计算框架。利用该计算框架，本项目研制开发了数据发布与服务、知识关联查询、风险识别、资本系分析和金融本体查询等功能模块，同时利用数据融合、信息抽取、文本分析挖掘等技术，构建了具有行业领先地位的金融大数据价值分析与协同创新应用平台。