资本市场企业信息系统人物和企业关系图谱的设计与实现

发布时间：2018-05-27 09:05

本文选题：关系图谱 + 知识图谱　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：在互联网+大数据时代,决策日益基于数据和分析做出,而非经验和直觉。近年来,随着信贷、消费等领域个人“用户画像”的成功应用,如何对资本市场企业和人物对象进行全方位、多角度的模型刻画正在成为金融监管和投融资的一个新热点。本文基于作者在证券交易所的实际开发项目,针对资本市场中证券交易所的监管需求,设计并实现了一个以资本市场人物和企业关系图谱为主要数据模型的信息系统。本文研究的人物和企业关系图谱是知识图谱技术在资本市场这一垂直领域的应用。知识图谱技术自从2012年Google发布以来,其在改进搜索引擎服务质量和效率方面作用明显。本文参考知识图谱技术的通用构建框架,提出了以实体获取和实体关系抽取为主要手段的关系图谱构建方案。在实体获取方面,应用深度学习技术,以长短时记忆学习网络作为语料特征学习模型,以条件随机场为序列标注模型,构建了在文本语料中识别命名实体的方案;在实体关系抽取方面,结合领域知识和业务需求,从公司公告年报等半结构化数据中以规则匹配抽取实体关系。此外,设计并实现了关系图谱在系统中的查询展示功能,提供了良好的可视化及交互性。本文对于资本市场人物和企业关系图谱的设计实现是基于多源异构数据的模型,具有信息价值密度大,抽象层次高以及应用范围广的特点。该业务可以广泛服务于证券交易所的上市公司持续监管、市场监察与执法、以及发行审核与投融资对接等业务,对中国多层次资本市场建设具有重要的支持价值。
[Abstract]:In the age of big data, decisions are increasingly based on data and analysis, rather than experience and intuition. In recent years, with the successful application of personal "user portrait" in the fields of credit, consumption and other fields, how to carry out all-round and multi-angle model portrayal of capital market enterprises and people is becoming a new hot spot in financial supervision and investment and financing. Based on the author's actual development project in the stock exchange, this paper designs and implements an information system based on the capital market figures and enterprise relationship atlas, aiming at the regulatory needs of the stock exchange in the capital market. The relationship map of people and firms is the application of knowledge map technology in the vertical field of capital market. Knowledge map technology has played an important role in improving the service quality and efficiency of search engine since the publication of Google in 2012. Referring to the general construction framework of knowledge atlas technology, this paper proposes a scheme of building relational atlas, which mainly uses entity acquisition and entity relation extraction as the main means. In the aspect of entity acquisition, the method of identifying named entity in text corpus is constructed by using depth learning technology, long and short memory learning network as corpus feature learning model and conditional random field as sequence tagging model. In the aspect of entity relation extraction, combined with domain knowledge and business requirement, entity relationship is extracted from semi-structured data such as annual report of company announcement by rule matching. In addition, the query and display function of relational atlas in the system is designed and realized, which provides good visualization and interactivity. In this paper, the design and implementation of capital market relationship atlas is based on a multi-source heterogeneous data model, with the characteristics of high information value density, high abstract level and wide application range. This business can be widely used in the continuous supervision of listed companies, market supervision and law enforcement, as well as the combination of issuance and audit with investment and financing, which has important support value for the construction of multi-level capital market in China.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.52

【参考文献】