金融对公知识图谱

来源：某战略合作持牌企业征信企业提供的企业工商数据。
产生方法：1）编写脚本，针对数据中的企业名称、文本地址、人名、数字、时间等字段进行数据清洗；2）针对预处理后的数据，进行实体关系抽取，构建原始图谱；3）对原始图谱进行度统计分析，挖掘热点数据，进一步识别异常数据；4）将异常数据剔除；5）针对剔除后的图谱，进行股权挖掘、同地址等关系挖掘，补充更多隐性关系
主要内容：包含企业间的高管、投资关系及其衍生关系
体量：实体数量超过1亿条，关系数量超过10亿条

Source: Corporate industrial and commercial data provided by a licensed corporate credit reporting enterprise with strategic cooperation partnerships.
Generation Method:
1) Develop automated scripts to perform data cleaning on fields such as enterprise names, textual addresses, personal names, numerical values, and timestamps in the dataset;
2) Conduct entity relation extraction on the preprocessed dataset to construct the initial knowledge graph;
3) Execute degree statistical analysis on the initial knowledge graph, mine hot-spot data, and further identify abnormal data entries;
4) Remove all identified abnormal data entries;
5) Carry out relationship mining tasks including equity connection extraction and co-located enterprise relationship identification on the cleaned knowledge graph to supplement additional implicit relationships.
Core Content: Includes executive connections, investment relationships and their derived inter-enterprise relationships.
Scale: The dataset contains over 100 million entities and more than 1 billion inter-entity relationships.

搜集汇总

背景与挑战

背景概述

该数据集为金融对公知识图谱，基于企业征信数据构建，通过数据清洗和实体关系抽取形成企业间的高管、投资等关系网络。其体量庞大，包含超过1亿个实体和10亿条关系，并经过异常数据剔除和隐性关系挖掘优化。