cool hit counter Thirty-seventh Data Analytics Mining Salon: Credit Risk Management Based on Knowledge Graph and Graph Mining_Intefrankly

Thirty-seventh Data Analytics Mining Salon: Credit Risk Management Based on Knowledge Graph and Graph Mining

Data Analysis Mining Salon

Time: 20 April, 15:00

Venue: Conference Room 912, Jintang International Finance Building

access: Head Office Management Information Department、 data center( Beijing, capital of People's Republic of China)、 Xiamen Branch

On April 20, 2018, the thirty-seventh Data Analytics Mining Salon was held as scheduled, and this salon invited Mr. Chen Meng from the Fifth Department of the Software Development Center of Agricultural Bank to share the knowledge of credit risk management based on knowledge graph and graph mining. The main venue of this salon was located in the Jintang International Finance Building, and the Management Information Department of the head office, Shanghai Data Center and Xiamen Branch participated in the learning and exchange through video access.

Project background and objectives

Ms. Chen Meng first explained the background and objectives of the project。 Central requirements、 Macro situation and banks' intrinsic motivation together point to innovative credit risk management models、 Meeting the risk management of the clientele、 The imminent need to support systemic financial risk monitoring and disposal, Need to work on addressing the willingness of client groups to prevent risk、 capacity、 limitation period、 Effectiveness issues, Do it proactively、 Scientific prevention of systemic risk, Achieving early detection、 early identification、 early warning、 Objective of early disposal。

Technical Architecture and Functionality

later, Mr. Chen Meng explained the technical architecture and functions of the project in detail, From the data、 information、 knowledge-related、 The overall process of the project is explained from an intelligent perspective。

Key Technologies and Innovations

The project has four main key technological innovations.

1.Create a data view (DV) and lay the foundation of information

By building a credit customer identification profile database, Increase the level of integration of data available within and outside the bank, This leads to a more comprehensive identification of inter-customer relationships, Build a full web view of your customer relationships, Further address the information asymmetry of banks to their customers。

2. Building a knowledge graph(KG), Building a knowledge base

Based on the customer relationship view, the knowledge graph of customer risk propagation is formed by considering the customer's own attributes and the association relationship between customers, and identifying the customer groups with strong association of risk transmission through the customer group segmentation algorithm.

3.Achieving knowledge discovery (KD) and laying the foundation for wisdom

Building CGEE, a customer segment risk insight engine based on credit risk relationship knowledge graph.

Core Customer Identification: Core customers have a key role in the transmission of risk to the customer base, is the intersection of potential credit risk, Targeting the core users of the relationship network, Raising the level of the warning, Can effectively prevent group risk。 Based on linkage and risk transfer characteristics, A two-factor aggregation algorithm based on the number of relations and the conduction coefficient is used。

Risk transmission measurement: When a customer in the customer base has a risk event, Timely identification of the most likely risk transmission pathways, Instruct account managers to prioritize、 intervene early、 ordered prevention, Both to gain golden time for risk management, It also cuts off the path of risk contagion。 Based on clientele subnetwork and transmission coefficient, Constructing a clientele risk transmission probability matrix, Applying the state migration algorithm, Measure the compound probability that each customer in the subnet is affected by other customers, be extremely distant from near or distant、 Customers with the highest probability of forming a risk transmission path per layer of depth。

Segment Risk Assessment: Consider the clientele subnetwork as a community of risk, Through information integration、 Data modeling、 Overall evaluation yields an overall risk rating for the clientele。

Customer relationship exploration: for customer credit relationship exploration, we have independently developed a high-performance graph calculation and display framework to support users' graphical, differentiated, customized and simple exploration of massive customer relationship information and large-scale complex relationship networks, supporting real-time grouping queries of 10 billion data, single Server 150TPS, response time 1-3s, thus covering the needs of exploring customer relationships and associated risks for 300,000 new businesses per month.

4.Comprehensive application of knowledge (KA) to enhance the efficiency of risk management

Credit Risk Knowledge Graph and Segment Risk Insight Engine Fully Applied to Credit Management Process, Improving the efficiency of risk management, Helping to prevent systemic risk。

The project was awarded several patents and published several articles。

Application effect

Finally, Ms. Chen Meng introduced the effect of the application of the project. The project has achieved the expected goal of early detection, early identification, early warning and early disposal, automatically identifying 5.21 million relationships, more than 9,800 customer groups and 730,000 members; calculating an annual average of nearly 2,000 risk transmission paths, sending an average of 225 signals per day along the risk transmission path, with a signal accuracy rate of 60%; risk front-loading cycle of 3 to 6 months, shortening the time to rank customer-related risks from 10 days to 3 days; helping the Agricultural Bank to continue to double-decrease non-performing, with the non-performing rate dropping from 2.37% to 1.81% in 2017 and the amount of new non-performing in FY2017 dropping by nearly 5.49% year-on-year.

Discussion and question-and-answer session

After Mr. Chen Meng's introduction, colleagues from the Software Development Center, Management Information Department and Shanghai Data Center showed strong interest in knowledge graph and graph computing technology, and there was a lively discussion.

Shared by: Chen Meng

From the Credit Support Group, Application Development V, Software Development Center, Has been involved in credit risk monitoring since joining the industry、 credit data applications, etc.。

look after ask: Zhao Weiping

Dong Xiaojie

Editor-in-Chief: Zhang Yong

Rotating Editor: Wu Wenbin

Li Zhuo Rong


Zhang Xiao

Rotating Reviewer: Wei Chao

ABC Big Data - Analytical Mining Squad

WeChat: ABC_DataMining

Long press to identify the QR code above and follow [Count it]

1、Cant go on Google cant find the literature cant download the full text to see here let you free seconds under HD literature
2、C voice on the use of pointers functions variables etc to give a subcategory description embedded source code program analysis
3、2018 College Salary Rankings Heavy Machinery No Better Than Specialist Programming
4、Mastercam2017 for layer understanding006
5、Watching you go from web front end geek to 32k guru dumbfounded

    已推荐到看一看 和朋友分享想法
    最多200字,当前共 发送