CN113724068A

CN113724068A - Method for constructing debtor decision strategy based on knowledge graph and related equipment

Info

Publication number: CN113724068A
Application number: CN202111012637.9A
Authority: CN
Inventors: 谢勇
Original assignee: Ping An Bank Co Ltd
Current assignee: Ping An Bank Co Ltd
Priority date: 2021-08-31
Filing date: 2021-08-31
Publication date: 2021-11-30

Abstract

The application belongs to the technical field of artificial intelligence, and particularly discloses a method for constructing a debtor decision strategy based on a knowledge graph and related equipment, wherein the method comprises the steps of constructing a corresponding financial risk knowledge graph through existing financial data, extracting financial risk characteristics of the financial data from the financial risk knowledge graph, taking the financial risk characteristics of the financial data as input, the preset machine learning model is output and trained by taking the decision strategy aiming at the financial data as a target to obtain the trained machine learning model, the debtor decision strategy can be automatically obtained according to the conditions of different debtors based on the machine learning model without artificial evaluation, a large amount of manpower and material resources are saved, therefore, the decision strategy of the debtors can be accurately judged, and the financial risks that the loans are difficult to withdraw and bad accounts appear in financial institutions such as banks and the like are effectively reduced.

Description

Method for constructing debtor decision strategy based on knowledge graph and related equipment

Technical Field

The application belongs to the technical field of artificial intelligence, and particularly relates to a method for establishing a debtor decision strategy based on a knowledge graph, a device for establishing the debtor decision strategy based on the knowledge graph, a computer readable medium and electronic equipment.

Background

Loans are one of the main business areas of financial institutions such as banks. After the loan is carried out, the financial institution often needs to monitor and evaluate the repayment capacity of the debtor so as to realize timely taking measures to urge back the loan and avoid bad accounts.

However, the existing monitoring of the clearing capacity of the debtors generally extracts features through structured data, then performs comparative analysis, finally forms evaluation of the clearing capacity of the debtors, and then executes different decision strategies of the debtors according to evaluation results, however, the existing decision strategies of the debtors are all artificially evaluated and confirmed, a large amount of manpower and material resources are consumed, and the structured data used for artificial evaluation causes low evaluation accuracy, which greatly affects the recovery of loans by financial institutions such as banks and the like.

It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present application and therefore may include information that does not constitute prior art known to a person of ordinary skill in the art.

Disclosure of Invention

The application aims to provide a method for establishing a debtor decision strategy based on a knowledge graph, a device for establishing the debtor decision strategy based on the knowledge graph, a computer readable medium and an electronic device, and at least solves the technical problems of high cost and low evaluation accuracy of manually determining the debtor decision strategy in the related technology to a certain extent.

Other features and advantages of the present application will be apparent from the following detailed description, or may be learned by practice of the application.

According to an aspect of an embodiment of the present application, there is provided a method for constructing a debtor decision strategy based on a knowledge graph, including:

constructing a corresponding financial risk knowledge graph based on the existing financial data;

mining financial risk characteristics of the financial data based on the financial risk knowledge graph, and constructing a decision strategy aiming at the financial data;

training a preset machine learning model by taking the financial risk characteristics of the financial data as input and taking a decision strategy aiming at the financial data as target output to obtain a trained machine learning model;

and extracting the financial risk characteristics of the debtors to be predicted, and inputting the financial risk characteristics of the debtors into the machine learning model to obtain a decision strategy output by the machine learning model and aiming at the debtors.

In some embodiments of the present application, based on the above technical solution, before constructing the corresponding financial risk knowledge-graph based on the existing financial data, the method further includes:

dividing the existing financial data into a behavior data dimension, a consumption data dimension and a basic information data dimension to form existing financial dimension data;

and cleaning and converting the existing financial dimension data to form financial data conforming to the knowledge graph modeling.

In some embodiments of the present application, based on the above technical solutions, a method for constructing a corresponding financial risk knowledge graph based on existing financial data includes:

constructing a debtor ontology and a contact ontology, an address ontology and a company ontology which are generated around the debtor ontology;

semantic annotation and semantic calibration are carried out on the existing financial data, and resolvable information between map entities is finally formed;

automatically identifying named entities, relationships between the entities and attribute information of the entities from existing financial data, wherein the named entities comprise the contact ontology, the address ontology and the company ontology;

and constructing a financial risk knowledge graph based on the automatically identified named entities, the relationship among the entities and the attribute information of the entities.

In some embodiments of the present application, based on the above technical solutions, the financial risk features include:

positive financial risk features including financial risk features with better debtor liquidation capacity,

a negative financial risk feature comprising a financial risk feature with poor debtor liquidation.

In some embodiments of the present application, based on the above technical solutions,

mining financial risk features of the financial data based on the financial risk knowledge graph, including:

performing sub-graph extraction on the financial risk knowledge graph based on the debtors, and extracting relevant triples of the debtors, wherein the relevant triples comprise entity, relation attribute or attribute value in the financial risk knowledge graph and triple information containing the debtors;

using the set of related triples as a financial risk feature.

In some embodiments of the present application, based on the above technical solutions, the decision policy of the financial data includes:

the positive decision-making strategy is used for expressing the decision-making strategy when the debtor compensation capacity is better, and comprises the steps of prolonging the collection time limit, reducing collection times, upgrading the client level or providing priority service;

the negative decision-making strategy is used for representing the decision-making strategy when the debtor compensation capacity is poor, and the negative decision-making strategy comprises the steps of shortening the collection hastening frequency, shortening the collection hastening time limit, setting a freezing reminder or listing as a key attention target.

In some embodiments of the present application, based on the above technical solutions, a method for training a preset machine learning model by using financial risk features of the financial data as input and using a decision strategy for the financial data as a target output to obtain a trained machine learning model includes:

continuously inputting the financial risk characteristics of the financial data into the machine learning model to obtain a decision strategy output by the machine learning model, and adjusting model parameters in the machine learning model based on an error between the decision strategy output by the machine learning model and the decision strategy aiming at the financial data until the error is smaller than a preset error threshold.

According to an aspect of an embodiment of the present application, there is provided an apparatus for constructing a debtor decision strategy based on a knowledge graph, including:

a knowledge graph construction module: the system is used for constructing a corresponding financial risk knowledge graph based on the existing financial data;

a decision strategy construction module: the financial risk knowledge graph is used for mining financial risk features based on the financial risk knowledge graph and constructing corresponding decision strategies aiming at the financial risk features;

a machine learning model building module: the system comprises a machine learning model, a decision strategy and a decision strategy, wherein the machine learning model is used for training a preset machine learning model by taking financial risk characteristics as input and taking the decision strategy as target output to obtain a trained machine learning model;

a prediction module: the method is used for extracting the financial risk characteristics of the debtors to be predicted, inputting the financial risk characteristics of the debtors into the machine learning model, and obtaining the decision strategies output by the machine learning model and aiming at the debtors.

In some embodiments of the present application, based on the above technical solution, the apparatus for constructing a debtor decision policy based on a knowledge graph further includes:

the financial data processing module is used for integrally dividing the existing financial data into a behavior data dimension, a consumption data dimension and a basic information data dimension to form the existing financial dimension data;

and the cleaning conversion module is used for cleaning and converting the existing financial dimension data to form financial data conforming to the knowledge graph modeling.

In some embodiments of the present application, based on the above technical solutions, the knowledge graph constructing module includes:

a body construction unit: the system comprises a database, a database and a database, wherein the database is used for constructing a debtor ontology and a contact ontology, an address ontology and a company ontology which are generated around the debtor;

the semantic annotation unit is used for performing semantic annotation and semantic calibration on the existing financial data to finally form resolvable information between map entities;

the entity feature extraction unit is used for automatically identifying named entities, relationships among the entities and attribute information of the entities from the existing financial data;

and the construction unit is used for constructing the financial risk knowledge graph based on the automatically identified named entities, the relationship among the entities and the attribute information of the entities.

In some embodiments of the present application, based on the above technical solutions, the financial risk feature further includes:

the graph relation type characteristic is used for calculating the out-degree or in-degree characteristic of the debtors;

the number of label values in the graph relationship is used for determining the number of debtor target relationship nodes not clearing;

the label value number of the small community is used for determining the unconfirmed client proportion of the area where the debtor is located or the unconfirmed client proportion of the business district where the debtor is located and the like;

a connected graph state used for representing the connected graph relation formed by the debtors;

and a certain relationship attribute or weight in a certain node time window period is used for representing the transaction record of the debtor in a certain time node.

In some embodiments of the present application, based on the above technical solutions, constructing a decision policy for the financial data includes:

constructing a positive decision strategy based on the positive financial risk characteristics, wherein the positive decision strategy is used for expressing a decision strategy when the debtor liquidation capacity is better, and the positive decision strategy comprises prolonging the income hastening period, reducing the income hastening times, upgrading the client level or providing priority service;

and constructing a negative decision strategy based on the negative financial risk characteristics, wherein the negative decision strategy is used for expressing a decision strategy when the debtor compensation capacity is poor, and the negative decision strategy comprises the steps of shortening the collection hastening frequency, shortening the collection period, setting a freezing reminder or listing as a key focus target.

In some embodiments of the present application, based on the above technical solution, the machine learning model building module is configured to continuously input the financial risk features of the financial data into the machine learning model, obtain the decision policy output by the machine learning model, and adjust the model parameters in the machine learning model based on an error between the decision policy output by the machine learning model and the decision policy for the financial data until the error is smaller than a preset error threshold.

According to an aspect of the embodiments of the present application, there is provided a computer readable medium, on which a computer program is stored, which when executed by a processor, implements a method for constructing a debtor decision strategy based on a knowledge graph as in the above technical solution.

According to an aspect of an embodiment of the present application, there is provided an electronic apparatus including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the executable instructions to perform a method of constructing debtor decision strategy based on knowledge-graph as in the above technical solution.

According to an aspect of embodiments herein, there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method for constructing the debtor decision strategy based on the knowledge graph as in the above technical solution.

According to the technical scheme provided by the embodiment of the application, the corresponding financial risk knowledge graph is constructed through the existing financial data, then the financial risk characteristics of the financial data are mined based on the financial risk knowledge graph, and a decision strategy for the financial data is constructed; the financial risk characteristics of the debtors can be input into the machine learning model according to the extracted financial risk characteristics of the debtors to be predicted based on the machine learning model, and the decision strategy for the debtors output by the machine learning model can be obtained. Therefore, the decision strategy of the debtor is automatically obtained according to the conditions of different debtors, manual evaluation is not needed, and a large amount of manpower and material resources are saved; and the financial risk characteristics obtained by mining the knowledge graph are used as the input of the machine learning model, so that the financial risk condition of the debtor can be accurately evaluated, the decision strategy of the debtor can be accurately judged, and the financial risks of the financial institutions such as banks and the like, such as difficult loan withdrawal and bad accounts and the like, are effectively reduced.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application. It is obvious that the drawings in the following description are only some embodiments of the application, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.

Fig. 1 schematically shows a block diagram of an exemplary system architecture to which the solution of the present application applies.

Figure 2 schematically illustrates a flow chart of a method of constructing a debtor decision strategy based on a knowledge graph.

FIG. 3 schematically illustrates a flow chart of a method of construction of a financial risk knowledge-graph.

FIG. 4 schematically illustrates a schematic of a financial risk knowledge-graph.

FIG. 5 schematically shows a diagram of a financial risk knowledge-graph in one embodiment.

Fig. 6 schematically shows a block diagram of an apparatus for constructing a debtor decision strategy based on a knowledge-graph.

FIG. 7 schematically shows a block diagram of the structure of the knowledge-graph building module.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the application. One skilled in the relevant art will recognize, however, that the subject matter of the present application can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations, or operations have not been shown or described in detail to avoid obscuring aspects of the application.

The block diagrams shown in the figures are functional entities only and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.

The flow charts shown in the drawings are merely illustrative and do not necessarily include all of the contents and operations/steps, nor do they necessarily have to be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.

With the increase of loan services of financial institutions such as banks and the like, after loans are carried out, the financial institutions often need to monitor and evaluate the repayment capacity of debtors, take measures to urge back the loans in time, and avoid bad accounts. However, monitoring of debtor liquidation capacity is generally characterized by structured data, which is data logically expressed and implemented by a two-dimensional table structure, strictly following data format and length specifications, and mainly stored and managed by a relational database. Therefore, the information about the settlement ability of the debtors, which can be obtained through the structured data, is limited, the actual settlement ability of the debtors cannot be evaluated accurately, and the problem that different decision strategies of the debtors are executed manually according to the evaluation result is more serious, and the decision strategies of the debtors cannot be judged and made accurately.

For example, a clerk of a financial institution considers that a debtor has good clearing ability according to a fact that a certain debtor pays a part of debt to the financial institution recently, so as to execute a decision policy for the debtor, such as reducing the number of times of earnings and paying the debtor to continue loan conditions, and the clerk may ignore the actual clearing ability of the debtor or the risk faced by the debtor, and therefore, the decision policy for the debtor manually made according to simple structured data is often wrong.

The method and the device can accurately evaluate the financial risk condition of the debtor, thereby realizing accurate judgment of decision strategies of the debtor, and effectively reducing the financial risks of difficult loan withdrawal and bad account occurrence of financial institutions such as banks and the like. The following technical scheme is proposed.

The embodiment of the application can acquire and process related data based on an artificial intelligence technology. Among them, Artificial Intelligence (AI) is a theory, method, technique and application system that simulates, extends and expands human Intelligence using a digital computer or a machine controlled by a digital computer, senses the environment, acquires knowledge and uses the knowledge to obtain the best result.

The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

As shown in fig. 1, system architecture 100 may include a terminal device 110, a network 120, and a server 130. The terminal device 110 may include various electronic devices such as a smart phone, a tablet computer, a notebook computer, and a desktop computer. The server 130 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud computing services. Network 120 may be a communication medium of various connection types capable of providing a communication link between terminal device 110 and server 130, such as a wired communication link or a wireless communication link.

The system architecture in the embodiments of the present application may have any number of terminal devices, networks, and servers, according to implementation needs. For example, the server 130 may be a server group composed of a plurality of server devices. In addition, the technical solution provided in the embodiment of the present application may be applied to the terminal device 110, or may be applied to the server 130, or may be implemented by both the terminal device 110 and the server 130, which is not particularly limited in this application.

According to an aspect of the embodiment of the present application, a method for constructing a debtor decision strategy based on a knowledge graph is provided, and the method may be applied to the terminal device 110, the server 130, or both the terminal device 110 and the server 130. As shown in fig. 2, fig. 2 schematically shows a flowchart of a method for constructing a debtor decision strategy based on a knowledge graph. Comprising step S210 to step S240.

Step S210: constructing a corresponding financial risk knowledge graph based on the existing financial data;

step S220: mining financial risk characteristics of financial data based on a financial risk knowledge graph, and constructing a decision strategy aiming at the financial data;

step S230: the financial risk characteristics of the financial data are used as input, a decision strategy aiming at the financial data is used as target output to train a preset machine learning model, and the trained machine learning model is obtained;

step S240: and extracting the financial risk characteristics of the debtors to be predicted, and inputting the financial risk characteristics of the debtors into the machine learning model to obtain a decision strategy output by the machine learning model and aiming at the debtors.

The decision strategy of the debtor can be automatically obtained according to the conditions of different debtors by utilizing the steps, manual evaluation is not needed, and a large amount of manpower and material resources are saved; and the financial risk characteristics obtained by mining the knowledge graph are used as the input of the machine learning model, so that the financial risk condition of the debtor can be accurately evaluated, the decision strategy of the debtor can be accurately judged, and the financial risks of the financial institutions such as banks and the like, such as difficult loan withdrawal and bad accounts and the like, are effectively reduced.

The contents of each step are specifically disclosed next.

In step S210: and constructing a corresponding financial risk knowledge graph based on the existing financial data.

The existing financial data are stored in financial institutions, the financial data mainly comprise debtors and other financial data related to the debtors, the financial data can comprise income and expenditure data, guarantee relations, loan data, credit investigation data and the like of the debtors and the related persons, the financial data corresponding to the financial institutions belonging to the Unionpay can be the existing financial data, the comprehensiveness of the data can be guaranteed, and the accurate acquisition of the settlement capacity of the debtors can be realized. The existing financial data are obtained on the premise of consent of the debtors, and generally, the situation of obtaining the financial data is clearly explained for the debtors during loan, so that the evaluation on the repayment capacity of the debtors is realized on the premise of protecting the privacy of the debtors.

The knowledge graph is a processed semantic network in nature, and is a data structure based on a graph and composed of nodes and edges. In the knowledge-graph, each node represents an "entity" existing in the real world, and each edge represents a "relationship" between entities. Knowledge-graphs are the most efficient way to represent relationships. Generally, a knowledge graph is a relational network obtained by connecting all different kinds of information together. Knowledge-graphs provide the ability to analyze problems from a "relational" perspective. And the relation network formed based on the knowledge graph is not structured data, so that more information can be obtained from the knowledge graph. Each node corresponding to the financial risk knowledge graph represents different debtors and the main body related to the debtors, each edge represents the relationship between each debtor and other people, and the actual settlement capacity of the corresponding debtor can be well determined by using the financial risk knowledge graph.

There are many ways to construct the knowledge graph, and in one embodiment of the present application, the present application discloses a method for constructing a corresponding financial risk knowledge graph based on existing financial data, as shown in fig. 3, and fig. 3 schematically illustrates a flow chart of the construction method of the financial risk knowledge graph. Comprising step S310-step S340.

Step S310: and constructing a debtor ontology and a contact ontology, an address ontology and a company ontology which are generated around the debtor.

All the financial risks and the relations of any debtor start from the debtor and then extend outwards continuously, wherein the debtor-related entities comprise a contact body, an address body related to the address of the debtor and a company body related to the enterprise corresponding to the debtor. The three ontologies represent the ontologies connected with the debtors, so that the contact ontology, the address ontology and the company ontology related to the debtors can be generated while the debtors ontology is constructed, and the main body related to the debtors can be clearly combed.

As shown in fig. 4, fig. 4 schematically shows a schematic diagram of a financial risk knowledge-graph. Included in fig. 4 are a debtor ontology 400, a contact ontology 410 generated around the debtor ontology 400, an address ontology 420, and a company ontology 430. The contact ontology 410 may be an immediate relative of the debtor ontology 400, and may include people who come and go in finance. And the address ontology 420 is the same as or related to the debtor ontology 400 address, e.g. belonging to a certain region with the debtor. And the company ontology 430 is a company related to the debtor ontology 400, such as a company with the debtor ontology 400 as a legal representative, a company invested in the debtor ontology 400, and so on.

Step S320: and performing semantic annotation and semantic calibration on the existing financial data to finally form resolvable information between map entities.

Semantic annotation refers to marking original data to make the original data contain certain semantic information. And performing semantic annotation and semantic calibration on the data recorded in the text, so that the semantic information in the data can be subjected to machine analysis. For example, the raw data may be converted into a semantic vector.

Step S330: named entities, relationships between entities, and attribute information for entities are automatically identified from existing financial data.

The named entities include the contact ontology 410, the address ontology 420, and the company ontology 430. And the relationship between the entities corresponds to the arrows in fig. 4. The attribute information 440 corresponds to attribute information 440 associated with the debtor ontology 400, such as debtor assets, lending relationships. There are not only one but also a plurality of the contact ontology 410, the address ontology 420, the company ontology 430 and the attribute information 440 in fig. 4.

Step S340: and constructing a financial risk knowledge graph based on the automatically identified named entities, the relationship among the entities and the attribute information of the entities.

Having obtained financial information about the debtor, a financial risk knowledge graph may be constructed based on the financial information. The related relationship between the debtor ontology and the generated contact ontology 410, address ontology 420 and company ontology 430, and the related relationship between the debtor ontology and the attribute information are mapped to each side of the financial risk knowledge graph, the sides in the financial risk knowledge graph represent the entity corresponding to the two connected nodes to be contacted, and different nodes include the contact ontology 410, the address ontology 420, the company ontology 430 and the attribute information 440. The financial risk knowledge graph is a graph-based data structure, and is a general expression mode of the knowledge graph based on triples, and the basic forms of the triples mainly comprise (entity 1-relation-entity 2) and (entity-relation attribute-attribute value) and the like.

The above is further illustrated by the following specific example, as shown in fig. 5, and fig. 5 schematically shows a schematic diagram of the financial risk knowledge-graph in one embodiment. Fig. 5 is a financial risk knowledge graph constructed based on a debtor a, wherein the debtor a includes five pieces of attribute information, five attribute relationships are established, and the five pieces of attribute information are respectively the debtor: 3 thousand of monthly expenses, 1 ten thousand of monthly income, 15 ten thousand of credit, 20 ten thousand of car credits and 100 ten thousand of houses are mortgage. And the entity information related to the debtor A comprises an entity E having a security relationship with the debtor A, an entity A having a couple relationship with the debtor A, an entity B having a borrowing relationship with the debtor, an entity C having the debtor as a legal representative, and an entity D having the same country with the debtor. Where entity a, entity B and entity E are all contact ontologies relating to debtors, entity C is a company ontology relating to debtors, and entity D is an address ontology 420 relating to debtors. Different attribute information also exists in each entity, which is not shown in the figure. Therefore, a corresponding financial risk knowledge graph can be constructed according to the existing financial data to determine the settlement capability of the debtors.

The present application is preceded by a step S210 for enabling construction of a corresponding financial risk knowledge-graph based on existing financial data. There is also a need to preprocess existing financial data. In one embodiment of the present application, prior to constructing the corresponding financial risk knowledge-graph based on the existing financial data, the method further comprises:

and cleaning and converting the existing financial dimension data to form financial data which accords with the knowledge graph modeling.

The part mainly converts financial data into labeled data with different dimensionalities, integrally divides the data into dimensionalities such as behavior data, consumption data and basic information data according to the provided financial data, and finally forms data conforming to the knowledge graph modeling by cleaning and converting the data. Meanwhile, the original data may have the problems of non-standard storage, non-uniform fields, Chinese and English mixing, data loss, multi-class variables and the like, and aiming at the problems, the original data is converted into regular data by adopting data cleaning, and the specific technical scheme is as follows: for the non-canonical field processing, there may be partial scrambled data and non-canonical data in the original data. For the messy code data, deleting processing is adopted; and for data which is stored in an irregular mode, converting non-standardized data into a unified standard form. For missing data processing: there may be a large number of missing fields in the original data, and there are different data processing modes for different missing situations. And deleting the contact person missing data of the contacted user for the contact data missing problem. And completing, deleting and the like the address information missing data in the address associated data.

After the financial risk knowledge graph is constructed, data mining and analysis are performed by using the financial risk knowledge graph, and the specific steps are as follows.

In step S220: and mining financial risk characteristics of the financial data based on the financial risk knowledge graph, and constructing a decision strategy aiming at the financial data.

In one embodiment of the present application, a method for mining financial risk features of financial data based on the financial risk knowledge graph includes: performing sub-graph extraction on the financial risk knowledge graph based on the debtors, and extracting relevant triples of the debtors, wherein the relevant triples comprise entity, relation attribute or attribute value in the financial risk knowledge graph and triple information containing the debtors; using the set of related triples as a financial risk feature.

The knowledge graph is a data structure based on a graph, and the basic form of the triples mainly comprises (entity 1-relation-entity 2) and (entity-relation attribute-attribute value) and the like based on the triples which are a general representation mode of the knowledge graph. Thus, there are many sub-features in the knowledge-graph, including specifically "entity, relationship attribute and attribute value", where an entity may correspond to a debtor a and a debtor B, and a relationship may indicate an association between two entities, such as a loan relationship, a guarantee relationship. The relationship attribute and the attribute value refer to some financial features associated with the entity, for example, debtor a owns 100 ten thousand deposits, then 100 ten thousand deposits are the attribute values, and owns are the corresponding relationship attributes, for example, debtor a is the main stockholder of company B, company B earns 100 ten thousand annually, then 100 ten thousand earns are the attribute values, and the relationship attributes are the stockholders. Therefore, the extraction of the knowledge graph is to extract any relevant triples related to the debtor, and the relevant triples are used as the financial risk characteristics. Step S220 of the present application is a process of extracting training samples, so the financial risk features extracted here are as many financial risk features as possible, so as to facilitate the prediction based on the financial risk features of step S240. Thus, step S220 is a process of extracting sufficient financial risk characteristics of the debtors.

And the financial risk characteristics of the financial data are a collection of characteristics that indicate the debtor's ability to clear. In one embodiment of the present application, the financial risk features include: positive financial risk characteristics including financial risk characteristics with better debtor liquidation capacity; the positive financial risk characteristics are used to indicate that the debtor has a high probability of being credited due, and for example, the financial analysis characteristics of 15 ten thousand deposits, 1 ten thousand monthly income and the like of the debtor a in fig. 4 are positive financial analysis characteristics. And the financial risk characteristics also include negative financial risk characteristics including financial risk characteristics with poor debtor liquidation. For example, in FIG. 5, 3 thousand months, 20 million car credits, a debt relationship with entity B, etc. The financial risk characteristics can be distinguished in the process of mining the financial risk knowledge graph, so that the debtor compensation capacity can be better judged.

In addition to the above simpler financial risk features, some more complex financial risk features may be formed according to mining and reprocessing of the financial risk knowledge graph, for example, in one embodiment of the present application, the financial risk features may further include:

graph relationship type feature: the system is used for calculating the out-degree or in-degree characteristics of the debtors; wherein the graph relationship type feature can calculate the income and the out degree of the debtor according to a specific certain relationship. For example, continuing with the example of fig. 5, the graph relationship feature may represent the out-to-in relationship between debtor a and entity C, corresponding to the disbursement paid out and the dividend earned during the period of debtor a being the legal representative of entity C. The graph relation type feature can be particularly the expense and income feature corresponding to any one edge of the debtor, so that all related expenses and incomes of the debtor can be clearly sorted by utilizing the graph relation type feature, and the clearing capacity of the debtor can be conveniently confirmed.

Number of label values in graph relationship: the debtor target relationship node settlement amount determining module is used for determining the debtor target relationship node settlement amount; the feature is mainly used for confirming the amount of the unsettled amount of a certain relationship node within a plurality of degrees of relationships of the debtor, for example, the debtor owes 10 entity money, wherein six entities are settled, the unsettled amount is 4, and the number of the label values in the corresponding graph relationship is 4. And can also presume the correspondent repayment amount of settlement, if repayment amount/total amount > 90%, can regard the debt as the settlement too. Therefore, by using the label value quantity in the graph relationship, the debtor can clearly know how much debt remains at present, and the settlement ability of the debtor can be evaluated.

The number of label values of the small community is as follows: the method is used for determining the unsettled customer proportion of the area where the debtor is located or the unsettled customer proportion of the business circle where the debtor is located and the like; the financial risk feature mainly plays a role in reference and borrowing reference, and the settlement capacity of the debtor is judged by referring to the unconfirmed client proportion of the area or business district where the debtor is located.

And (3) a connected graph state for representing connected graph relations formed by the debtors: for example, if a strong connectivity graph is formed to identify the mutual security behaviors of the debtors, then the clearing capacity for the debtor is poor and may be affected by the insured person associated with the debtor, resulting in the occurrence of a total overdue.

Certain relationship attribute or weight in certain node time window period: for representing transaction records within a time node of the debtor. For example, debtors transfer in/out transaction times, money amounts and the like within a certain time period after payment, and whether a client uses the loan application or not can be well explained by utilizing the characteristics; the method also comprises the transferring-in/transferring-out times, the amount and the like during the time period near the repayment date, and has a better explanation effect on whether the customer has the behavior of hiding the property.

The financial risk features are obtained by processing after mining and extracting based on the financial risk knowledge graph, for example, a certain relationship attribute or weight in a certain node time window period can be obtained by selecting time, and the income-related triples of all debtors are extracted by setting to only inquire income to form income-based financial risk features.

And the decision strategy for the financial data refers to some decision strategies about debtors made against the financial risk characteristics of the financial data. The proposal for dealing with the debtors is obtained after evaluating the liquidation capacity of the debtors. In one embodiment of the present application, constructing a decision policy for the financial data comprises: constructing a positive decision-making policy based on the positive financial risk characteristics and constructing a negative decision-making policy based on the negative financial risk characteristics. The positive decision-making strategy and the positive financial risk characteristics correspond, and the negative decision-making strategy and the negative financial risk characteristics correspond. Specifically, the method comprises the following steps:

the positive decision strategy is used for representing the decision strategy when the debtor compensation capacity is better, and the positive decision strategy can comprise prolonging the collection time limit, reducing the collection times, upgrading the client level or providing priority service and the like. The positive decision-making strategy corresponds to the positive financial risk characteristics, for example, the financial risk knowledge graph of one debtor is mined to obtain the positive financial risk characteristics, so the decision-making strategy corresponding to the debtor is the positive decision-making strategy, the positive decision-making strategy has various schemes, the income hastening period can be prolonged, and the original hastening once a month is changed into hastening once a quarter; the times of collection can be reduced, and the original three times of collection in one quarter can be changed into one time of collection in one quarter; the client level can be upgraded, for example, the client is upgraded to a member client to enjoy the treatment of the member; priority services may also be provided for the debtor. Besides the above active decision strategies, other decision strategies beneficial to the debtors can be included, such as sending small gifts, sending industry information and greetings, etc. periodically.

The financial data decision strategy also comprises a negative decision strategy, the negative decision strategy is used for representing the decision strategy when the debtor compensation capacity is poor, and the negative decision strategy comprises the steps of shortening the collection hastening frequency, shortening the collection hastening time limit, setting a freezing reminding or listing as a key focus target. Negative decision strategies correspond to negative financial risk features, e.g., where the financial risk knowledge-graph mining of a debtor is negative financial risk features, then the decision strategy corresponding to that debtor is a negative decision strategy. The negative decision strategy has various schemes, namely the collection hastening frequency can be shortened, the collection hastening is carried out once in one month, and the collection hastening is carried out three times in one month at present; the time limit of the induced harvest can be shortened, the original induced harvest at the bottom of each month is changed into the induced harvest at the beginning of each month; freezing reminding can be set, and when the excavated negative financial risk features are more and the loan cannot be cleared, the financial institution can be reminded to freeze the deposit of the debtor in time, so that bad accounts are avoided; the method can also be classified as a key focus target, and each debtor who corresponds to the key focus target can pay close attention to the input and output and various financial behaviors, so that the problem that loans cannot be settled is avoided. The negative decision strategies not only comprise the above strategies, but also comprise other decision strategies which are beneficial to returning loans and reducing the bad account risk of the financial institution.

After the financial risk features of the financial data are mined from the financial risk knowledge graph and a decision strategy for the financial data is constructed, artificial intelligence can be used for learning, and the decision strategy for automatically obtaining the financial data based on the financial risk features of the financial data is realized.

In step S230: and (4) training a preset machine learning model by taking the financial risk characteristics of the financial data as input and taking the decision strategy aiming at the financial data as target output to obtain the trained machine learning model.

The task of training the completed machine learning model is to input a financial risk feature to obtain a decision strategy based on the financial risk feature. The application utilizes a large enough number of financial risk features and a sufficient number of decision strategies during training, and different financial risk features or combinations of financial risk features correspondingly generate different decision strategies. The rule that the corresponding relation between the financial risk characteristics and the decision strategy is followed by the method is that positive financial risk characteristics correspond to positive decision strategies, and negative financial risk characteristics correspond to negative decision strategies. And for the hybrid financial risk characteristics and the passive financial risk characteristics, comprehensive evaluation is needed to determine the corresponding decision strategies.

The correspondence between positive decision strategies and positive financial risk characteristics is not simply a one-to-one or one-to-many relationship. For example, continuing with the example of fig. 5, the debtor a in fig. 5 has 15 ten thousand deposits, the monthly income is 1 ten thousand, and the decision strategy obtained by combining the two conditions is to extend the term of the induced income, but if the two conditions are mixed with other conditions, the corresponding decision strategy is changed, for example, if the debtor a has 20 ten thousand car credits although there are 15 ten thousand deposits, the decision strategy at this time will not extend the term of the induced income, and may not perform any strategy, and the current situation is maintained. Keeping the current situation is also a decision strategy.

Different grades can be set for the positive decision-making strategy and the negative decision-making strategy, different grades can be set for the positive financial risk characteristic and the negative financial risk characteristic, and the client or the server can correspond to the different grades when determining the corresponding relation between the financial risk characteristic and the decision-making strategy. For example, taking a negative decision policy as an example, a decision policy of shortening the hastening frequency and shortening the hastening period may be defined as a primary decision policy, and a freezing reminder is set and a primary decision policy is set as a secondary decision policy. Correspondingly, the negative financial risk characteristics can be graded according to the negative degree of the debtor, for example, a debtor only borrows 2 thousands of money from other people, the influence on the slight degree of the clearing capacity can be defined as a primary financial risk characteristic, the corresponding primary decision strategy is, for a debtor mortgage house, the influence on the clearing capacity of the debtor is greater by providing high-value guarantee for other people, and the like, the corresponding secondary financial risk characteristic can be defined as a secondary decision strategy. The setting of the positive decision-making strategy is consistent with the negative decision-making strategy, and the setting can be carried out based on financial risk characteristics of different levels, so that the machine learning model can fully learn.

The financial risk characteristics of debtors in life are complex and are mixed positive financial risk characteristics and negative financial risk characteristics, and the decision strategy can be made in various ways for the decision strategy. The first way can be set by setting a value for different financial risk characteristics and also based on the above influence degree on the debtor clearing capacity, the value is larger for a large influence degree, the corresponding value is larger, the value is positive for a positive financial risk characteristic, the negative financial risk characteristic is negative, finally the values obtained by the debtor are added, if the obtained value is positive, the comprehensive clearing capacity of the debtor is better, a positive decision strategy can be carried out, and if the obtained value is negative, the comprehensive clearing capacity of the debtor is poorer, the negative decision strategy can be carried out. For example, continuing with the example of fig. 5, if the debtor a in fig. 5 has a deposit value of +2 for 15 ten thousand, a monthly income value of +1 for 1 ten thousand, a car loan value of-1 for 20 ten thousand, and a house value of-1 for 100 ten thousand, the value obtained by the comprehensive addition is +1, and a positive decision strategy, such as extending the term of collection, can be performed.

The second method can correspond the financial risk characteristics and the decision strategies in a mode of not influencing each other, namely, the positive financial risk characteristics and the negative financial risk characteristics of a debtor correspond to different decision strategies respectively, and a mode of paralleling a plurality of decision strategies is realized. For example, continuing with the example of fig. 5, if debtor a in fig. 5 has 15 ten thousand deposits, the monthly income is 1 ten thousand, the corresponding decision policy is to extend the hastening period, and the decision policy for 20 ten thousand car credits and mortgage 100 ten thousand houses is to shorten the hastening frequency, then debtor a can be subjected to the decision policy for both extending the hastening period and shortening the hastening frequency.

The decision-making strategy and the financial risk characteristics are correspondingly arranged to facilitate the learning of the machine learning model, so that the correct decision-making strategy can be accurately obtained based on the compensation capability, and the loss of a financial institution is avoided. Therefore, different corresponding relations can be preset according to different financial institutions, and the method is not limited to the above corresponding relations.

For the machine learning module of the present application, there are various specific training modes, and in an embodiment of the present application, a method for training a preset machine learning model by using financial risk features of financial data as input and using a decision policy for the financial data as a target output to obtain a trained machine learning model is disclosed, which specifically includes:

and continuously inputting the financial risk characteristics of the financial data into the machine learning model to obtain a decision strategy output by the machine learning model, and adjusting model parameters in the machine learning model based on the error between the decision strategy output by the machine learning model and the decision strategy aiming at the financial data until the error is smaller than a preset error threshold.

The output of the machine learning model can be adjusted through the steps, and large deviation is avoided.

After the machine learning model is constructed, a decision strategy can be automatically obtained based on the machine learning model, and the specific steps are as follows.

In step S240: and extracting the financial risk characteristics of the debtors to be predicted, and inputting the financial risk characteristics of the debtors into the machine learning model to obtain a decision strategy output by the machine learning model and aiming at the debtors.

Here, the financial risk characteristics of the debtor to be predicted are a part of the financial risk characteristics of the financial data in step S220, and the decision policy for the debtor is one or more of the decision policies for the financial data in step S220. After the machine learning model is obtained, the financial risk characteristics of the debtor to be predicted can be input into the machine learning model, and the machine learning model is used for outputting one or more decision strategies aiming at financial data based on the input of one or more financial risk characteristics aiming at the financial data. The financial risk characteristics of the debtor to be predicted are input, the decision strategy for the debtor can be obtained, and therefore corresponding behaviors are carried out on the debtor according to the decision strategy, and the risk that a financial institution cannot recover loans is reduced.

The method comprises the steps of constructing a corresponding financial risk knowledge graph through existing financial data, then mining financial risk characteristics of the financial data based on the financial risk knowledge graph, and constructing a decision strategy for the financial data; the financial risk characteristics of the debtors can be input into the machine learning model according to the extracted financial risk characteristics of the debtors to be predicted based on the machine learning model, and the decision strategy for the debtors output by the machine learning model can be obtained. Therefore, the decision strategy of the debtor is automatically obtained according to the conditions of different debtors, manual evaluation is not needed, and a large amount of manpower and material resources are saved; and the financial risk characteristics obtained by mining the knowledge graph are used as the input of the machine learning model, so that the financial risk condition of the debtor can be accurately evaluated, the decision strategy of the debtor can be accurately judged, and the financial risks of the financial institutions such as banks and the like, such as difficult loan withdrawal and bad accounts and the like, are effectively reduced.

The method for constructing the debtor decision strategy based on the knowledge graph is introduced above, and the device corresponding to the method is introduced next.

According to an aspect of the embodiment of the present application, there is provided an apparatus 600 for constructing a debtor decision policy based on a knowledge graph, as shown in fig. 6, where fig. 6 schematically shows a block diagram of an apparatus for constructing a debtor decision policy based on a knowledge graph. The method comprises the following steps:

the knowledge-graph building module 610: the system is used for constructing a corresponding financial risk knowledge graph based on the existing financial data;

decision policy building module 620: the system is used for mining financial risk features based on a financial risk knowledge graph and constructing a corresponding decision strategy aiming at the financial risk features;

machine learning model building module 630: the system comprises a machine learning model, a decision strategy and a decision strategy, wherein the machine learning model is used for training a preset machine learning model by taking financial risk characteristics as input and taking the decision strategy as target output to obtain a trained machine learning model;

the prediction module 640: the method is used for extracting the financial risk characteristics of the debtors to be predicted, inputting the financial risk characteristics of the debtors into the machine learning model, and obtaining the decision strategies output by the machine learning model and aiming at the debtors.

In an embodiment of the present application, the apparatus for constructing a debtor decision policy based on a knowledge graph further includes:

In one embodiment of the present application, as shown in FIG. 7, FIG. 7 schematically illustrates a block diagram of the structure of the knowledge-graph building module. The knowledge-graph building module 610 includes:

the ontology building unit 710: the system comprises a database, a database and a database, wherein the database is used for constructing a debtor ontology and a contact ontology, an address ontology and a company ontology which are generated around the debtor;

the semantic annotation unit 720 is used for performing semantic annotation and semantic calibration on the existing financial data to finally form resolvable information between map entities;

an entity feature extraction unit 730, configured to automatically identify named entities, relationships between entities, and attribute information of the entities from existing financial data;

a construction unit 740 for constructing a financial risk knowledge graph based on the automatically identified named entities, the relationships between the entities, and the attribute information of the entities.

The financial risk characteristics of the financial data mentioned in the apparatus 600 for constructing a debtor decision policy based on a knowledge graph according to the present application are mentioned in step S220, and are not described herein again.

In one embodiment of the present application, the machine learning model building module 630 of the present application is configured to continuously input the financial risk features of the financial data into the machine learning model, obtain the decision-making strategy output by the machine learning model, and adjust the model parameters in the machine learning model based on an error between the decision-making strategy output by the machine learning model and the decision-making strategy for the financial data until the error is smaller than a preset error threshold.

The above section discloses the content of the apparatus 600 for constructing debtor decision strategy based on knowledge-graph of the present application, and then proceeds to disclose other aspects of the present application.

According to an aspect of an embodiment of the present application, there is provided an electronic apparatus including: a processor; and a memory for storing executable instructions for the processor; wherein the processor is configured to execute the method of constructing debtor decision strategy based on knowledge-graph as in the above technical solution via executing the executable instructions.

It should be noted that although the various steps of the methods in this application are depicted in the drawings in a particular order, this does not require or imply that these steps must be performed in this particular order, or that all of the shown steps must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions, etc.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

It should be noted that although in the above detailed description several modules or units of the device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the application. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.

Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains.

It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. A method for constructing a decision strategy of a debtor based on a knowledge graph is characterized by comprising the following steps:

2. The method for knowledge-graph-based construction of debtor decision strategies according to claim 1, wherein prior to constructing a corresponding financial risk knowledge graph based on existing financial data, the method further comprises:

3. The method for knowledge-graph-based construction of debtor decision strategies according to claim 1, wherein constructing a corresponding financial risk knowledge graph based on existing financial data comprises:

4. The method for knowledge-graph-based construction of debtor decision strategies according to claim 1, wherein the financial risk features comprise:

5. The method of knowledge-graph based construction of debtor decision strategy of claim 1, wherein mining financial risk features of the financial data based on the financial risk knowledge graph comprises:

using the set of related triples as a financial risk feature.

6. The method of knowledgegraph-based construction of debtor decision strategies according to claim 4, wherein constructing a decision strategy for the financial data comprises:

7. The method for constructing a decision strategy for a debtor based on a knowledge graph according to claim 1, wherein the step of training a preset machine learning model by taking financial risk characteristics of the financial data as input and taking a decision strategy for the financial data as target output to obtain a trained machine learning model comprises the steps of:

8. An apparatus for constructing a decision strategy for a debtor based on knowledge graph, comprising:

9. A computer readable medium having stored thereon a computer program which, when executed by a processor, implements the method of knowledge-graph based construction of debtor decision strategy according to any of claims 1 to 7.

10. An electronic device, comprising:

a processor; and

a memory for storing executable instructions of the processor;

wherein the processor is configured to perform the method of knowledge-graph-based construction of debtor decision strategy of any of claims 1 to 7 via execution of the executable instructions.