CN113849618A

CN113849618A - Strategy determination method and device based on knowledge graph, electronic equipment and medium

Info

Publication number: CN113849618A
Application number: CN202111118226.8A
Authority: CN
Inventors: 李烨; 夏小华
Original assignee: Ping An Bank Co Ltd
Current assignee: Ping An Bank Co Ltd
Priority date: 2021-09-23
Filing date: 2021-09-23
Publication date: 2021-12-28

Abstract

The embodiment of the application discloses a strategy determination method, a strategy determination device, electronic equipment and a strategy determination medium based on a knowledge graph, and relates to the technical field of artificial intelligence and digital medical treatment. The method can comprise the following steps: when a handling policy request for a target resource of a target object is detected, object data of the target object is acquired; and performing knowledge extraction based on at least one piece of dimension information and object data to obtain an object data knowledge graph, determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, determining a score of each piece of dimension information for each candidate strategy according to the attribute information and the adaptation range information, and determining a target disposal strategy according to the score of at least one piece of dimension information of each candidate strategy. The method and the device are beneficial to improving the accuracy of the strategy determination. The embodiment of the application can also be applied to the field of artificial intelligence, for example, the object data of the target object can be processed based on the artificial intelligence technology to obtain the object data knowledge graph.

Description

Strategy determination method and device based on knowledge graph, electronic equipment and medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, an electronic device, and a medium for determining a policy based on a knowledge graph.

Background

Resources for a subject typically need to determine their corresponding handling policies, such as determining a handling policy for bad resources of the subject. Determining the treatment policy corresponding to the resource of the subject is generally based on experience of policy determination personnel, performing analysis and judgment based on some basic information of the subject, and then determining the corresponding treatment policy. However, the experience of the policy determination personnel is usually compared in such a manner, and if the experience of the policy determination personnel is low, it is very likely that a proper policy cannot be accurately determined, which causes an error in disposing of resources and has low accuracy of policy determination.

Disclosure of Invention

The embodiment of the application provides a strategy determination method, a strategy determination device, electronic equipment and a strategy determination medium based on a knowledge graph, and the method, the device, the electronic equipment and the medium are favorable for improving the accuracy of strategy determination.

In one aspect, an embodiment of the present application discloses a method for determining a policy based on a knowledge graph, where the method includes:

when a handling policy request for a target resource of a target object is detected, acquiring object data of the target object;

determining at least one dimension information for the treatment policy request, and performing knowledge extraction based on the at least one dimension information and the object data to obtain an object data knowledge graph for the target object; the at least one dimension information is information of a reference dimension when determining a treatment policy for the target object, and each target relationship in the object data knowledge graph corresponds to the at least one dimension information respectively;

obtaining a plurality of candidate strategies, wherein each candidate strategy is associated with adaptation range information aiming at the at least one dimension information;

determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, and determining a score of each piece of dimension information of each candidate strategy of the target object according to the attribute information of each piece of dimension information of the target object and the adaptive range information corresponding to each piece of dimension information of each candidate strategy;

determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of at least one dimension information of each candidate strategy;

determining a target handling policy for the handling policy request from the plurality of candidate policies according to the total score corresponding to each candidate policy.

In another aspect, an embodiment of the present application discloses a knowledge-graph-based policy determination apparatus, including:

an acquisition unit configured to acquire object data of a target object when a handling policy request for a target resource of the target object is detected;

a processing unit, configured to determine at least one dimension information for the treatment policy request, and perform knowledge extraction based on the at least one dimension information and the object data, to obtain an object data knowledge graph for the target object; the at least one dimension information is information of a reference dimension when determining a treatment policy for the target object, and each target relationship in the object data knowledge graph corresponds to the at least one dimension information respectively;

the obtaining unit is further configured to obtain a plurality of candidate policies, where each candidate policy is associated with adaptation range information for the at least one dimension information;

the processing unit is further configured to determine attribute information corresponding to each piece of dimension information according to the object data knowledge graph, and determine a score of each piece of dimension information of each candidate policy for the target object according to the attribute information of each piece of dimension information of the target object and adaptation range information corresponding to each piece of dimension information of each candidate policy;

the processing unit is further configured to determine a total score corresponding to each candidate policy corresponding to the target object according to the score of the at least one piece of dimension information of each candidate policy;

the processing unit is further configured to determine a target handling policy for the handling policy request from the plurality of candidate policies according to the total score corresponding to each candidate policy.

In yet another aspect, an embodiment of the present application provides an electronic device, which includes a processor and a memory, where the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to perform the following steps:

In another aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer program instructions are stored, and when executed by a processor, the computer program instructions are configured to perform the following steps:

In yet another aspect, embodiments of the present application disclose a computer program product or a computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read by a processor of a computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to execute the knowledge-graph based policy determination method.

In the embodiment of the present application, when a handling policy request for a target resource of a target object is detected, object data of the target object may be acquired; determining at least one piece of dimension information for the treatment strategy request, extracting knowledge based on the at least one piece of dimension information and the object data to obtain an object data knowledge graph for the target object, determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, determining a score of the target object for each piece of dimension information of each candidate strategy according to the attribute information of each piece of dimension information of the target object and adaptation range information corresponding to each piece of dimension information of each candidate strategy, and determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of the at least one piece of dimension information of each candidate strategy; a target handling policy for the handling policy request is determined from the plurality of candidate policies according to the total score corresponding to each candidate policy. Therefore, multi-dimensional attribute information of the target object can be extracted based on the knowledge graph technology, the target object is analyzed in a multi-dimensional mode, the target disposal strategy is determined, and accuracy of strategy determination is improved.

Drawings

In order to more clearly illustrate the technical strategy of the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without inventive labor.

FIG. 1 is a schematic flow chart diagram illustrating a knowledge-graph-based policy determination method according to an embodiment of the present disclosure;

FIG. 2 is a schematic diagram illustrating the effect of an object data knowledge-graph provided by an embodiment of the present application;

FIG. 3 is a flowchart illustrating a total score obtaining process according to an embodiment of the present disclosure;

FIG. 4 is a schematic flow chart diagram illustrating a knowledge-graph-based policy determination method according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a knowledge-graph-based policy determination apparatus according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

Technical strategies in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The application provides a strategy determination scheme based on a knowledge graph, which can acquire object data of a target object when a treatment strategy request aiming at target resources of the target object is detected, then determine at least one dimension information, obtain a target data knowledge graph based on the at least one dimension information and the object data to determine attribute information of the target object aiming at each dimension information, and further determine a target treatment strategy according to the attribute information of the target object at each dimension information and a candidate scheme, so that the attribute information corresponding to at least one dimension information for strategy determination can be quickly extracted by constructing the knowledge graph, and the target treatment strategy is determined by quantizing scores corresponding to each dimension information, thereby being beneficial to improving the accuracy of strategy determination.

The technical solution of the present application may be applied to an electronic device, where the electronic device may be a terminal, a server, or other devices for performing policy determination based on a knowledge graph, and the present application is not limited. The application is operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In a possible implementation manner, the embodiment of the present application may be applied to the field of artificial intelligence, for example, object data of a target object may be processed based on an artificial intelligence technology to obtain an object data knowledge graph. 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.

In a possible embodiment, the solution of the present application can also be applied in the field of medical technology. Specifically, the target resource of the target object may be medical expenses which are not paid in time by the target object, and then when a disposal policy request for the medical expenses which are not paid in time by the target object is detected, object data of the target object is acquired, attribute information corresponding to at least one piece of dimensional information for policy determination is quickly extracted by constructing a knowledge graph, and a target disposal policy is determined by quantifying scores corresponding to the dimensional information, so that the medical expenses which are not paid in time by the target object are disposed.

Based on the above description, the embodiment of the present application provides a strategy determination method based on a knowledge graph. Referring to fig. 1, fig. 1 is a schematic flowchart of a method for determining a knowledge-graph-based policy according to an embodiment of the present application. The method may be performed by the above mentioned electronic device. The method may include the following steps.

S101, when a handling strategy request aiming at a target resource of a target object is detected, object data of the target object is obtained.

The target object is any object needing to determine a disposal policy, and the target resource is a resource needing to be disposed. For example, when it is detected that the target resource of the target object satisfies the policy determination condition, a handling policy request for the target resource of the target object is generated. The policy determination condition may be a trigger condition for handling policy requests. If the target object is not paid for the payment beyond a certain period, a request for a handling policy for the non-paid resource of the target object (i.e. the target resource) is generated.

The handling policy request may be a request to determine a handling policy for the target resource. Optionally, the handling policy request may carry an object identifier of the target object, such as an object name, a unique identification code, and the like of the target object, which is not limited herein.

The object data of the target object may be base object data and retrieval object data for the target object. The basic object data may be information for recording some bases for the target object, and the basic object data may be object data derived based on a business system for each object, such as data of an area of the object, a main business, whether to go to a company, guaranteed object information, object transaction information, and the like, which is not limited herein. The retrieval object data can be data which is retrieved in a financial report issuing system, an enterprise inquiry system, a news issuing system and the like with authorization and aims at a target object. For example, if the target object is a person, the search may be performed based on information such as a name and a employment organization of the target object, and information such as complaint information that the search target data for the target object may be the target object is obtained, which is not limited herein; for another example, if the target object is an enterprise, the target object may be searched based on information such as a name of the target object, a legal representative name, and a name of a related object, so as to obtain information such as complaint information, enterprise financial reports, and enterprise financing conditions, which may be the target object, for the searched customer data of the target object, which is not limited herein. Optionally, the basic object data may be structured data describing attribute values corresponding to various categories of information of the target object, and the retrieved object data may be related text data retrieved by the target object.

S102, determining at least one dimension information for the treatment strategy request, and performing knowledge extraction based on the at least one dimension information and the object data to obtain an object data knowledge graph for the target object.

Wherein the at least one dimension information is information of a reference dimension when determining the treatment policy for the target object. For example, the at least one dimension information may be information of dimensions of an area of the object, a main business, a company on sale, a company income, guarantee object information, object complaint information, object transaction information, and the like, which is not limited herein. The at least one dimension information may then be referenced when determining information for a target resource of the target object.

The object data knowledge graph is a knowledge graph constructed based on the object data, and each target relation in the object data knowledge graph corresponds to at least one piece of dimension information respectively. The target relationship may be a relationship that two entities connected have in the object data knowledge graph. For example, if the target relationship between the entity a and the entity b in the object data knowledge graph is the number of object complaints that the entity b is the entity a, the dimension information corresponding to the target relationship is the number of object complaints.

It is to be understood that the object data indication map may be composed of a plurality of triple data, the triple data may represent { subject, predicate, object }, and in the knowledge map, the triple data may be { entity, relationship, entity }, or { entity, attribute }, where entity and attribute are collectively referred to as entity, relationship and attribute are collectively referred to as relationship, that is, the subject in the triple data is an entity, and the object in the triple data is referred to as an entity regardless of whether it is an attribute value of the entity corresponding to the subject or an associated entity, and the predicate in the triple data is a relationship between the entity corresponding to the subject and the entity corresponding to the object. For example, it can be extracted from the text data that the profit of the enterprise a is 1 hundred million yuan that the first entity (i.e., the entity corresponding to the subject in the three sets of data) may be "enterprise a", the second entity (i.e., the entity corresponding to the subject in the three sets of data) may be "1 hundred million yuan", the relationship between the two entities may be the profit, and the corresponding triple data may be { enterprise a, profit, 1 hundred million yuan }; for another example, from the text data "three legal representatives of business a" may be extracted that the first entity may be "business a" and the second entity may be "three legal representatives", and the relationship between the two may be the legal representatives.

In a possible implementation, when the knowledge extraction is performed based on at least one dimension information and the object data to obtain the object data knowledge graph for the target object, a corresponding knowledge extraction method may be adopted for the basic object data and the retrieval object data. For example, if the basic object data may be structured data, the basic object data may be converted into triple data by using a D2R (a tool capable of converting structured data into triple data), which is not described herein. For another example, the retrieval object data may be retrieved text data, and then entity extraction may be performed on the retrieval object data based on a supervised learning method, and then relationship extraction is performed based on the extracted entity, so as to obtain triple data for the retrieval object data.

In a possible implementation, performing knowledge extraction based on at least one dimension information and object data to obtain an object data knowledge graph for a target object, may include the following steps: determining a plurality of entity data in the object data; determining a corresponding target relationship between each entity data according to the entity data and the object data; forming at least one ternary group of data according to the entity data with the target relation and the corresponding target relation; an object data knowledge graph for the target object is constructed from the at least one triple data set.

The object data may be text data corresponding to the search object data. The entity data may be information of people, objects, places, organizations, attribute values, etc. in the object data, and is not limited herein. For example, based on the retrieval target data "xxx business annual income of last year up to 1 million yuan", the entity data "xxx business" "1 million yuan" can be extracted. After obtaining the plurality of entity data, the target relationships between the entities may be extracted from the object data, and the target relationships may be any one of the target relationships respectively corresponding to the at least one piece of dimension information. Optionally, in the above entity data, there may be some entity data having no relationship therebetween. The triple data may be triple data configured based on entity data having a target relationship and a corresponding target relationship, the entity data having the target relationship may be a subject and an object in the triple data, respectively, and the corresponding target relationship may be a predicate in the triple data. For example, if the objective relationship between the entity data "xxx business" and "1 million yuan" is annual income, the triple data { xxx business, annual income, 1 million yuan } is obtained.

When the object data knowledge graph for the target object is constructed according to the triple data, two entities in the triple data can be respectively determined as vertexes, the relation in the triple data is determined as an edge connecting the two vertexes, that is, each vertex in the object data knowledge graph represents an entity in the triple data, each edge represents the relation between the entities corresponding to the two vertexes connected by the edge, then, based on each triple data in the at least one triple data, the corresponding two vertexes and edges can be determined, and the same vertexes of the multiple triple data are combined into the same vertex, so that the knowledge graph for the object data of the target object can be obtained. For example, please refer to fig. 2, fig. 2 is a schematic diagram illustrating an effect of an object data knowledge graph according to an embodiment of the present application. If the extracted at least one triplet data includes { object a, complaint frequency, 5}, { object a, annual income, 3 billion }, { object a, legal person representative, object B }, then an object data knowledge graph as shown in fig. 2 can be obtained based on the at least one triplet data.

In a possible implementation manner, determining a plurality of entity data in the object data may train an entity extraction model for identifying the entity data in the object data based on a deep learning method, and then the object data may be input into the entity extraction model, so as to obtain a plurality of entity data for the object data. For example, the entity extraction model may be a model combining LSTM (Long Short-Term Memory network) and CRF (Conditional Random Fields). After the initial entity extraction model is trained through the sample object data with the entity label information, an entity extraction model capable of identifying the entity data in the object data can be obtained. Wherein the entity tag information is used to indicate actual entity data in the sample object data.

In a possible implementation manner, the determining the corresponding target relationship between the entity data may be inputting a plurality of entity data and object data into a target classification model, so as to obtain the corresponding target relationship between the entity data. The target classification model may be a classification model such as SVM, NN, Naive Bayes, etc., and is not limited herein. Therefore, the target relation corresponding to the at least one dimension information can be used as the optional classification of the target classification model, the relation extraction process is converted into a classification problem, and the target relation between the entity data can be determined through the target classification model by inputting the object data and the entity data.

In a possible implementation manner, before a plurality of entity data and object data are input into the target classification model to obtain corresponding target relationships between the entity data, the target classification model may be trained, which specifically includes the following steps: determining a relation set used for constructing an object data knowledge graph, wherein at least one target relation in the relation set corresponds to at least one dimension information; acquiring sample data, wherein the sample data is provided with a sample label, the sample label is used for indicating an actual relation among all entity data in the sample data, and the actual relation belongs to a relation set; inputting sample data into an initial classification model to obtain a prediction relation aiming at entity data in the sample data, wherein the prediction relation belongs to a relation set; and correcting the model parameters of the initial classification model according to the actual relation between the prediction relation and the entity data to obtain a target classification model.

The relationship set may include at least one target relationship for constructing an object data knowledge graph, and the at least one target relationship in the relationship set corresponds to at least one dimension information respectively. That is, when determining the relationship in the object data knowledge graph, only information corresponding to a reference dimension that needs to be policy-determined subsequently may be determined, thereby improving the efficiency of generating the object data knowledge graph.

The sample data may include sample object data and entity data in the sample object data, and the sample tag may indicate an actual relationship between the entity data in the sample data, where the actual relationship is a target relationship in the relationship set. Optionally, the initial classification model in the sample data may be a classification model that is not trained based on the sample data, and it can be understood that the model structure of the initial classification model is the same as that of the target classification model. And then after the sample data is input into the initial classification model, a prediction relation aiming at each entity data in the sample data can be obtained, wherein the prediction relation is also one target relation in the relation set. And correcting the model parameters of the initial classification model according to the actual relation between the prediction relation and the entity data, calculating a corresponding loss value for calling a loss function based on the prediction relation and the actual relation, determining the initial detection model as a target detection model if the loss value is less than a threshold value, correcting the model parameters of the initial detection model if the loss value is more than the threshold value, and then training the initial detection model after the model parameters are corrected through the sample data until the loss value meets a preset condition. It can be appreciated that the initial classification model can be trained over multiple sample data, such that the initial classification model has better ability for relationship extraction.

S103, obtaining a plurality of candidate strategies.

Wherein each candidate policy is associated with adaptation range information for at least one dimension information. The multiple candidate policies may be different policies for handling the resource, and the adaptation information range of each candidate policy is not identical. For example, the candidate policy may be to reclaim resources in the form of prosecution, or to reclaim resources in the form of solicitation, or to reduce losses in the form of replacement mortgages. The adaptation range information may be information indicating a range to which the candidate policy is applicable, and the adaptation range information for each dimension information includes an adaptation range corresponding to the dimension information.

For example, the at least one dimension information includes complaint number information, and the candidate policy a and the candidate policy b both have corresponding adaptation range information for the complaint number information, such as the adaptation range included in the adaptation range information corresponding to the complaint number information (i.e. one dimension information) of the candidate policy a is 1 to 5, that is, the candidate policy a is applicable to the subject with complaint number between 1 and 5, and the adaptation range included in the adaptation range information corresponding to the complaint number information (i.e. one dimension information) of the candidate policy b is 6 to 10, that is, the candidate policy b is applicable to the subject with complaint number between 6 and 10.

S104, determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, and determining the score of the target object for each piece of dimension information of each candidate strategy according to the attribute information of each piece of dimension information of the target object and the adaptive range information corresponding to each piece of dimension information of each candidate strategy.

The attribute information may be information corresponding to each dimension information of the target object. The attribute information may include attribute values, that is, the attribute information corresponding to each dimension information includes the attribute value corresponding to each dimension information. For example, the at least one piece of dimension information includes complaint frequency information, and if the complaint frequency of the target object is 3 times, which can be obtained from the target data knowledge graph, the attribute value of the attribute information corresponding to the complaint frequency information (i.e., one piece of dimension information) of the target object is 3.

In a possible implementation manner, the determining, according to the object data knowledge graph, the attribute information corresponding to each piece of dimension information may be determining, according to the dimension information, a target relationship corresponding to the dimension information, determining two pieces of entity data associated with the target relationship, and determining, as the attribute information corresponding to the dimension information, another piece of entity data, except the piece of entity data corresponding to the target object, of the two pieces of entity data. For example, as shown in the object data knowledge graph shown in fig. 2, the target object is the object a, and if the attribute information for the number of complaints (i.e., one dimension information) is to be determined, the target relationship corresponding to the number of complaints may be determined, and further, the entity data associated with the target relationship corresponding to the number of complaints, that is, as shown in fig. 2, objects a and 5, is determined, another entity data excluding the object a is determined as the attribute information (i.e., 5), that is, the attribute value of the attribute information for the dimension information, which is the number of complaints, is 5. Similarly, the attribute value of the attribute information for the dimensional information of the annual profit is 3 hundred million, and the attribute value of the attribute information for the legal representative of the dimensional information is the object B.

In a possible implementation manner, the at least one piece of dimension information includes target dimension information, the attribute information corresponding to each piece of dimension information includes an attribute value corresponding to each piece of dimension information, and the adaptation range information of each piece of dimension information includes an adaptation range of each piece of dimension information. Further, determining the score of the target object for each piece of dimension information of each candidate policy according to the attribute information of each piece of dimension information of the target object and the adaptation range information corresponding to each piece of dimension information of each candidate policy may include the following steps: when the adaptation range corresponding to the adaptation range information of the target dimension information in the candidate strategy is detected, including the attribute value corresponding to the attribute information of the target dimension information in the dimension information of the target object, determining the score of the target dimension information in the candidate strategy as a first score; when the adaptation range corresponding to the adaptation range information of the target dimension information in the candidate strategy is detected, the attribute value corresponding to the attribute information of the target dimension information in the dimension information of the target object is not included, the score of the target dimension information in the candidate strategy is determined to be a second score, and the first score is larger than the second score.

The first score may be a score corresponding to the dimension information when the adaptation range includes the attribute value, and the first score is greater than zero. The second score may be a score corresponding to the dimension information when the adaptation range does not include the attribute value, and the second score is greater than zero and smaller than the first score. For example, the first score is preset to be 10, the second score is 2, the adaptation range of the candidate policy for the adaptation range information of the dimension information of the number of complaints is 1 to 5, if the attribute value of the target object for the attribute information of the dimension information of the number of complaints is 3, the attribute value belongs to the adaptation range, and the score of the target object for the dimension information of the number of complaints of the candidate policy is 10, which is the first score; if the attribute value of the target object for the attribute information of the dimension information of the number of complaints is 6, the attribute value does not belong to the adaptation range, and the score of the target object for the dimension information of the number of complaints of the candidate policy is 5, which is the second score. Therefore, the score corresponding to each dimension information of each candidate strategy can be obtained.

In a possible implementation, the at least one dimension information includes target dimension information, and determining the adaptation range information that the candidate policy has for the at least one dimension information may include the following steps: obtaining a plurality of historical policy data, wherein each historical policy data comprises historical object data of a historical object and a historical disposal policy, the historical disposal policies in the plurality of historical policy data are the same, and the historical disposal policy corresponds to any one of the candidate policies; respectively determining attribute information aiming at the target dimension information according to the object data of each historical object to obtain a plurality of attribute information aiming at the target dimension information; and determining adaptation range information of the target dimension information in the candidate strategies corresponding to the historical treatment strategies according to the plurality of attribute information.

Wherein, the target dimension information may be any dimension information in the at least one dimension information.

The historical policy data may be policy data that works well for the disposal of resources. The historical object data is object data of an object for which the historical policy data is intended, and the historical disposal policy is a disposal policy of a resource for which the historical policy data is intended. The data source of the historical policy data may be the object data of the object corresponding to the resource that has been handled in the business system and the handling policy corresponding to the resource, for example, may be generated from the business system. Attribute information for the target dimension information is determined according to the historical object data of each historical object, and a corresponding historical object data knowledge graph can be generated based on the historical object data. After obtaining the plurality of attribute information for the target dimension information, the adaptation range information for the target dimension information may be determined according to the plurality of attribute information. For example, a union of the plurality of attribute information may be determined as adaptation range information; for another example, an attribute value of attribute information, of which the number of times of the same attribute information is greater than a threshold value, among the plurality of attribute information may be used as one attribute value in the adaptation range of the adaptation range information. Similarly, the adaptation range information corresponding to each dimension information of the candidate policy can be obtained. It is to be understood that a plurality of historical policy data may be obtained for each candidate policy, so that the adaptation range information of each dimension information for each candidate policy may be obtained.

S105, determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of the at least one dimension information of each candidate strategy.

Wherein the total score may be a score determined according to a score of the at least one dimension information of the candidate policy. For example, the scores of each dimension information may be directly added to obtain a total score.

In a possible implementation manner, determining the total score corresponding to each candidate policy corresponding to the target object may further include the following steps: determining weight information corresponding to each dimension information in at least one dimension information; and determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of each dimension information and the corresponding weight information.

The weight information may be a numerical value indicating a degree of influence of each dimension information on determining the treatment policy. The larger the weight information is, the higher the degree of influence of the dimension information corresponding to the weight information on the determination of the treatment policy is. Determining weight information corresponding to each piece of dimension information in at least one piece of dimension information, which may be to determine a corresponding target weight rule from a plurality of candidate weight rules according to an object type of a target object, where each candidate weight rule has weight information corresponding to the at least one piece of dimension information, and then taking the weight information corresponding to the at least one piece of dimension information that the target weight rule has as the weight information corresponding to each piece of dimension information in the at least one piece of dimension information of the target object, respectively. Wherein the object type may indicate that the target object is a business or a person, the influence degree of the same dimension information may be different when determining the treatment policy of objects of different object types, and further the weight information for the same dimension information of different object types may be different. For example, the dimensional information such as complaint information for a business is influenced to a greater degree by a business, and the dimensional information such as income information for an individual is influenced to a greater degree by an individual.

The total score may be a score obtained by multiplying the score of each dimension information by the weight information and then adding the product. For example, if the score for the dimension information a is 10, the weight information is 5, the score for the dimension information b is 1, the weight information is 3, and the score for the dimension information c is 10, and the weight information is 1, the total score can be obtained by 10 × 5+1 × 3+10 × 1, that is, the total score is 63.

For example, please refer to fig. 3, fig. 3 is a flowchart illustrating a process for obtaining a total score according to an embodiment of the present application. According to the embodiment of the application, a first matrix can be constructed based on the score of at least one piece of dimension information of each candidate strategy, each row of data of the first matrix can correspond to different dimension information, each column of data can correspond to different candidate strategies, and each numerical value in the first matrix is the score of the dimension information corresponding to the row of the candidate strategy corresponding to the column. As shown in 301 in fig. 3, each column of data of the first matrix corresponds to a policy a, a policy b, and a policy c, and each column of data corresponds to a dimension 1, a dimension 2, a dimension 3, a dimension 4, and a dimension 5. And constructing a second matrix based on the weight information for the at least one piece of dimension information, wherein the first matrix comprises 1 row of data and at least one column of data, each column of data can correspond to different pieces of dimension information, and each numerical value in the first matrix is the weight information of the dimension information corresponding to the column. As shown at 302 in fig. 3, each column of data of the first matrix corresponds to dimension 1, dimension 2, dimension 3, dimension 4, and dimension 5, respectively. And further performing matrix multiplication operation on the first matrix and the second matrix to obtain a target matrix, wherein the target matrix has 1 row of data and multiple columns of data, each column of data can correspond to different candidate strategies, and each numerical value in the target matrix is the total score of the candidate strategies corresponding to the column in which the numerical value is located. As shown by 303 in fig. 3, each column of data of the target matrix corresponds to a policy a, a policy b, and a policy c, that is, the total score corresponding to the policy a, the policy b, and the policy c can be obtained.

And S106, determining a target handling policy aiming at the handling policy request from the plurality of candidate policies according to the total score corresponding to each candidate policy.

Wherein the target handling policy may be a policy for handling a target resource of a target object.

In one possible implementation, the target handling policy is determined from the plurality of candidate policies according to the total score corresponding to each candidate policy, and the candidate policy with the largest total score in the plurality of candidate policies may be the target handling policy.

In a possible implementation, a plurality of target handling policies may also be determined from the plurality of candidate policies to provide the user with a plurality of target handling policies for reference. For example, the top K candidate policies ranked in total score are determined as the target handling policies.

In one possible implementation, after obtaining the target handling policy for the target object, the target handling policy may also be sent to the client, so that the client displays the target handling policy for the target resource for handling the target object. Optionally, the processing method and the processing device for the target processing policy may further obtain at least one processing procedure record of the historical policy data with a good processing effect corresponding to the target processing policy, and send the processing procedure record to the client, so that the client displays the processing procedure record for the user to refer to.

Referring to fig. 4, fig. 4 is a schematic flowchart of a method for determining a knowledge-graph-based policy according to an embodiment of the present application. The method may be performed by the above mentioned electronic device. The method may include the following steps.

S401, detecting resource transaction information of the target object, and determining that the resource corresponding to the resource transaction information is bad resource when detecting that the resource transaction information of the target object meets a preset condition.

The resource transaction information may be information when a resource transaction is performed with a target object. For example, if the resource transaction with the target object is to borrow a money from the target object, the resource transaction information may include information of a money amount to be borrowed, a time to be borrowed, an agreed payment time, an amount not yet paid, an amount paid, an overdue time, and the like, which is not limited herein.

Bad resources are bad resources for bad accounts in accounting subjects. In some scenarios, the undesirable resource may be referred to as an undesirable asset. For example, if the target object does not return the original information by amount in time after the loan is made, the overdue and unpaid resources of the target object may be called bad resources. Optionally, the unpaid resources corresponding to other loan businesses of the target object may be determined as bad resources.

The preset condition may be a condition for determining the resource as a bad resource. For example, if the preset condition is that the expiry time included in the resource transaction information is greater than a threshold, the unpaid resource of the target object may be determined as a bad resource.

S402, generating a disposal strategy request aiming at the bad resources of the target object.

The handling policy request for the bad resources of the target object may be a handling policy request generated according to an object identifier of the target object with the bad resources, that is, the handling policy request may carry the object identifier of the target object. So as to determine the object data of the target object according to the object identification.

S403, when the treatment strategy request aiming at the bad resource of the target object is detected, object data of the target object is obtained.

S404, determining at least one dimension information for the treatment strategy request, and performing knowledge extraction based on the at least one dimension information and the object data to obtain an object data knowledge graph for the target object.

S405, obtaining a plurality of candidate strategies.

S406, determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, and determining the score of the target object for each piece of dimension information of each candidate strategy according to the attribute information of each piece of dimension information of the target object and the adaptive range information corresponding to each piece of dimension information of each candidate strategy.

S407, determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of the at least one dimension information of each candidate strategy.

And S408, determining a target handling policy for the handling policy request from the plurality of candidate policies according to the total score corresponding to each candidate policy.

The relevant description of steps S403 to S408 may refer to steps S101 to S106, which is not described herein. This determines a target handling policy for the handling policy request, i.e. a handling policy for bad resources of the target object.

In a possible implementation manner, when a target disposal resource is used to dispose a bad resource, as the disposal process of the bad resource advances, the situation of the target object may change, and the target data of the target object may be re-acquired at intervals, so as to update the target data knowledge graph of the target object, thereby re-determining the disposal strategy for the bad resource of the target object, further improving the disposal effect for the bad resource, and avoiding that the bad resource of the target object cannot be disposed well due to the fact that the movement of the target object is not grasped in time.

In the embodiment of the application, the resource transaction information of the target object can be detected, when the resource transaction information of the target object is detected to meet the preset condition, the resource corresponding to the resource transaction information is determined to be the bad resource, a disposal policy request for the bad resource of the target object is generated, and when the disposal policy request for the bad resource of the target object is detected, the object data of the target object is obtained; determining at least one piece of dimension information for the treatment strategy request, extracting knowledge based on the at least one piece of dimension information and the object data to obtain an object data knowledge graph for the target object, determining attribute information corresponding to each piece of dimension information according to the object data knowledge graph, determining a score of the target object for each piece of dimension information of each candidate strategy according to the attribute information of each piece of dimension information of the target object and adaptation range information corresponding to each piece of dimension information of each candidate strategy, and determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of the at least one piece of dimension information of each candidate strategy; a target handling policy for the handling policy request is determined from the plurality of candidate policies according to the total score corresponding to each candidate policy. Therefore, multi-dimensional attribute information of the target object can be extracted based on the knowledge graph technology, the target object is analyzed in a multi-dimensional mode, the target disposal strategy is determined, and accuracy of strategy determination is improved.

Referring to fig. 5, fig. 5 is a schematic structural diagram of a knowledge-graph-based policy determining apparatus according to an embodiment of the present application. Alternatively, the knowledge-graph-based policy determination apparatus may be disposed in the electronic device. As shown in fig. 5, the knowledge-graph-based policy determination apparatus described in this embodiment may include:

an obtaining unit 501, configured to obtain object data of a target object when a handling policy request for a target resource of the target object is detected;

a processing unit 502, configured to determine at least one dimension information for the treatment policy request, and perform knowledge extraction based on the at least one dimension information and the object data, to obtain an object data knowledge graph for the target object; the at least one dimension information is information of a reference dimension when determining a treatment policy for the target object, and each target relationship in the object data knowledge graph corresponds to the at least one dimension information respectively;

the obtaining unit 501 is further configured to obtain a plurality of candidate policies, where each candidate policy is associated with adaptation range information for the at least one dimension information;

the processing unit 502 is further configured to determine, according to the object data knowledge graph, attribute information corresponding to each piece of dimension information, and determine, according to the attribute information of each piece of dimension information of the target object and adaptation range information corresponding to each piece of dimension information of each candidate policy, a score of each piece of dimension information of each candidate policy for the target object;

the processing unit 502 is further configured to determine, according to the score of the at least one piece of dimension information of each candidate policy, a total score corresponding to each candidate policy corresponding to the target object;

the processing unit 502 is further configured to determine a target handling policy for the handling policy request from the plurality of candidate policies according to the total score corresponding to each candidate policy.

In an implementation manner, the processing unit 502 is specifically configured to:

determining a plurality of entity data in the object data;

determining a corresponding target relationship between the entity data according to the entity data and the object data;

forming at least one ternary group of data according to the entity data with the target relation and the corresponding target relation;

constructing an object data knowledge graph for the target object from the at least one triple data set.

In one implementation, the processing unit 502 is further configured to:

determining a relationship set used for constructing an object data knowledge graph, wherein at least one target relationship in the relationship set corresponds to the at least one dimension information;

obtaining sample data, wherein the sample data is provided with a sample label, the sample label is used for indicating an actual relation among entity data in the sample data, and the actual relation belongs to the relation set;

inputting the sample data into an initial classification model to obtain a prediction relation aiming at entity data in the sample data, wherein the prediction relation belongs to the relation set;

correcting the model parameters of the initial classification model according to the actual relation between the prediction relation and the entity data to obtain a target classification model;

the determining the corresponding target relationship between the entity data according to the entity data and the object data includes:

and inputting the entity data and the object data into the target classification model to obtain corresponding target relations among the entity data.

In an implementation manner, the at least one piece of dimension information includes target dimension information, the attribute information corresponding to each piece of dimension information includes an attribute value corresponding to each piece of dimension information, and the adaptation range information of each piece of dimension information includes an adaptation range of each piece of dimension information; the processing unit 502 is specifically configured to:

when the adaptation range corresponding to the adaptation range information of the target dimension information in the candidate strategy is detected, and the adaptation range corresponding to the attribute information of the target dimension information in the dimension information of the target object is included, determining the score of the target dimension information in the candidate strategy as a first score;

when the adaptation range corresponding to the adaptation range information of the target dimension information in the candidate strategy is detected and the attribute value corresponding to the attribute information of the target dimension information in the dimension information of the target object is not included, determining the score of the target dimension information in the candidate strategy to be a second score, wherein the first score is larger than the second score.

In one implementation, the processing unit 502 is further configured to:

obtaining a plurality of historical policy data, each historical policy data including historical object data of a historical object and a historical disposal policy, the plurality of historical policy data including the same historical disposal policy, the historical disposal policy corresponding to any one of the candidate policies;

respectively determining attribute information aiming at the target dimension information according to the object data of each historical object to obtain a plurality of attribute information aiming at the target dimension information;

determining adaptation range information of the target dimension information in the candidate strategies corresponding to the historical treatment strategies according to the attribute information.

determining weight information corresponding to each dimension information in the at least one dimension information;

and determining a total score corresponding to each candidate strategy corresponding to the target object according to the score of each dimension information and the corresponding weight information.

In an implementation manner, the target resource is a bad resource, and the processing unit 502 is further configured to:

detecting resource transaction information of the target object, and when the resource transaction information of the target object is detected to meet a preset condition, determining that a resource corresponding to the resource transaction information is a bad resource;

generating a handling policy request for bad resources of the target object.

Referring to fig. 6, fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. The electronic device described in this embodiment includes: a processor 601, a memory 602. Optionally, the electronic device may further include a network interface 603 or a power supply module. The processor 601, the memory 602, and the network interface 603 may exchange data therebetween.

The Processor 601 may be a Central Processing Unit (CPU), and may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The network interface 603 may include an input device such as a control panel, a microphone, a receiver, etc., and/or an output device such as a display screen, a transmitter, etc., to name but a few. For example, in an application embodiment, the network interface may include a receiver and a transmitter.

The memory 602 may include both read-only memory and random access memory, and provides program instructions and data to the processor 601. A portion of the memory 602 may also include non-volatile random access memory. Wherein, the processor 601, when calling the program instruction, is configured to perform:

In one implementation, the processor 601 is specifically configured to:

determining a plurality of entity data in the object data;

In one implementation, the processor 601 is further configured to:

In an implementation manner, the at least one piece of dimension information includes target dimension information, the attribute information corresponding to each piece of dimension information includes an attribute value corresponding to each piece of dimension information, and the adaptation range information of each piece of dimension information includes an adaptation range of each piece of dimension information; the processor 601 is specifically configured to:

In one implementation, the processor 601 is further configured to:

In one implementation, the processor 601 is specifically configured to:

In one implementation, the target resource is a bad resource, and the processor 601 is further configured to:

generating a handling policy request for bad resources of the target object.

Optionally, the program instructions may also implement other steps of the method in the above embodiments when executed by the processor, and details are not described here.

The present application further provides a computer-readable storage medium, in which a computer program is stored, the computer program comprising program instructions, which, when executed by a processor, cause the processor to perform the above method, such as performing the above method performed by an electronic device, which is not described herein in detail.

Optionally, the storage medium, such as a computer-readable storage medium, referred to herein may be non-volatile or volatile.

Alternatively, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to the use of the blockchain node, and the like. The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism and an encryption algorithm. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block includes information of a batch of network transactions for verifying the validity (anti-counterfeiting) of the information and generating a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

It should be noted that, for simplicity of description, the above-mentioned embodiments of the method are described as a series of acts or combinations, but those skilled in the art should understand that the present application is not limited by the order of acts described, as some steps may be performed in other orders or simultaneously according to the present application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.

Embodiments of the present application also provide a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. The computer instructions are read by a processor of a computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the steps performed in the embodiments of the methods described above. For example, the computer device may be a terminal, or may be a server.

The method, the device, the electronic device and the storage medium for determining the strategy based on the knowledge graph provided by the embodiment of the application are introduced in detail, a specific example is applied in the description to explain the principle and the implementation of the application, and the description of the embodiment is only used for helping to understand the method and the core idea of the application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A strategy determination method based on knowledge graph is characterized by comprising the following steps:

2. The method of claim 1, wherein the extracting knowledge based on the at least one dimension information and the object data to obtain an object data knowledge graph for the target object comprises:

determining a plurality of entity data in the object data;

3. The method of claim 2, further comprising:

4. The method according to claim 1, wherein the at least one dimension information includes target dimension information, the attribute information corresponding to each dimension information includes an attribute value corresponding to each dimension information, and the adaptation range information of each dimension information includes an adaptation range of each dimension information;

the determining, according to the attribute information of each piece of dimensional information of the target object and the adaptation range information corresponding to each piece of dimensional information of each candidate policy, a score of the target object for each piece of dimensional information of each candidate policy includes:

5. The method of claim 4, further comprising:

6. The method of claim 1, wherein the determining a total score value corresponding to each candidate policy corresponding to the target object according to the score value of the at least one dimension information of each candidate policy comprises:

7. The method of claim 1, wherein the target resource is a bad resource, and wherein the method further comprises:

generating a handling policy request for bad resources of the target object.

8. A knowledge-graph based policy determination apparatus, comprising:

9. An electronic device comprising a processor, a memory, wherein the memory is configured to store a computer program comprising program instructions, and wherein the processor is configured to invoke the program instructions to perform the method of any of claims 1-7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program comprising program instructions that, when executed by a processor, cause the processor to carry out the method according to any one of claims 1-7.