CN114565457A

CN114565457A - Risk data identification method and device, storage medium and electronic equipment

Info

Publication number: CN114565457A
Application number: CN202210036046.3A
Authority: CN
Inventors: 郭芷秀; 刘润佳
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2022-01-11
Filing date: 2022-01-11
Publication date: 2022-05-31

Abstract

The application discloses a risk data identification method and device, a storage medium and electronic equipment. Relating to the field of financial science and technology, the method comprises the following steps: acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; and determining the risk level of the target object according to the indexes and the corresponding target weights. Through the method and the device, the problem that the accuracy rate of risk identification in the related technology is poor by adopting a single empowerment method is solved.

Description

Risk data identification method and device, storage medium and electronic equipment

Technical Field

The application relates to the technical field of financial science and technology, in particular to a risk data identification method and device, a storage medium and electronic equipment.

Background

The novel technical means that the financial product innovation process is accelerated and is continuously emerged provides richer and more convenient financial services for customers, and meanwhile derives more complex illegal fund transfer risks which are difficult to identify. In addition, the continuous promotion of the international progress of the RMB, the cross-border fund transaction frequency and the absolute value are continuously increased, and the illegal fund transfer, terrorist financing and other crimes are more diversified and concealed through a technical chain transmission mode, so that new impact is brought to the illegal fund transfer risk prevention and control of the traditional commercial bank. Along with the situation of anti-illegal fund transfer supervision in the world and the country becoming stricter, the compliance pressure of financial institutions continuously increases, and the illegal fund transfer risk needs to be effectively identified and coped with by more advanced technological means so as to realize supervision compliance.

Because the starting time of the illegal fund transfer risk management is late, relatively unified standards are lacked in the industry, the requirements of financial institutions on data confidentiality are high, the management of the illegal fund transfer risk is more dependent on internal models and management methods of the financial institutions, and the management of the illegal fund transfer risk mainly has the following three problems at present:

1. the customer identification aspect: customer identification is the first work performed by anti-illegal fund transfer personnel, but as financial institutions wish to continuously improve availability and convenience of financial services, the collection of customer identification information is not highly valued and is relatively comprehensive. Particularly for the unnatural customer, the identification process includes identification of all persons who finally benefit, but because the stock right structure of part of the unnatural customer is complex, identification difficulty exists or identification has misoperation risk when all persons who finally benefit from manual identification.

2. And (3) aspects of due diligence survey process: due diligence can supplement the missing of the customer identity information to a certain extent, but due to the fact that the customer identity background and the transaction background have large differences, due to the fact that the network point due diligence personnel may have the problem of low specialty, the problem of invalid due diligence may be caused. Currently, commercial banks try to prompt due diligence points by issuing due diligence guidance, but a standardized due diligence process is adopted for non-standardized customers, so that the effectiveness of the due diligence is low.

3. Application aspect of risk characteristic identification achievement: since the illegal fund transfer risk management is always a concept of post management, the financial institution identifies the customers and the transactions after the fact, and the application of the risk level has certain limitation, so that the subjective initiative of the financial institution on the illegal fund transfer risk management is low. The reason for this is that, on one hand, there is a certain periodicity and a certain time lag for identifying the risk of illegal fund transfer, and on the other hand, the cost of active management is high due to the lack of the characteristic view of the risk of illegal fund transfer of the whole financial institution.

Aiming at the problem of poor accuracy rate of a single weighting method adopted by risk identification in the related technology, an effective solution is not provided at present.

Disclosure of Invention

The application mainly aims to provide a risk data identification method, a risk data identification device, a storage medium and electronic equipment, so as to solve the problem that the accuracy of risk identification in the related technology is poor by adopting a single empowerment method.

In order to achieve the above object, according to one aspect of the present application, there is provided a risk data identification method including: acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes respectively according to the first weight and the entropy weight; and determining the risk level of the target object according to the plurality of indexes and the corresponding target weights.

Optionally, determining the entropy weight of the first weight sum of the plurality of indexes comprises: determining a first weight of a plurality of metrics; determining entropy weights of a plurality of indexes; determining a target weight for the metric as a function of the first weight and the entropy weight comprises: and determining the target weight through a first weight and a corresponding first preset proportion, and an entropy weight and a corresponding second preset proportion, wherein the sum of the first preset proportion and the second preset proportion is one.

Optionally, determining the first weight of the plurality of metrics includes: generating a judgment matrix by comparing the importance of a plurality of indexes of a low level relative to the indexes of a high level and combining a preset scale, wherein each element of the judgment matrix represents the importance degree of a first index and a second index relative to the indexes of the high level, and the plurality of indexes of the low level belong to the next level of the indexes of the high level; determining the relative weight of a plurality of indexes according to the maximum eigenvalue of the judgment matrix and the corresponding eigenvector; generating an index matrix according to the relative weights of a plurality of indexes and the original data of the plurality of indexes, wherein each element of the index matrix represents the product of one index and one relative weight; a first weight of a plurality of metrics is determined from the metric matrix.

Optionally, after determining the relative weights of the multiple indexes by using the maximum eigenvalue of the judgment matrix and the corresponding eigenvector, the method further includes: determining a consistency index of the judgment matrix; determining a consistency ratio index of the judgment matrix according to the consistency index; determining that the consistency of the judgment matrix is qualified under the condition that the consistency ratio index is in a preset value range, and determining the relative weight of a plurality of indexes through the maximum eigenvalue of the judgment matrix and the corresponding eigenvector; and under the condition that the consistency ratio index is not in the preset value range, determining that the consistency of the judgment matrix is unqualified, and modifying the elements of the judgment matrix until the consistency of the judgment matrix is qualified.

Optionally, the determining the first weight of the plurality of indexes through the index matrix includes: carrying out non-dimensionalization on the original data in the index matrix; determining an optimal value in a row/column corresponding to each index in an index matrix through the non-dimensionalized index matrix, wherein the optimal value is a maximum value or a minimum value, the optimal value is the maximum value under the condition that the type of the index is a profitability index, and the optimal value is the minimum value under the condition that the type of the index is a cost index; and determining a first weight of the index according to the optimal value and the relative weight corresponding to the index.

Optionally, determining the entropy weights of the multiple indexes includes: carrying out dimensionless processing on the original data of the index; calculating an entropy value and a difference coefficient of the index according to the processed original data of the index; and determining the entropy weight of the index according to the entropy value and the difference coefficient.

Optionally, the target object includes: customer, business, territory.

Optionally, after determining the risk level of the target object according to the multiple indicators and the corresponding target weights, the method further includes: updating risk data of the risk of the client according to the risk grade; identifying the risk of the client according to the updated risk data; the risk data of the client risk comprise internal data and external data, wherein the internal data are data of a client account inside the system, and the external data are client data outside the system.

Optionally, after determining the risk level of the target object according to the multiple indexes and the corresponding target weights, the method further includes: updating risk data of business risk according to the risk grade; training a business risk identification model according to the updated risk data, wherein the business risk identification model is a semi-supervised fuzzy clustering algorithm (FCM) model; wherein the risk data of business risk includes business risk data and customer data related to the customer.

Optionally, after determining the risk level of the target object according to the multiple indicators and the corresponding target weights, the method further includes: updating risk data of the region risk according to the risk grade; generating region risk characteristics according to the updated risk data; the risk data of the regional risk comprise regional system data, service data corresponding to regions and related to services and customer data related to customers.

In order to achieve the above object, according to another aspect of the present application, there is provided a risk data identifying apparatus including: the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module acquires risk data of a target object in multiple dimensions, and the risk data comprises multiple indexes of different levels in each dimension; a first confirmation module that determines an entropy weight of a first weight sum of a plurality of indicators; the second determining module is used for determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; and the third determining module is used for determining the risk level of the target object according to the plurality of indexes and the corresponding target weights.

In order to achieve the above object, according to another aspect of the present application, there is provided a computer-readable storage medium storing a program, wherein the program executes to perform the risk data identification method of any one of the above.

To achieve the above object, according to another aspect of the present application, there is provided an electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the risk data identification method of any one of the above.

By the application, the following steps are adopted: acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; the risk grade of the target object is determined according to the multiple indexes and the corresponding target weights, the first weight and the entropy weight are determined by the multiple dimensions of the target object and the multiple indexes of multiple levels, the target weights of the multiple indexes are obtained comprehensively, the risk grade of the target object is further determined, and the problem that accuracy of risk identification in the related technology is poor due to the fact that a single weighting method is adopted is solved. And then the effect of improving the high efficiency and accuracy of risk identification is achieved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application. In the drawings:

fig. 1 is a flowchart of a risk data identification method provided according to an embodiment of the present application;

FIG. 2 is a system flow diagram of a method for identifying and visualizing a risk characteristic of a customer-centric illegal funds transfer according to an embodiment of the present application;

FIG. 3 is a flow chart of a combined evaluation method of an AHP method and an entropy weight method according to an embodiment of the present application;

FIG. 4 is a flow chart of a deep neural network based integrated analysis system provided in accordance with an embodiment of the present application;

fig. 5 is a flow chart of external public opinion information processing provided according to an embodiment of the present application;

fig. 6 is a flow chart for sample screening provided in accordance with an embodiment of the present application;

FIG. 7 is a flow chart of a comprehensive risk level calculation provided according to an embodiment of the present application;

FIG. 8 is a schematic diagram of a risk data identification device provided in accordance with an embodiment of the present application;

fig. 9 is a schematic diagram of an electronic device provided according to an embodiment of the application.

Detailed Description

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.

In order to make the technical solutions better understood by those skilled in the art, the technical solutions 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 only partial embodiments of the present application, but not all embodiments. 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.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It should be understood that the data so used may be interchanged under appropriate circumstances such that embodiments of the application described herein may be used. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

The present invention is described below with reference to preferred implementation steps, and fig. 1 is a flowchart of a risk data identification method provided according to an embodiment of the present application, and as shown in fig. 1, the method includes the following steps:

step S101, acquiring risk data of a target object in multiple dimensions, wherein the risk data comprises multiple indexes of different levels in each dimension;

step S102, determining a first weight sum entropy weight of a plurality of indexes;

step S103, determining target weights corresponding to a plurality of indexes according to the first weight and the entropy weight;

and step S104, determining the risk level of the target object according to the indexes and the corresponding target weights.

The executing body of the steps can be a processor or a controller, and through the steps, risk data of the target object in multiple dimensions are obtained, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; according to the multiple indexes and the corresponding target weights, the risk level of the target object is determined, the first weight and the entropy weight are determined by utilizing the multiple dimensions of the target object and the multiple indexes of multiple levels, the target weights of the multiple indexes are obtained comprehensively, the risk level of the target object is further determined, and the problem that accuracy of risk identification in the related technology is poor due to the fact that a single weighting method is adopted is solved. And then the effect of improving the high efficiency and accuracy of risk identification is achieved.

For example, the target object may be a customer risk, a business risk, or a regional risk, in order to ensure accurate assessment of the risk level of the target object, a plurality of evaluation dimensions may be divided in the target object in the prior art, and a plurality of indexes of different levels may be set in each dimension, the plurality of evaluation dimensions may be used as a first-level evaluation index, a lower-level index may be set under the first-level evaluation index, such as a second-level index, a third-level index, and a plurality of index levels of multiple levels, and each level index may include a plurality of indexes, for example, fig. 2 is a system flowchart of an illegal fund transfer risk feature identification and visualization method centering on a customer according to the embodiment of the present application, a customer risk, a business risk, or a regional risk, The evaluation indexes and the sub-items of the evaluation indexes can be set in the regional risk dimension, and it should be noted that the indexes of the plurality of different levels may include the indexes in fig. 2, but are not limited to the number of the indexes in fig. 2, and the index level.

The first weight of the plurality of indexes may be a relative weight of a plurality of indexes of a lower hierarchy except the first hierarchy of risk data of different dimensions with respect to an upper hierarchy of the lower hierarchy, and may be compared by importance of the plurality of indexes of the lower hierarchy with respect to an index of a higher hierarchy, wherein the degree of importance may be subjectively divided by a preset scale, an index weight determination matrix may be generated by the importance and the preset scale, and the first weight of the plurality of indexes may be determined by an operation of the index matrix. In this embodiment, the first weight may be a subjective weight determined by an analytic hierarchy process AHP.

The entropy weight can be another objective evaluation method of the multiple indexes, the influence on the index weight caused by unit or order difference among the multiple indexes can be eliminated by carrying out non-dimensionalization processing on the original data of the indexes, the entropy value of each index can be calculated by converting the original data matrix into a non-dimensionalized index matrix, the difference coefficient can be determined through the entropy value, and the entropy weights of the multiple indexes can be determined according to the entropy value and the difference coefficient.

The target weight may be a first weight of the plurality of indicators and a comprehensive weight of the plurality of indicators determined by the entropy weight, and the target weight of any one of the plurality of indicators is determined comprehensively by setting a first preset proportion of the first weight and a second preset proportion of the entropy weight, where the sum of the first preset proportion and the second preset proportion is one, that is, the first weight and the entropy weight respectively occupy different preset proportions in the target weight.

It should be noted that, by determining the target weights of the multiple indexes through the first weight and the entropy weight, the technical problems that a risk evaluation system adopted in the prior art is complex and a single weighting method is poor in accuracy rate can be effectively solved.

The risk level may be a risk level of the target object, and may be configured to visually present a risk condition of the target object, and the risk level may be determined by target weights corresponding to the plurality of indexes and the plurality of indexes, and the risk data of the target object may be updated by the risk level determination method, so as to perform risk identification on the target object.

The first weight and the entropy weight are determined by using multiple dimensions of the target object and multiple indexes of multiple levels, the target weights of the multiple indexes are obtained comprehensively, the risk level of the target object is further determined, and the problem that accuracy of risk identification in the related technology is poor by adopting a single weighting method is solved. And then the effect of improving the high efficiency and accuracy of risk identification is achieved.

Optionally, determining the entropy weight of the first weight sum of the plurality of indexes comprises: determining a first weight of a plurality of metrics; determining entropy weights of a plurality of indexes; determining a target weight of the indicator based on the first weight and the entropy weight comprises: and determining the target weight through the first weight and a corresponding first preset proportion, and the entropy weight and a corresponding second preset proportion, wherein the sum of the first preset proportion and the second preset proportion is one.

Determining a first weight of the plurality of indexes, that is, determining a weight of relative importance of the plurality of indexes to a level above the plurality of indexes, wherein the importance can be divided into importance degrees through a preset scale. The entropy weights for determining the indicators, that is, the weights for determining the objective importance of the indicators to the previous level thereof, may be calculated from the risk raw data by using entropy values and difference coefficients.

The target weight may be a composite weight of any one of the plurality of indicators, and may be determined by a predetermined ratio of the first weight and the second weight in the target weight, where the first weight corresponds to a first predetermined ratio in the target weight, the entropy weight corresponds to a second predetermined ratio in the target weight, and the target weight may be determined by the first predetermined ratio and the second predetermined ratio, and a sum of the first predetermined ratio and the second predetermined ratio is one. For example, the optimal combination weight result (i.e., the target weight) determined by the minimum entropy principle is 50% of the AHP subjective weight (i.e., the first weight) and the entropy weight (i.e., the entropy weight), and the combination weight w is represented as: w is 0.5w_i+0.5w_jWherein w represents a combination weight, wi represents a subjective weight of an AHP method, and wj represents an objective weight of an entropy weight method; the final weight calculation results from the weights are shown in table 1.

The target weight of the index is comprehensively determined through the first weight and the second weight, the technical problem that the wind direction evaluation form of the index is single and inaccurate in the prior art is solved, and the technical effect of improving the accuracy of the risk evaluation result is achieved.

Optionally, determining the first weight of the plurality of metrics includes: generating a judgment matrix by comparing the importance of a plurality of indexes of a low level relative to the indexes of a high level and combining a preset scale, wherein each element of the judgment matrix represents the importance degree of a first index and a second index relative to the indexes of the high level, and the plurality of indexes of the low level belong to the next level of the indexes of the high level; determining the relative weight of a plurality of indexes by judging the maximum eigenvalue of the matrix and the corresponding eigenvector; generating an index matrix according to the relative weights of the indexes and the original data of the indexes, wherein each element of the index matrix represents the product of one index and one relative weight; a first weight of a plurality of metrics is determined from a metric matrix.

The first weight of the multiple indexes can be compared with the importance of the indexes of the previous level of the multiple indexes, that is, the importance of the multiple indexes of the low level with respect to the indexes of the high level is compared, the indexes of the low level can be the indexes of the next level of the high level, the importance degree between any two indexes with respect to the previous level can be identified through a preset scale, for example, the importance degrees of the first index and the second index can introduce a 1-9 scale method, and further a judgment matrix of the weight can be generated, and the maximum eigenvalue of the judgment matrix and the eigenvector corresponding to the maximum eigenvalue can be calculated through mathematical operation through the judgment matrix.

The relative weights of the indexes can be the index weights of the indexes relative to the indexes of the upper layer, the relative weights of the indexes can be determined by the maximum eigenvalue of the judgment matrix and the corresponding eigenvector, and an index matrix can be generated according to the relative weights of the indexes and the original data of the indexes, wherein each element of the index matrix can represent the product of one index and one relative weight; a first weight of a plurality of metrics may be determined from the metric matrix.

Optionally, after determining the relative weights of the multiple indexes by determining the maximum eigenvalue of the matrix and the corresponding eigenvector, the method further includes: determining a consistency index of the judgment matrix; determining a consistency ratio index of the judgment matrix according to the consistency index; determining that the consistency of the judgment matrix is qualified under the condition that the consistency ratio index is in the preset value range, and determining the relative weight of a plurality of indexes through the maximum eigenvalue and the corresponding eigenvector of the judgment matrix; and under the condition that the consistency ratio index is not in the preset value range, determining that the consistency of the judgment matrix is unqualified, and modifying the elements of the judgment matrix until the consistency of the judgment matrix is qualified.

After the relative weights of a plurality of indexes are determined through the maximum characteristic value and the corresponding characteristic vector of the judgment matrix, consistency check can be carried out on the judgment matrix, whether the matrix consistency is qualified or not can be judged through the consistency index of the judgment matrix, wherein the consistency index can be set in a preset mode, and if the consistency ratio index of the judgment matrix is within a preset range deviating from the consistency in the judgment matrix, the element value in the judgment matrix can be modified.

In the case that the consistency ratio index is in the preset value range, the consistency of the judgment matrix can be determined to be qualified, and the step of determining the relative weight of the indexes through the maximum eigenvalue and the corresponding eigenvector of the judgment matrix is executed.

Optionally, the determining the first weight of the plurality of indexes through the index matrix includes: carrying out dimensionless transformation on the original data in the index matrix; determining an optimal value in a row/column corresponding to each index in the index matrix through the non-dimensionalized index matrix, wherein the optimal value is a maximum value or a minimum value, the optimal value is the maximum value under the condition that the type of the index is a profitability index, and the optimal value is the minimum value under the condition that the type of the index is a cost index; and determining a first weight of the index according to the optimal value and the relative weight corresponding to the index.

In the process of determining the first weights of the multiple indexes through the index matrix, in order to avoid the influence of the magnitude order or the number unit of the raw data of the multiple indexes on the calculation result of the first weights of the multiple indexes, the raw data in the indexes can be subjected to non-dimensionalization, the raw data can be subjected to non-dimensionalization by adopting an extremum method, and the optimal value in the row/column corresponding to each index in the index matrix can be determined through the non-dimensionalized index matrix, wherein the optimal value can be a maximum value or a minimum value, when the type of the index is a profitable index, the optimal value can be a maximum value, and when the type of the index is a cost index, the optimal value can be a minimum value. According to the optimal value and the relative weight corresponding to the index, the first weight of the index can be determined.

The entropy weight can be another objective evaluation method of the indexes, the influence on the index weight caused by different units or orders of magnitude among the indexes can be eliminated by carrying out non-dimensionalization processing on the original data of the indexes, the entropy value of each index can be calculated by converting the original data matrix into a new evaluation matrix after non-dimensionalization, and the difference coefficient can be determined through the entropy value, so that the entropy weights of the indexes can be determined.

Optionally, the target object includes: customer, business, territory.

The target object may include multiple aspects of risk for the financial asset, which may be customer risk, business risk, or regional risk.

Optionally, after determining the risk level of the target object according to the multiple indexes and the corresponding target weights, the method further includes: updating risk data of the risk of the client according to the risk level; identifying the risk of the client according to the updated risk data; the risk data of the client risk comprise internal data and external data, wherein the internal data are data of a client account inside the system, and the external data are client data outside the system.

The risk data may include internal data, such as customer portrait data in an organization system data warehouse, feature data in a core business system, anti-illegal fund transfer monitoring data, and external data, such as external public opinion information in a public data platform, administration penalty information, credit information, judicial actions, and the like.

After the risk level of the target object is determined according to the multiple indexes and the corresponding target weights, the risk data of the client risk can be updated according to the risk level, the identity risk condition of the client can be identified according to the updated risk data, and response measures for the financial asset safety protection can be made in time.

The external data set used for the identification of the client identity risk may be used as a technical means for weighting the plurality of indexes, and the change of the index weight may also affect the identification of the client identity risk, that is, may also affect the client risk level.

The risk grade of the target object is determined by utilizing the target weights respectively corresponding to the indexes, so that the risk of the client is identified, the problem that the risk of the client is easy to make mistakes in the prior art is solved, and the technical effect of improving the effectiveness of the identity risk identification of the client is achieved.

Optionally, after determining the risk level of the target object according to the multiple indexes and the corresponding target weights, the method further includes: updating risk data of business risk according to the risk grade; training a business risk identification model according to the updated risk data, wherein the business risk identification model is a semi-supervised fuzzy clustering algorithm (FCM) model; wherein the risk data of business risk comprises business risk data and customer data related to the customer.

In order to improve the automatic identification capability of illegal fund transfer in suspicious new products and complex transactions, after the risk level of a target object is determined according to a plurality of indexes and corresponding target weights respectively, the risk data of the business risk can be updated according to the risk level, the business risk identification model can be trained according to the updated business risk data and client data related to clients, namely, the business risk identification model, namely a semi-supervised fuzzy clustering algorithm (FCM) model can be trained through a semi-supervised fuzzy clustering algorithm (FCM), the suspicious transactions in the business can be monitored by utilizing the trained semi-supervised fuzzy clustering algorithm (FCM), suspicious reports can be automatically generated, and dynamic monitoring and early warning can be carried out on the suspicious transactions.

By training the business risk identification model by using the updated business risk data and the customer data related to the customer, the problems of low business risk identification efficiency and high error probability in the prior art are solved, and the technical effect of improving the safety of business transaction is realized.

After the customer risk and the business risk are determined according to the target weights of the indexes, the indexes can be updated according to the updated customer risk and the updated business risk, so that the target weights of the indexes are updated, and feedback and dynamic self-adjustment are formed.

Optionally, after determining the risk level of the target object according to the multiple indexes and the corresponding target weights, the method further includes: updating risk data of the region risk according to the risk grade; generating region risk characteristics according to the updated risk data; the risk data of the regional risk comprise regional system data, service data corresponding to regions and related to services and customer data related to customers.

The risk data of the regional risk may include regional system data, such as illegal fund transfer risk data of the financial institution, business data related to business corresponding to the region, or customer data related to customers.

After the risk level of the target object is determined according to the multiple indexes and the corresponding target weights, the risk data of the region risk can be updated according to the risk level, and the region risk characteristics can be generated according to the updated risk data, so that the risk of the system in the region can be reflected.

For example, the illegal fund transfer risk characteristics of all specific countries and regions can be obtained by using the data in the specific countries and regions and calculating by using the region risk dimension in the evaluation system established by the AHP-entropy weight method. The method combines external rating data and illegal fund transfer risk characteristic data of the financial institution, evaluates and monitors the illegal fund transfer risk of the regional dimension of the financial institution from the aspects of systematic risk and non-systematic risk, guides a due-time investigation post in a specific region to carry out targeted due-time investigation on the identity and transaction characteristics of a client, and identifies the hidden danger of the illegal fund transfer risk in advance.

By updating the region risk data by using the risk grade and generating the region risk characteristics, the technical problem that the risk characteristics of clients and businesses of the financial institution in various regions are neglected by the region risk identification and evaluation method in the prior art is solved, and the technical effect of expanding the region risk identification range is achieved.

It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer-executable instructions and that, although a logical order is illustrated in the flowcharts, in some cases, the steps illustrated or described may be performed in an order different than presented herein.

This embodiment also provides an alternative embodiment, which is described in detail below.

The present embodiment addresses the shortcomings and drawbacks of the prior art. A customer-centric illegal funds transfer risk characteristic identification and visualization method is provided.

Fig. 2 is a system flowchart of a customer-centric illegal fund transfer risk characteristic identification and visualization method according to an embodiment of the present application, and the overall system flowchart is as shown in fig. 2.

In order to solve the existing problems, the method is carried out from the following aspects:

visualization of illegal fund transfer risk characteristics based on an AHP-entropy weight method;

according to the implementation mode, the monitoring of the illegal fund transfer risk is divided into three dimensions of client, business and region according to the requirements of financing risk safety guide and the current situation of a financial institution. Respectively performing (1) risk distribution characteristics of a single dimension on the basis of evaluation of the single dimension by the current financial institution; (2) the illegal fund transfer risk characteristics of the financial institution are visualized from two angles of comprehensive dimension statistical analysis, and a basis is provided for the active management of the illegal fund transfer risk from the angle of comprehensive risk management.

1. Visualization of risk distribution characteristics of a single dimension;

(1) evaluating the dimensionality;

table 1 is a single-dimensional risk distribution characteristic visualization evaluation table, and the evaluation dimensions are shown in table 1:

TABLE 1 visual evaluation table for risk distribution characteristics of single dimension

Note: the financial institution can adjust the evaluation dimensions of various risks according to the financial institution;

(2) an evaluation method;

and taking the single risk dimension as a module, respectively carrying out access calculation on the sub-items of each evaluation index, and drawing a statistical chart.

2. Comprehensive dimension statistical analysis;

(1) evaluating the dimensionality;

based on the management principle that the risk is the main factor, the embodiment comprehensively analyzes the illegal fund transfer risk characteristics (A) related to high risk.

2. Comprehensive dimension statistical analysis;

(1) evaluating the dimension;

table 2 is a comprehensive dimension evaluation table, and as shown in table 2, the present embodiment comprehensively analyzes the illegal fund transfer risk characteristics (a) related to high risk, based on the management principle based on risk as the principal.

TABLE 2 comprehensive dimension evaluation Table

(2) an evaluation method;

according to the embodiment, the AHP method and the entropy weight method are combined to carry out combined weighting on all dimensionality indexes of the illegal fund transfer risk, the subjective weighting result of the AHP method and the objective weighting result of the entropy weight method are weighted and averaged, the weight of each index in the illegal fund transfer risk evaluation is determined, the limitation of adopting a single weighting method is avoided, the consistency of the measuring and calculating result and the actual situation is ensured, and the whole illegal fund transfer risk level of a financial institution is better reflected. The specific steps for determining the index weight are as follows:

firstly, an AHP (Analytic Hierarchy Process) method is used for obtaining the subjective weight of the index, wherein a target layer is the illegal fund transfer risk level of a financial institution, a criterion layer is a first-level index, and a scheme layer is 25 second-level indexes.

The indexes in the level are compared pairwise to determine the relative importance of the indexes relative to the indexes in the criterion layer. I.e. the index e of layer D, using the risk dimension B as a criterion₁，e₂，…，e_mEvaluation for criteria B, e_iAnd e_jWhich is more important. This decision is quantified by introducing a suitable scale and represented in matrix form, i.e. the decision matrix is constructed. Let the n-th order matrix denote that A ═ α_ij)_m×m，a_ijRelative to the index e_iAnd e_jIndex e_iRatio e_jTo a significant (or insignificant) degree. Table 3 shows the index e_iAnd e_jIn the paired comparison table, as shown in table 3, the indexes were quantitatively compared in pairs by using a 1-9 scale method. The matrix a obtained by this method has the following properties:

wherein, a_ijIs relative to the index e_iAnd e_jM, n are natural numbers, e₁，e₂，…，e_mThe matrix a satisfies the above formula as an index of the layer D, and it is known that the matrix a is a reciprocal matrix.

TABLE 3 index e_iAnd e_jPaired comparison table

Note: if e_iAnd e_jThe values of which differ by an extent between two of the above-mentioned adjacent levels are taken to be respectively 2, 4, 6, 8,

secondly, according to the judgment matrix A obtained in the last step, an index e is obtained₁，e₂，…，e_mThe index weight relative to which the index B is dominant. In calculating the weights, it is assumed that A has consistency, i.e. a_ij×a_jk＝a_ik. The maximum eigenvalue of the calculated determination matrix a and the eigenvector corresponding to the maximum eigenvalue satisfy a condition where a ω ═ λ_maxWherein ω is (ω)₁，ω₂，…，ω_m)；

Wherein, ω is a feature vector corresponding to the maximum feature value of the judgment matrix a, and is the maximum feature value of the matrix a, and m is a natural number.

Then (omega)₁，ω₂，…，ω_m) Is e₁，e₂，…，e_mRelative weight values for criterion B.

And thirdly, carrying out consistency check on the matrix A. The judgment matrix A given by expert evaluation in general does not satisfy the complete consistency, i.e. does not satisfy a completely_ij×a_jk＝a_ik. A consistency criterion may be given and when it is determined that matrix a meets the given criterion, a may be considered to meet consistency.

The consistency index CI of matrix a is defined to measure the consistency of a:

namely, the consistency degree of the judgment matrix A is checked according to the change of the maximum characteristic root of the judgment matrix A. The larger the CI is, the larger the deviation degree of the judgment matrix A from the consistency is; conversely, the smaller CI, the smaller the degree to which the judgment matrix A deviates from consistency.

For the judgment matrix with a larger order, the applicability of the consistency checking method is reduced, and the consistency index CI needs to be corrected. The average random consistency index RI is introduced here, and its calculation process is as follows:

i. for m-th order decision matrix, from

In the random value taking, repeating

Secondly, wherein m is a natural number, each time is independent, the value is used as an upper triangular element of a new matrix, a main diagonal element of the matrix is 1, a lower triangular element is the reciprocal of the upper triangular element, and a random forward and inverse matrix is constructed;

calculating a consistency index CI of the obtained matrix;

repeat the above steps to obtain enough samples, calculate the mean of the samples for CI, and take this mean as RI, Table 4 is a table of values associated with R1 and m, it is clear that RI is a value associated with m (as shown in Table 4).

Table 4 is a table of values associated with R1 and m

m	1	2	3	4	5	6	7	8	9	…
											RI	0.00	0.00	0.58	0.90	1.12	1.24	1.32	1.41	1.45	…

When m is 1, 2, RI is 0, i.e. the decision matrix always has complete consistency.

When m is more than 2, introducing a consistency ratio index CR, defining

And correcting the consistency criterion of the judgment matrix into:

when CR is less than 0.1, judging that the matrix A has acceptable consistency; on the contrary, the element value in the matrix a needs to be modified if the degree of deviation of the judgment matrix a from the consistency is too large.

After the index weight of a group of indexes relative to the upper layer of domination indexes is obtained, the index weights of all the indexes relative to the upper layer of domination indexes can be obtained according to the method. The index weights are weighted by layer according to the dominance relationship. And finally, obtaining the index weight of each index at the bottommost layer relative to the index at the highest layer. Suppose a k-1 th layer n_k-1The weight vector of each index relative to the highest index is

N on the k-th layer_kThe weight vector of each index to the jth index of the k-1 layer is

Wherein the weight of the index not dominated by the jth index is 0, the weight vector of the index on the kth layer relative to the index on the highest layer is beta_i：

Wherein β i is a weight vector of the k-th layer index relative to the highest layer index, k represents the number of layers, and n is a natural number.

By analogy, the index weight of the bottommost layer index relative to the highest layer, namely the total target, can be calculated from top to bottom.

And finally, evaluating the quality of each object to be evaluated by combining the obtained index weight value and the original data. Suppose that m evaluated objects exist in the system, and the index weight of n indexes at the bottommost layer relative to the total target is omega_jJ is 1, 2. The index value corresponding to the evaluated object can be represented in the following matrix form R:

because the data of each index value may not be in an order of magnitude, in order to facilitate comparison, the original data value needs to be subjected to dimensionless processing by adopting an extremization method.

Firstly, carrying out dimensionless processing on original data: let the optimal value of each column in R be R_j*：

Wherein, the profitability index means that the index value is better as the index value is larger, and the cost index means that the index value is better as the index value is smaller;

let S be (S)_ij)_m×nThen, then

Wherein S is a matrix formed by Sij defined by the S.

The relative evaluation value of the ith object is

According to x_iThe value of the evaluation target is judged to be good or bad. x is the number of_iThe larger the number, the more preferable the ith object is.

Secondly, determining the objective weight of the index by using an entropy weight method, wherein the calculation method comprises the following steps:

processing original data: performing non-dimensionalization on the index data by using a linear scaling method (same as above), and converting the original data matrix e into a new evaluation matrix which is non-dimensionalized and is S-S (S)_ij)_mn(m is the index number, n is the number of years of evaluation).

② calculating entropy value H of ith index_i：

Wherein the content of the first and second substances,

and (i is 1, 1.. multidot.m, and j is 1, 2.. multidot.n) represents the proportion of the ith index value in calculation, and ln m means that a natural logarithm is obtained on m.

Calculating the difference coefficient alpha of the i index_i：

a_j＝1-H_i，(j＝1，2，…，n)

Information entropy H of certain index_iDifference a between 1 and_idetermines the information utility value, alpha, of the index_iThe magnitude of the direct influence weight is larger, and the larger the information utility value is, the larger the importance of the index to the evaluation is, and the larger the weight is.

Fourthly, calculating the entropy weight omega of the ith index_i：

Finally, the optimal combination weight result determined by the minimum information entropy principle is that the subjective weight of the AHP method and the objective weight of the entropy weight method respectively account for 50 percent, and the combination weight w is expressed as:

w＝0.5w_i+0.5w_j

wherein w is the optimal combining weight, w_iIs subjective weight of AHP method, w_jIs an objective weight of entropy weight method.

Table 5 is an objective weight result table for determining the index by using the entropy weight method, and the final weight calculation result obtained according to the weight is shown in table 5:

TABLE 5 Objective weight result table for index determination using entropy weight method

Wherein the content of the first and second substances,

wherein x1, x2 …, xm are relative evaluation values of the mth object, and S is defined S_ijAnd omega is a natural number corresponding to the characteristic vector m of the maximum characteristic value of the judgment matrix A.

Table 6 is a table of calculation results of the index layer, and the calculation results of the index layer obtained according to the weights are shown in table 6:

TABLE 6 index layer calculation results Table

Customer risk/%)	Business risk/%)	Regional risk/%)	Illegal funds transfer risk level/%)
				Y₁	Y₂	Y₃	Y

Wherein the content of the first and second substances,

at the moment, the illegal fund transfer risk of each dimension of the financial structure and the overall illegal fund transfer risk level are visually processed, and the financial institution can periodically evaluate, so that the dynamic change of each risk dimension can be monitored from the global perspective, the time for managing the illegal fund transfer risk is advanced, and the subjective activity is improved to a certain extent.

Fig. 3 is a flowchart of a combined evaluation method of the AHP method and the entropy weight method according to an embodiment of the present application, and the flowchart of the above calculation process is shown in fig. 3.

(II) customer identification technology;

in order to automatically capture as much customer identity information as possible through the system, a customer identity information identification mode is further supplemented on the basis that the existing financial institution carries out integrity judgment on nine elements of individual customers and 24 elements of legal customers and integrity judgment on all beneficiaries.

In the acquisition and identification of personal customer information, the application range of the biological identification technology is expanded, such as face identification, voiceprint, fingerprint and the like, and the biological identification module is comprehensively incorporated into a bank non-face-to-face and core business system to control the higher degree of potential illegal fund transfer risk in non-face-to-face transaction.

In the identification of public customers and beneficiaries, a comprehensive analysis system based on a deep neural network is constructed by utilizing internal data of financial institutions and external data such as workers, credit investigation and customs, and penetrating identification of the finally beneficiaries is realized to deal with operation risks in the manual judgment process under a complex stock right structure, fig. 4 is a flow chart of the comprehensive analysis system based on the deep neural network provided according to the embodiment of the application, and the technical implementation steps are shown in fig. 4. The method comprises the following specific steps:

(1) extracting portrait data of a public client, characteristic data in a core business system and anti-illegal fund transfer monitoring data in a financial institution data warehouse to form an internal data set;

(2) extracting external public opinion data such as business information, administrative penalty information, credit information, judicial litigation and the like in an external public data platform to form an external data set;

(3) extracting internal and external standardized information to form basic information fields in the due diligence questionnaire;

(4) and carrying out data cleaning and semantic analysis on non-standardized information such as external public sentiment, identifying all people who are finally benefited by the enterprise, and providing further identification for the full-time investigation process.

Fig. 5 is a flow chart of external public opinion information processing provided according to an embodiment of the application, and a process is as shown in fig. 5, and by integrating internal data and external data, on one hand, a final beneficial owner under a complex stock right structure can be identified by combining an association relationship between beneficial owners, on the other hand, beneficial owner information which may be hidden can be identified according to the external public opinion information, and effectiveness of client identity identification is improved by accessing such information in a due diligence process.

Thirdly, monitoring the service risk based on semi-supervised FCM clustering;

in order to improve the automatic identification capability of the suspicious model on illegal fund transfer risks in new products and complex transactions, the scheme improves the monitoring capability on transaction risks by introducing a semi-supervised fuzzy clustering algorithm (FCM). The method is mainly characterized in that based on the existing suspicious monitoring model and manual screening results of a financial institution as label samples, massive label-free samples are trained, and the recognition capability of the existing model is improved.

First, in the FCM algorithm, it is assumed that the number of samples of the cluster data set X is n, that is, X is (X ═ X)₁，x₂，...，x_n) The cluster center is c_iThen the objective function can be expressed as:

wherein U is a membership matrix of the function; u. of_ik∈[0，1]A degree of attribution representing a corresponding category, wherein a larger value thereof indicates a higher probability of attribution to the cluster center; d_ik＝||x_k-c_iI represents the k-th data to the clustering center c_iThe Euclidean distance of; m ∈ [0, + ∞) is a weighted index that controls the degree of ambiguity of the objective function, typically m ═ 2.

Computing a targetWhen the function is carried out, if the objective function value is smaller than the threshold value or the iteration times are larger than the maximum times, the algorithm is terminated to obtain a clustering center c_i(ii) a And (5) recalculating the clustering centers by using the anti-regularization for iteration.

In that

Under the constraint condition of (3), the minimum value of the target function is obtained through Lagrange (Lagrange) number multiplication to obtain a membership matrix u_ikAnd a clustering center c_i：

Secondly, on the basis, the screened suspicious samples are used as prior information and are included into an objective function of the FCM algorithm to guide the clustering process, and the new objective function is as follows:

wherein f is_ijRepresenting x for membership matrix containing existing label information_jAttribution c_iDegree of (d); b_jIs a virtual variable, where x_jRepresenting a sample data set:

further, α is an adjustment coefficient for balancing unsupervised information and supervised information. Obtaining the minimum value of the objective function by Lagrange (Lagrange) number multiplication to obtain a membership matrix u_ikAnd a cluster center c_iThe iterative expression of (c) is:

wherein d is_ik＝||x_k-c_iI represents the k-th data to the clustering center c_iThe Euclidean distance of; f. of_ijIs a membership matrix containing information of existing tags.

When calculating the objective function, if the objective function value is less than the threshold value or the iteration times is more than the maximum times, terminating the algorithm to obtain the clustering center c_i(ii) a And (5) recalculating the clustering centers by using the anti-regularization for iteration.

Fig. 6 is a flow chart of sample screening provided according to an embodiment of the present application, and as shown in fig. 6, after an algorithm is completed, on one hand, a suspicious report may be automatically generated for a monitored suspicious transaction, and dynamic monitoring and early warning may be performed on the transaction; on the other hand, the method can realize the point pushing of the due-time investigation of the target client, so that network operators can carry out the due-time investigation on the fund source and the application of the client more pertinently, and the effectiveness of the due-time investigation is improved.

(IV) regional risk monitoring technology;

currently, most of the financial institutions classify regional risks directly by rating of external institutions, and the rating method evaluates national/regional risks based on macroscopic variables and reflects systematic risks in the regions. However, for the illegal fund transfer risk management of the financial institution, the classification method is directly adopted to ignore the risk characteristics of the clients and businesses of the financial institution in each region, namely, the non-systematic risk table 7 region risk characteristic evaluation table, as shown in table 7, in the embodiment, the illegal fund transfer risk characteristics of all specific countries and regions can be obtained by using the data in the specific countries/regions and calculating by using the region risk dimension in the evaluation system established by the AHP-entropy method in the foregoing.

TABLE 7 regional Risk characteristics evaluation Table

Fig. 7 is a flowchart of a comprehensive risk level calculation according to an embodiment of the present application, and as shown in fig. 7, the evaluation method combines external rating data and illegal fund transfer risk characteristic data of a financial institution itself, evaluates and monitors illegal fund transfer risk of a regional dimension of the financial institution from two aspects of systematic risk and non-systematic risk, guides a due diligence post in a specific region to perform targeted due diligence on a client identity and a transaction characteristic, and identifies an illegal fund transfer risk in advance.

The embodiment can realize the visualization of the overall illegal fund transfer risk characteristics of the financial institution for the whole financial institution, and improve the possibility of providing the active management process of the illegal fund transfer risk; from the branch institution, the identification achievement of each dimension risk characteristic can be applied, the full-time investigation key points are automatically pushed at the system level, the effectiveness of the full-time investigation process is improved, the integrity of the customer identity information storage and the identification accuracy are further improved, the function of the second defense line of the financial institution risk management module is fully exerted, and more effective support is provided for each business system.

The embodiment of the present application further provides a risk data identification device, and it should be noted that the risk data identification device of the embodiment of the present application may be used to execute the method for identifying risk data provided by the embodiment of the present application. The risk data identification device provided by the embodiment of the application is described below.

Fig. 8 is a schematic diagram of a risk data identification device according to an embodiment of the application. As shown in fig. 8, the apparatus includes: the acquisition module 82, the first determination module 84, the second determination module 86, and the third determination module 88, which will be described in detail below.

The acquiring module 82 is used for acquiring risk data of the target object in multiple dimensions, wherein the risk data comprises multiple indexes of different levels in each dimension; a first determining module 84, connected to the acquiring module, for determining a first weight and an entropy weight of the plurality of indicators; a second determining module 86, connected to the first determining module, for determining target weights corresponding to the plurality of indicators according to the first weight and the entropy weight; and a third determining module 88, connected to the second determining module, for determining the risk level of the target object according to the plurality of indicators and the corresponding target weights.

By the above device, the obtaining module 82 obtains risk data of the target object in multiple dimensions, where the risk data includes multiple indexes of different levels in each dimension; a first determination module 84 that determines an entropy weight of a first weight sum of the plurality of metrics; a second determining module 86, configured to determine target weights corresponding to the plurality of indicators according to the first weight and the entropy weight; the third determining module 88 determines the risk level of the target object according to the multiple indexes and the corresponding target weights, determines the first weight and the entropy weight by using the multiple dimensions of the target object and the multiple indexes of multiple levels, and obtains the target weights of the multiple indexes in a comprehensive manner, so as to determine the risk level of the target object, thereby solving the problem that the accuracy of risk identification using a single weighting method in the related art is poor. And then the effect of improving the high efficiency and accuracy of risk identification is achieved.

The risk data identification device comprises a processor and a memory, wherein the acquiring module 82, the first determining module 54, the second determining module 86, the third determining module 88 and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.

The processor comprises a kernel, and the kernel calls the corresponding program unit from the memory. The kernel can be set to be one or more, whether the target image has the target object or not is detected through the target object detection model by adjusting the kernel parameters, and in the case of having the target object, whether the target object is abnormal or not is recognized.

The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), including at least one memory chip.

An embodiment of the present invention provides a computer-readable storage medium on which a program is stored, which, when executed by a processor, implements the risk data identification method.

The embodiment of the invention provides a processor, which is used for running a program, wherein the risk data identification method is executed when the program runs.

As shown in fig. 9, an embodiment of the present invention provides an electronic device, where the electronic device 90 includes a processor, a memory, and a program stored in the memory and executable on the processor, and the processor executes the program to implement the following steps:

acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; and determining the risk level of the target object according to the indexes and the corresponding target weights.

Optionally, determining the entropy weight of the first weight sum of the plurality of indexes comprises: determining a first weight of a plurality of metrics; determining entropy weights of a plurality of indexes; determining a target weight for the metric based on the first weight and the entropy weight comprises: and determining the target weight through the first weight and a corresponding first preset proportion, and the entropy weight and a corresponding second preset proportion, wherein the sum of the first preset proportion and the second preset proportion is one.

Optionally, the determining the first weight of the multiple indexes through the index matrix includes: carrying out dimensionless transformation on the original data in the index matrix; determining an optimal value in a row/column corresponding to each index in an index matrix through the non-dimensionalized index matrix, wherein the optimal value is a maximum value or a minimum value, the optimal value is the maximum value under the condition that the type of the index is a profitability index, and the optimal value is the minimum value under the condition that the type of the index is a cost index; and determining a first weight of the index according to the optimal value and the relative weight corresponding to the index.

Optionally, the target object includes: customer, business, territory.

The device herein may be a server, a PC, a PAD, a mobile phone, etc.

The present application further provides a computer program product adapted to perform a program for initializing the following method steps when executed on a data processing device: acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension; determining an entropy weight of a first weight sum of a plurality of indexes; determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight; and determining the risk level of the target object according to the indexes and the corresponding target weights.

Optionally, determining the first weight of the plurality of indicators includes: generating a judgment matrix by comparing the importance of a plurality of indexes of a low level relative to the indexes of a high level and combining a preset scale, wherein each element of the judgment matrix represents the importance degree of a first index and a second index relative to the indexes of the high level, and the plurality of indexes of the low level belong to the next level of the indexes of the high level; determining the relative weight of a plurality of indexes by judging the maximum eigenvalue of the matrix and the corresponding eigenvector; generating an index matrix according to the relative weights of the indexes and the original data of the indexes, wherein each element of the index matrix represents the product of one index and one relative weight; a first weight of a plurality of metrics is determined from a metric matrix.

Optionally, the determining the first weight of the plurality of indexes through the index matrix includes: carrying out dimensionless transformation on the original data in the index matrix; determining an optimal value in a row/column corresponding to each index in an index matrix through the non-dimensionalized index matrix, wherein the optimal value is a maximum value or a minimum value, the optimal value is the maximum value under the condition that the type of the index is a profitability index, and the optimal value is the minimum value under the condition that the type of the index is a cost index; and determining a first weight of the index according to the optimal value and the relative weight corresponding to the index.

Optionally, the target object includes: customer, business, territory.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in the process, method, article, or apparatus that comprises the element.

The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for identifying risk data, comprising:

acquiring risk data of a target object in multiple dimensions, wherein the risk data comprise multiple indexes of different levels in each dimension;

determining an entropy weight of a first weight sum of a plurality of indexes;

determining target weights corresponding to the multiple indexes respectively according to the first weight and the entropy weight;

and determining the risk level of the target object according to the plurality of indexes and the corresponding target weights.

2. The method of claim 1, wherein determining the entropy weight of the first weighted sum of the plurality of metrics comprises:

determining a first weight of a plurality of metrics;

determining entropy weights of a plurality of indexes;

determining a target weight for the metric as a function of the first weight and the entropy weight comprises:

and determining the target weight through a first weight and a corresponding first preset proportion, and an entropy weight and a corresponding second preset proportion, wherein the sum of the first preset proportion and the second preset proportion is one.

3. The method of claim 2, wherein determining a first weight for a plurality of metrics comprises:

generating a judgment matrix by comparing the importance of a plurality of indexes of a low level relative to the indexes of a high level and combining a preset scale, wherein each element of the judgment matrix represents the importance degree of a first index and a second index relative to the indexes of the high level, and the plurality of indexes of the low level belong to the next level of the indexes of the high level;

determining the relative weight of a plurality of indexes according to the maximum eigenvalue of the judgment matrix and the corresponding eigenvector;

generating an index matrix according to the relative weights of a plurality of indexes and the original data of the plurality of indexes, wherein each element of the index matrix represents the product of one index and one relative weight;

a first weight of a plurality of metrics is determined from the metric matrix.

4. The method of claim 3, wherein after determining the relative weights of the plurality of metrics by the maximum eigenvalue of the decision matrix and the corresponding eigenvector, the method further comprises:

determining a consistency index of the judgment matrix;

determining a consistency ratio index of the judgment matrix according to the consistency index;

determining that the consistency of the judgment matrix is qualified under the condition that the consistency ratio index is in a preset value range, and determining the relative weight of a plurality of indexes through the maximum eigenvalue of the judgment matrix and the corresponding eigenvector;

and under the condition that the consistency ratio index is not in the preset value range, determining that the consistency of the judgment matrix is unqualified, and modifying the elements of the judgment matrix until the consistency of the judgment matrix is qualified.

5. The method of claim 4, wherein determining a first weight for a plurality of metrics from the metric matrix comprises:

carrying out non-dimensionalization on the original data in the index matrix;

determining an optimal value in a row/column corresponding to each index in an index matrix through the non-dimensionalized index matrix, wherein the optimal value is a maximum value or a minimum value, the optimal value is the maximum value under the condition that the type of the index is a profitability index, and the optimal value is the minimum value under the condition that the type of the index is a cost index;

and determining a first weight of the index according to the optimal value and the relative weight corresponding to the index.

6. The method of claim 2, wherein determining entropy weights for a plurality of metrics comprises:

carrying out dimensionless processing on the original data of the index;

calculating an entropy value and a difference coefficient of the index according to the processed original data of the index;

and determining the entropy weight of the index according to the entropy value and the difference coefficient.

7. The method of any one of claims 1 to 6, wherein the target object comprises: customer, business, territory.

8. The method of claim 7, wherein after determining the risk level of the target object based on the plurality of indicators and the corresponding target weights, the method further comprises:

updating risk data of the risk of the client according to the risk grade;

identifying the risk of the client according to the updated risk data;

the risk data of the client risk comprise internal data and external data, wherein the internal data are data of a client account inside the system, and the external data are client data outside the system.

9. The method of claim 7, wherein after determining the risk level of the target object based on the plurality of indicators and the corresponding target weights, the method further comprises:

updating risk data of business risk according to the risk grade;

training a service risk identification model according to the updated risk data, wherein the service risk identification model is a semi-supervised fuzzy clustering algorithm (FCM) model;

wherein the risk data of business risk includes business risk data and customer data related to the customer.

10. The method of claim 7, wherein after determining the risk level of the target object based on the plurality of indicators and the corresponding target weights, the method further comprises:

updating risk data of the region risk according to the risk grade;

generating region risk characteristics according to the updated risk data;

the risk data of the regional risk comprise regional system data, service data corresponding to regions and related to services and customer data related to customers.

11. A risk data identification device, comprising:

the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module acquires risk data of a target object in multiple dimensions, and the risk data comprises multiple indexes of different levels in each dimension;

a first confirmation module that determines a first weight and an entropy weight of a plurality of indicators;

the second determining module is used for determining target weights corresponding to the multiple indexes according to the first weight and the entropy weight;

and the third determining module is used for determining the risk level of the target object according to the plurality of indexes and the corresponding target weights.

12. A computer-readable storage medium, characterized in that the storage medium is used to store a program, wherein the program when running executes the risk data identification method of any of claims 1 to 10.

13. An electronic device comprising one or more processors and memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the risk data identification method of any of claims 1-10.