CN115526700A - Risk prediction method and device and electronic equipment - Google Patents
Risk prediction method and device and electronic equipment Download PDFInfo
- Publication number
- CN115526700A CN115526700A CN202210983442.7A CN202210983442A CN115526700A CN 115526700 A CN115526700 A CN 115526700A CN 202210983442 A CN202210983442 A CN 202210983442A CN 115526700 A CN115526700 A CN 115526700A
- Authority
- CN
- China
- Prior art keywords
- risk
- objects
- data
- information
- object information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a risk prediction method and device and electronic equipment. Relating to the field of financial science and technology or other related fields, the method comprises the following steps: acquiring object information of an object to be predicted, wherein the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted; the method comprises the steps of processing object information of an object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to object information of N objects, the object information at least comprises structural feature data, and the structural feature data represent incidence relation data among the N objects. The invention solves the technical problem of low risk grade prediction efficiency in the prior art.
Description
Technical Field
The invention relates to the field of financial science and technology or other related fields, in particular to a risk prediction method and device and electronic equipment.
Background
The credit business is a foundation for the development of banking business, and the application degree of financial technology determines the highest point of future banking business. At present, in the prior art, most of business modes are established on the basis of single production type or business type enterprises, and risk assessment is carried out on business objects only by means of financial information and business information, but the business modes have the problems of insufficient internal and external data integration application, low automation degree, excessive data representation and the like, and are key factors for restricting further scale development and risk control of bank credit business.
Meanwhile, with the emergence of emerging business models such as internet platform economy and the like, various symbiotic coexistence relations occur between a plurality of enterprises and industry chains, and the enterprises are not single production type or business type enterprise individuals any more, but are cross-country group enterprises with complex stock right and organization structure, multiple industries in layout and wide operation range, even different laws and supervision environments, and the factors are in continuous dynamic change, so that the problem that the real operation risk condition is difficult to be comprehensively reflected exists in a risk prediction mode based on reports such as finance and tax.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the invention provides a risk prediction method, a risk prediction device and electronic equipment, and at least solves the technical problem of low risk level prediction efficiency in the prior art.
According to an aspect of an embodiment of the present invention, there is provided a risk prediction method, including: acquiring object information of an object to be predicted, wherein the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted; the method comprises the steps of processing object information of an object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to object information of N objects, the object information at least comprises structural feature data, and the structural feature data represent incidence relation data among the N objects.
Further, the object model is generated by: acquiring object information of N objects; carrying out feature extraction on object information of the N objects to obtain feature data, wherein the feature data at least comprise structural feature data, basic feature data and risk feature data, the basic feature data represent inherent attribute features of the objects to be predicted, and the risk feature data represent indexes for evaluating whether the N objects are at risk or not; and training the preset model based on the characteristic data to generate a target model.
Further, after object information of the N objects is acquired, an associated object corresponding to each object in the N objects is determined based on the object information; and generating an association relation table based on the N objects and the associated objects.
Further, the risk prediction method further comprises: determining a relationship network based on the association relationship table, wherein the relationship network at least comprises a weight value for representing the association relationship between each object and the associated object, and different association relationships correspond to different weight values; and extracting the characteristics of the relation network to obtain structural characteristic data.
Further, the risk prediction method further comprises: and processing the object information based on the early warning rules contained in the preset risk early warning model to generate risk characteristic data.
Further, generating risk profile data by at least one of: determining turnover funds required by the first type of object in a first preset period based on the goods turnover rate, the account turnover rate and the income information corresponding to the first type of object, and generating risk characteristic data when the turnover funds required in the first preset period are smaller than the outstanding loan amount of the first type of object, wherein the first type of object represents an object for selling goods; determining the sum of loans corresponding to the second type of objects based on the amount of outstanding loans of the second type of objects and the income amount of the second type of objects in a second preset period, and generating risk characteristic data when the sum of loans is greater than the income amount of the second type of objects in the second preset period, wherein the second type of objects represent objects for trading real estate articles; and determining loan information of the target shareholder based on the shareholder information of the target shareholder corresponding to the N objects and credit investigation data corresponding to the target shareholder, and generating risk characteristic data when the loan information of the target shareholder is abnormal, wherein the target shareholder is the shareholder with the largest share ratio among the M shareholders.
Further, the risk prediction method further comprises the following steps: acquiring feature data of N objects in a third-party risk rating mechanism; obtaining risk grade sample data based on the characteristic data and the characteristic data of the third-party risk rating mechanism; and training the preset model based on the risk grade sample data to generate a target model.
Further, after the object information is processed based on the early warning rules contained in the preset risk early warning model to generate risk characteristic data, an early warning record is generated based on the risk characteristic data; determining risk levels and object attributes corresponding to the N objects based on the early warning records, wherein different object attributes correspond to different classes of objects; determining a risk detection process based on the risk level and the object attributes, wherein different object attributes correspond to different risk detection processes; and detecting the early warning record based on the risk detection process to obtain a detection result.
According to another aspect of the embodiments of the present invention, there is also provided a risk prediction apparatus, including: the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring object information of an object to be predicted, and the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted; the processing module is used for processing object information of the object to be predicted based on the target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in the current time period and the next time period of the current time period, the target model is generated according to the object information of the N objects, the object information at least comprises structural feature data, and the structural feature data represent data of incidence relations among the N objects.
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned risk prediction method when running.
According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including one or more processors; a memory for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement a method for predicting risk as described above, wherein the program is arranged to execute when executed.
According to another aspect of embodiments of the present invention, there is also provided a computer program product comprising computer programs/instructions which, when executed by a processor, implement the above-mentioned risk prediction method.
In the embodiment of the invention, a mode of processing object information of an object to be predicted based on a target model to obtain a prediction result is adopted, firstly, the object information of the object to be predicted is obtained, wherein the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted, then the object information of the object to be predicted is processed based on the target model to obtain the prediction result, the prediction result represents credit risk levels of the object to be predicted in the current time period and the next time period of the current time period, the target model is generated according to the object information of N objects, the object information at least comprises structural characteristic data, and the structural characteristic data represents the association relation between the N objects.
In the process, the target model is generated by training according to the object information of a plurality of target objects, the object information at least comprises structural characteristic data representing the incidence relation between N objects, and because the real operation risk condition is difficult to be fully reflected in the risk prediction mode based on statements such as finance, tax and the like in the prior art, the model is trained through the structural characteristic data, the topological structure of the object in a complex environment is mined, the limitation that the prior art is only limited to the object is broken through, the influence of external environmental factors is also integrated into the risk early warning of the object, the credit risk management is facilitated to be improved in the risk prediction efficiency, and credit risk is better prevented.
Therefore, the method and the device achieve the purpose of processing the object information of the object to be predicted based on the target model to obtain the prediction result, achieve the technical effect of improving the risk prediction efficiency of credit risk management, and further solve the technical problem of low risk level prediction efficiency in the prior art.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a schematic diagram of an alternative risk prediction method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of an alternative risk prediction method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an alternative risk prediction device according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of an alternative electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention 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 is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. 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.
It should be noted that the related information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) related to the present invention are information and data authorized by the user or sufficiently authorized by each party. For example, an interface is provided between the system and the relevant user or institution, and before obtaining the relevant information, an obtaining request needs to be sent to the user or institution through the interface, and after receiving the consent information fed back by the user or institution, the relevant information needs to be obtained.
Example 1
In accordance with an embodiment of the present invention, there is provided an embodiment of a risk prediction method, it should be noted that the steps illustrated in the flowchart of the drawings 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 flowchart, in some cases, the steps illustrated or described may be performed in an order different than that herein.
Fig. 1 is a schematic diagram of a risk prediction method according to an embodiment of the present invention, as shown in fig. 1, the method includes the following steps:
step S101, object information of an object to be predicted is obtained, wherein the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted.
In step S101, object information of the object to be predicted may be acquired based on a system, a server, an electronic device, or the like, and in this implementation, the object information of the object to be predicted may be acquired by the system. Alternatively, the object to be predicted may be a customer, for example, a customer of a bank, including a business and an individual.
Optionally, when the object to be predicted may be an enterprise client, the object information further includes scientific and technological innovation achievement data of the enterprise, for example, authorized technical patent data.
Step S102, processing object information of an object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to the object information of the N objects, the object information at least comprises structural feature data, and the structural feature data represent data of incidence relations among the N objects.
In step S102, the system processes object information of the object to be predicted through the trained target model to obtain a prediction result, wherein a relationship network between objects is constructed based on the object information of a plurality of objects, and feature extraction is performed on the relationship network to obtain structural feature data.
Based on the schemes defined in steps S101 to S102, it can be known that, in the embodiment of the present invention, a manner of obtaining a prediction result by processing object information of an object to be predicted based on a target model is adopted, first, by obtaining object information of the object to be predicted, where the object information at least includes credit investigation data, property data, and transaction data of the object to be predicted, and then processing the object information of the object to be predicted based on the target model, so as to obtain the prediction result, where the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to the object information of N objects, the object information at least includes structural feature data, and the structural feature data is data representing an association relationship between the N objects.
It is easy to note that, in the above process, the target model is generated by training according to the object information of a plurality of target objects, the object information at least includes structural feature data representing the association relationship between N objects, and since it is difficult to comprehensively reflect the real business risk status in the risk prediction mode based on the reports of finance, tax, and the like in the prior art, the model is trained through the structural feature data, the topological structure of the object in the complex environment is mined, the limitation that the prior art is only limited to the object itself is broken, the influence of external environmental factors is also integrated into the risk early warning of the object, which is beneficial to improving the risk prediction efficiency of credit risk management and better preventing credit risk.
Therefore, the method and the device achieve the purpose of processing the object information of the object to be predicted based on the target model to obtain the prediction result, achieve the technical effect of improving the risk prediction efficiency of credit risk management, and further solve the technical problem of low risk level prediction efficiency in the prior art.
In an alternative embodiment, the object model of the system is generated by: acquiring object information of N objects; then, carrying out feature extraction on the object information of the N objects to obtain feature data, wherein the feature data at least comprise structural feature data, basic feature data and risk feature data, the basic feature data represent inherent attribute features of the objects to be predicted, and the risk feature data represent indexes for evaluating whether the N objects are at risk or not; and training the preset model based on the characteristic data to generate a target model.
Optionally, as shown in fig. 2, feature engineering is performed on the object information, where the basic feature data includes feature data of the object, such as litigation amount, negative information amount, authorized patent number, qualification certificate number, registrar number, and evaluation index, where the evaluation index is a difference between the positive evaluation and the negative evaluation. The risk characteristic data is generated based on a preset risk early warning model and consists of early warning numbers generated in a single batch and risk grades.
Optionally, the structural feature data is a fusion feature obtained by establishing a relationship between target objects through different relationships, and the variation factors of non-target objects are also included in the analysis of research objects, so that not only is the individual limitation broken through, but also the problem of insufficient data integration application is solved. Meanwhile, the structural characteristics can describe the topological structural characteristics of the object in a real complex environment, better classification capability is provided, and the defect that credit analysis is only performed around the representation data in the prior art is overcome.
It should be noted that both the basic feature data and the risk feature data are feature mining based on the own dimension of the object, belong to single object analysis, and do not break the individual limitation, but in combination with the structural feature of the structural feature data, the variation factors of the non-target object are also brought into the analysis of the research object, so that the individual limitation is broken through, and the problem of insufficient data integration application is solved.
Further, the system determines an associated object corresponding to each object in the N objects based on the object information after acquiring the object information of the N objects; and then generating an association relation table based on the N objects and the associated objects.
Optionally, as shown in fig. 2, the system preprocesses the acquired data, constructs an association table based on relationships among shareholders, high governments, investments, branches, guarantees, operations, transactions, proprietary technologies, litigation, and operation scopes in the object information, and stores the association table in a preset relational database.
It should be noted that, by constructing the association relationship table, basic data is provided for subsequently constructing the relationship network, and accuracy of obtaining the feature data is improved.
Further, the system performs feature extraction on the object information of the N objects to obtain feature data, including: determining a relationship network based on an association relationship table, wherein the relationship network at least comprises a weight value for representing the association relationship between each object and the associated object, and different association relationships correspond to different weight values; and then, extracting the characteristics of the relation network to obtain structural characteristic data.
Optionally, the system constructs, based on the association relationship table, a relationship network G = (V, E) as shown in fig. 2, where V represents a node set, E represents an edge set, the node represents a credit client, the edge represents that there is some association relationship between two credit objects, and weights of the edges are differentiated positively and negatively, that is, the edge weight that is a mutual aid relationship between the credit objects is a positive number, the edge weight that is a mutual exclusion relationship between the credit objects is a negative number, where the mutual aid relationship is a cooperation relationship between the objects, and the mutual exclusion relationship is a competition relationship between the objects.
Optionally, feature extraction is performed based on a relational network to obtain structural feature data, where the structural feature data is based on node neighbor number, path, feature vector, network random walk and other methods, and the specific features include degree centrality, aggregation coefficient, k-shell, near centrality, betweenness centrality, centrifugal centrality, feature vector centrality, katz centrality, random walk and the like. The degree centrality is the importance of the nodes measured by the connection degree between the nodes, and the influence is larger when the relationship of the nodes is more; the aggregation coefficient is used for describing the degree coefficient of clustering among the nodes, namely the degree of interconnection among the neighbor nodes of one node; the k-shell is a fine-grained node importance classification method and is a result obtained by decomposing a network on the basis of node degree; the proximity between the nodes is described by the proximity centrality, and the proximity is defined as the average value of the distance between the target node and other nodes in the network; the betweenness centrality is the importance of the node measured by calculating the number of the shortest paths passing through the target node in the shortest paths of all node pairs in the network; the centrifugal centrality is the maximum value of the shortest distance between a target node and other nodes in the network, and the smaller the centrifugal centrality is, the more important the node is; the feature vector centrality evaluates the importance of the target node based on the importance of the neighbor nodes of the target node; katz centrality is an evaluation index for increasing the importance of a target node on the basis of the centrality of a feature vector; random walk is an algorithm for personalized ranking of nodes in a network.
It should be noted that, definition of the edge relationship is added on the basis of the traditional network analysis, and the analysis capability of the complex relationship is improved.
In another optional embodiment, the system performs feature extraction on the object information of the N objects to obtain feature data, including: and processing the object information based on the early warning rules contained in the preset risk early warning model to generate risk characteristic data.
Optionally, the service early warning model constructed as shown in fig. 2 is triggered to generate risk feature data of the batch once, early warning of different batches can obtain a time sequence risk feature data set, and then a time sequence feature matrix is extracted by a TsFresh method. The dynamic monitoring of the entity risk can be realized based on the time-series risk characteristic data, and the future trend can be predicted favorably.
Optionally, the preset early warning model is defined based on the aspects of the operation state, negative information, loan state, financing condition, cross default condition, mortgage investigation and sealing condition, income and expense matching condition, guarantee condition, non-silver financing and the like of the object.
Further, the risk characteristic data is generated by at least one of: determining turnover funds required by the first type of object in a first preset period based on the goods turnover rate, the account turnover rate and the income information corresponding to the first type of object, and generating risk characteristic data when the turnover funds required in the first preset period are smaller than the outstanding loan amount of the first type of object, wherein the first type of object represents an object for selling goods; determining a total loan amount corresponding to the second type of object based on the outstanding loan amount of the second type of object and the income amount of the second type of object in a second preset period, and generating risk characteristic data when the total loan amount is larger than the income amount of the second type of object in the second preset period, wherein the second type of object represents the object for trading real estate articles; and determining loan information of the target shareholder based on the shareholder information of the target shareholder corresponding to the N objects and credit investigation data corresponding to the target shareholder, and generating risk characteristic data when the loan information of the target shareholder is abnormal, wherein the target shareholder is the shareholder with the largest share ratio among the M shareholders.
Optionally, the preset early warning model includes an excessive financing model, early warning monitoring of credit and operation conditions of the credit customer is achieved, and monitoring contents include information such as outstanding credit information, sales amount and fund demand. Wherein, when the object represents an object selling goods, such as a wholesale and retail enterprise, the early warning rule is defined as follows:
wherein, delta is the stock turnover rate,
the receivable turnover rate is MI which is the income of the main operation business, MI 'which is the income of the main operation business at the last year, R which is the receivable, R' which is the receivable in the previous year, MC which is the cost of the main operation business, CL which is the sum of outstanding credit and loan, and PL which is the sum of outstanding small loan. If the above rules are met, i.e. the enterprise does not settle a loan for more than its one cash cycle, the model triggers an early warning and the risk level is defined as low risk.
Where an object characterizes an object that trades real estate items, such as a real estate enterprise, the warning rules are defined as follows:
CL + PL > MI ', MI' >0, alpha = 'Normal'
Wherein CL is the sum of outstanding credit loan, PL is the sum of outstanding small loan, MI' is the main business income in the last year, and alpha is the five-level classification state of loan. If the rules are met, namely the amount of outstanding loans of the enterprise exceeds the sales amount of the last half year, the model triggers early warning, and the risk level is defined as low risk.
When the object is manufacturing and other enterprises, the early warning rules are defined as follows:
CL+PL>MI′,MI′>0
wherein CL is the sum of outstanding credit loans, PL is the sum of outstanding small loans, and MI' is the income of main business in the last year. If the rules are met, namely the amount of outstanding loans of the enterprise exceeds the sales amount of the last half year, the model triggers early warning, and the risk level is defined as low risk.
Optionally, the preset early warning model includes a cross default model, so that early warning monitoring of default information of shareholder clients is realized, whether the maximum shareholder information is in a shareholder list of a default list is monitored, wherein the default list is a set obtained by traversing all credit investigation data and screening the shareholder information of enterprises which meet the requirements that loan in the unclassified credit information is classified into secondary, suspicious and lost. The early warning rules are defined as follows:
ls∩DL＝ls
wherein ls and DL are data in the same period, ls is the maximum shareholder information of a certain enterprise, and DL is a default list set. If ls ≠ NULL, which indicates that the loan classification of the largest stockholder of the enterprise in another enterprise is bad, the model triggers early warning, and the risk level is defined as high risk.
Optionally, the preset early warning model includes a mortgage checking model, so as to realize early warning monitoring on the real estate ticket change condition of the credit customer, and the monitoring content includes the mortgage or checking condition of the real estate ticket. The early warning rules are defined as follows:
where RE is a real estate ticket number of a target organization (e.g., a bank) that all stocks have not released a real estate mortgage, SR is latest one-month audit information of real estate mortgage data, and MR is latest one-month mortgage registration information of real estate mortgage data. If it is used
The real estate title number indicating that the enterprise does not release the real estate mortgage in the target organization stock has a newly added mortgage or verified real estate title number, the model triggers early warning, and the risk level is defined as medium risk.
Optionally, the preset early warning model includes a revenue and expenditure trend model, so as to realize early warning monitoring on the production and operation conditions of credit customers, and the monitoring content mainly refers to the correlation between the business revenue of main business and the electricity utilization condition. The early warning rules are defined as follows:
wherein MI is the main business income of this year, MI 'is the main business income of the previous year, e is the electricity payment amount of this year, e' is the electricity payment amount of the previous year, and
if S17 ≧ b 4 When the relation between the business income increase and the power consumption reduction of the enterprise is not matched, the model triggers early warning, and a is more than or equal to S17 4 The risk level is defined as middle risk when S17>a 4 The risk class is defined as the intermediate risk, wherein b 4 Is a first warning threshold, a 4 The second early warning threshold value is larger than the first early warning threshold value.
Optionally, the preset early warning model includes an excessive guarantee model to realize early warning monitoring of the external guarantee condition of the credit customer, and the monitoring content mainly refers to a relationship between the external guarantee amount of the enterprise and the rights and interests of the owner of the enterprise. The early warning rules are defined as follows:
S18＝FA-OE
wherein, FA is the non-unsecured financing guarantee amount in the financing information of the financial institution, and OE is the owner's equity sum in the balance sheet of the latest first credit data. If S18>0, it indicates that the enterprise has exceeded its owner' S interest in the wagering, the model triggers an early warning, and the risk level is defined as intermediate risk.
Optionally, the preset early warning model comprises a non-silver financing model, so that early warning monitoring on the condition of the small credit of the credit customer is realized, and the monitoring content mainly refers to the change condition of the unclassified small credit financing in the early years. The early warning rules are defined as follows:
wherein PL' is the annual unsettled petty loan, and PL is the latest annual unsettled petty loan. If the rule of S19 is satisfied, namely the enterprise has newly increased unbeared small credit financing, the model triggers early warning, and the risk level is defined as medium risk.
Optionally, the preset early warning model comprises an operation state model, early warning monitoring of the operation condition of the credit customer is achieved, and the monitoring content comprises the operation state, stockholder change, social security number change, water, electricity, gas payment information condition and enterprise income tax payment condition of the customer. The early warning rules are defined as follows:
wherein S1 is the latest first-term business state. If S1 exists, the operation state of the enterprise is abnormal, the model triggers early warning, and the risk level is defined as high risk.
S2＝G1∪G2∪G3
Wherein, G1 is a shareholder name set, G2 is a shareholder type set, G3 is a funding proportion set, and S2 is latest shareholder information of the first period. S2' = G1 '. U G2 '. U G3' is information of the stockholder in the last period, if S2 ≠ S2 '. Noteq.NULL, it is indicated that the information of the stockholder of the enterprise is changed, the model triggers early warning, and the risk level is defined as medium risk.
Wherein S' is the number of social security at the beginning of the year, S is the number of the latest social security at the first period, S3 is the social security change proportion, and if S3 is more than or equal to b 0 The model triggers early warning when the social security payment number of the enterprise declines to a larger extent in the early years, and a is more than or equal to S3 0 The risk class is defined as low risk when S3>a 0 The risk class is defined as the intermediate risk, wherein b 0 Is a third warning threshold value, a 0 Is a fourth pre-warning threshold, the third pre-warning threshold is less than the fourth pre-warning threshold.
Wherein e' is the current electricity payment amount in the same period of the last year, e is the latest current electricity payment amount in the first period, S4 is the change ratio of the electricity payment information, and if S4 is more than or equal to b 1 The method indicates that the electricity consumption payment of the enterprise is reduced by a larger extent than the same period of the last year, the model triggers early warning, and a is more than or equal to S4 1 Risk level is defined as low windDanger, when S4>a 1 The risk class is defined as the intermediate risk, wherein b 1 Is a fifth warning threshold, a 1 And the fourth early warning threshold is a sixth early warning threshold, and the fifth early warning threshold is smaller than the sixth early warning threshold.
Wherein w' is the payment amount of the same-period water in the last year, w is the latest payment amount of the first-period water, S5 is the change proportion of the payment information of the water, and if S5 is more than or equal to b 2 The water consumption payment of the enterprise is shown to be reduced by a larger extent than the same period of the last year, the model triggers early warning, and a is more than or equal to S5 2 The risk level is defined as low risk if S5>a 2 The risk class is defined as the intermediate risk, wherein b 2 Is a seventh warning threshold, a 2 And the seventh early warning threshold value is smaller than the eighth early warning threshold value.
Wherein c' is the tax amount paid by the same-period enterprise in the last year, c is the latest tax amount paid by the same-period enterprise, S6 is the change proportion of the tax obtained by the same-period enterprise, and if S6 is more than or equal to b 3 The model triggers early warning when the income tax of the enterprise paid by the enterprise is reduced by a larger extent than the same period of the last year, and a is more than or equal to S6 3 The risk level is defined as low risk when S6>a 3 The risk class is defined as the intermediate risk, wherein b 3 Is the ninth warning threshold, a 3 The tenth early warning threshold value, the ninth early warning threshold value is smaller than the tenth early warning threshold value.
Optionally, the preset early warning model comprises a negative information model, early warning monitoring of complaint information and other negative information of the credit customer is achieved, monitoring contents comprise plan information, plan settlement information, information of a person who is out of credit and is executed, debt payment and overdue information, punishment information, production listing correction information, information of abnormal users of tax departments, information of enterprises losing customs and environment-friendly and other negative evaluation information of the customer, and the negative information does not give an early warning repeatedly. The early warning rules are defined as follows:
S7≠NULL
wherein, S7 is the information related to the filing of a case. If S7 is not equal to NULL, the enterprise is indicated to have newly-added case information, the model triggers early warning, if the enterprise is legal customer, the risk level is defined as low risk, and if the enterprise is popular customer, the risk level is defined as medium risk.
S8>m(t≤6)
Wherein, S8 is the information related to the case closing, m is the threshold value of the case closing mark, and t is the case closing date (unit: month). If S8 is not equal to NULL, the enterprise is indicated to have newly added settlement information, if the settlement date is within 6 months and the settlement mark is larger than m ten thousand yuan, the model triggers early warning, and the risk level is defined as low risk.
S9≠NULL
Wherein, S9 is information of the person to be executed without information. If S9 is not equal to NULL, the enterprise is indicated to have newly added information of the information-losing executed person, the model triggers early warning, and the risk level is defined as high risk.
S10＝F0∪F1∪F2∪F3∪F4∪F5
Wherein, F0 is information of debt and overdue, F1 is information of punishment, F2 is information of production listing correction, F3 is information of abnormal household of tax department, F4 is information of customs information loss enterprise, F5 is information of environmental protection and other negative evaluations, and S10 is negative information. If F0 is not equal to NULL, new debt payment or overdue information is indicated, the model triggers early warning, and the risk level is defined as high risk; f2, F3, F4, F5 are the same as F0.
If F1 ≠ NULL, it is defined as follows:
wherein, forfeit is a fine amount, if F1= other, namely no fine appears in the fine information, the model triggers early warning, the risk grade is defined as high risk, if forfeit belongs to F1, namely the fine information is a fine record, the model triggers early warning, and when forfeit is not more than F 1 The risk level is defined as low risk when f 1 <forfeit≤f 2 The risk class is defined as medium risk, when forteit>f 2 The risk level is defined as high risk, wherein f 1 Is the eleventh warning threshold, f 2 Is a twelfth pre-warning threshold value, the eleventh pre-warning threshold value is smaller than the twelfth pre-warning threshold value.
Optionally, the preset early warning model includes a loan abnormity model, early warning monitoring on unsettled credit information of a credit customer is achieved, and monitoring contents include unsettled loan information and unsettled micropayment information. The early warning rules are defined as follows:
S11＝count(α,β,γ)
wherein, α is the five-level classification state of the loan, β is the financial institution unit, γ is the loan type (outstanding credit information, outstanding small loan information), S11 is the abnormal situation of the loan of other lines in the latest period, and S11' is the abnormal situation of the loan of other lines in the last period. The specific situation is as follows:
when alpha = ' attention ', beta ≠ ' own institution ', and gamma = ' un-cleared credit information ', if S11> S11', the new attention class is added to the loan of the other lines of the client, the model triggers early warning, and the risk level is defined as high risk.
When alpha = ' secondary/suspicious/loss ', beta ≠ ' own institution ', and gamma = ' un-cleared credit information ', if S11> S11', the model indicates that the three types of failures are caused after the new loan on other lines of the client, the model triggers early warning, and the risk level is defined as high risk.
When alpha is not equal to ' normal ', gamma is not equal to ' unpaid small loan information ', if S11 is greater than S11', the customer is indicated to newly add abnormal small loans, the model triggers early warning, and the risk level is high risk.
It should be noted that the object information is processed according to the pre-set early warning rules of the early warning model, clear quantitative indexes are provided, the early warning model is defined systematically, comprehensively and differentially, the defect that the existing credit early warning method is 'one-look-the-same-kernel' is overcome, and the early warning accuracy is improved.
Further, the system trains the preset model based on the characteristic data to generate a target model, and comprises: obtaining characteristic data of N objects in a third-party risk rating mechanism; then based on the characteristic data and the characteristic data of the third-party risk rating mechanism, obtaining risk level sample data; and finally, training the preset model based on the risk grade sample data to generate a target model.
Optionally, the system generates risk level sample data by performing label definition on the feature data acquired by the third-party risk rating mechanism and the feature data, and the definition is as follows:
C＝∑c i ω i
wherein the customer evaluation comprises the organization and the mainstream rating organization, c i For customer evaluation of the ith organization, ω i For the weight scored for the ith customer, the C is graded to include six grades, A +, A-, B +, B, and B-.
Further, a LightGBM algorithm is adopted to construct a distributed gradient lifting model, sample data is screened by adopting a gradient unilateral sampling (GOSS) method, wherein the gradient unilateral sampling refers to that samples with large gradients are reserved according to a proportion a and samples with small gradients are randomly sampled according to a proportion b by calculating gradients of all samples. And then, reducing the number of features by adopting an Exclusive Feature Binding (EFB) method, wherein the exclusive feature binding refers to binding the exclusive features into one feature, and solving the problem of model efficiency reduction possibly caused by feature data sparsity.
The risk level sample data is split according to the information Gain, the sum of the sample gradients of different value ranges of each characteristic is calculated through a histogram algorithm, the corresponding values are substituted into a Gain formula Gain, and the Gain of each characteristic divided by each different value range is calculated. Gain is defined as follows:
wherein a and b are both proportional parameters, A l For a set of samples with a large gradient, A r For r sample sets with large gradients, B l For a set of samples with small gradient, B r For r samples with small gradientCollection, g i Is the gradient value of a certain feature of the i sample,
in order to obtain the total number of samples of the T node after the samples are screened by the GOSS method,
in order to obtain the total sample number of the left subnodes of the T node after the samples are screened by the GOSS method,
and the total sample number of the right child nodes of the T node after the samples are screened by the GOSS method.
And finally, finding out the node with the maximum gain in all leaf nodes as an optimal segmentation point by adopting a leaf-wise leaf growth strategy with depth limitation, and segmenting the data of the leaf nodes corresponding to the optimal segmentation point into two batches. Recursion is continued until the condition for stopping generation is satisfied, resulting in a target model, such as the credit customer evaluation model shown in FIG. 2.
It should be noted that, by combining the model performance and the feature importance evaluation method, the model prediction accuracy is improved by continuously iterative optimization.
Further, after the system processes the object information based on the early warning rules contained in the preset risk early warning model and generates risk characteristic data, an early warning record is generated based on the risk characteristic data; then, determining risk levels and object attributes corresponding to the N objects based on the early warning records, wherein different object attributes correspond to different classes of objects; determining a risk detection process based on the risk level and the object attributes, wherein different object attributes correspond to different risk detection processes; and finally, detecting the early warning record based on a risk detection process to obtain a detection result.
Optionally, the system sends the early warning record generated by triggering the early warning model to the user management manager for processing through the early warning management shown in fig. 2, where the risk level of the early warning record may be low risk, medium risk, or high risk; the early warning type can be classified into reading and verification according to risk level; the verification process is jointly determined by two factors of risk level and client attribute, the client attribute is divided into legal (enterprise) client and general (personal) client, namely the risk level is the same, and the verification processes of the legal client and the general client are different, so that the legal client and the general client are ensured to know as much as possible and are implemented to specific implementers.
It should be noted that, different object attributes correspond to different risk detection flows, so that the efficiency of early warning management is improved and computing resources are saved while service flexibility and necessary service processing flows are ensured.
Therefore, the invention provides a new risk prediction method, which gives clear quantitative indexes to the early warning model, systematically, comprehensively and differentially defines the early warning model, makes up the defect of 'one view with one kernel' of the existing credit early warning method, and can better guide the development of credit business. Compared with the prior art, in the aspect of early warning management, the early warning process is defined through the risk level and the customer attribute, and the definition mode supports customization, so that the early warning management efficiency is improved and the computing resource is saved while the business flexibility and the necessary business processing process are ensured. In addition, on the basis of optimizing the prior art architecture, the invention innovatively provides that the topological structure of an enterprise in a complex environment is mined by constructing a credit customer relationship network from a complex system theory, so that the limitation of the prior art only to customers is broken, the influence of external environmental factors is also integrated into the risk early warning of credit customers, the effect of credit risk management on risk prediction or classification is favorably improved, and credit risk is better prevented.
Example 2
According to an embodiment of the present invention, an embodiment of a risk prediction apparatus is provided, where fig. 3 is a schematic diagram of an alternative risk prediction apparatus according to an embodiment of the present invention, as shown in fig. 3, the apparatus includes:
an obtaining module 301, configured to obtain object information of an object to be predicted, where the object information at least includes credit investigation data, property data, and transaction data of the object to be predicted; the processing module 302 is configured to process object information of an object to be predicted based on a target model to obtain a prediction result, where the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to object information of N objects, the object information at least includes structural feature data, and the structural feature data is data representing an association relationship between the N objects.
It should be noted that the acquiring module 301 and the processing module 302 correspond to steps S101 to S102 in the above embodiment, and the two modules are the same as the examples and application scenarios realized by the corresponding steps, but are not limited to the disclosure of the above embodiment 1.
Optionally, the risk prediction apparatus further comprises: the device comprises a first acquisition module, a first extraction module and a first generation module; the first acquisition module is used for acquiring the object information of the N objects; the first extraction module is used for carrying out feature extraction on object information of the N objects to obtain feature data, wherein the feature data at least comprise structural feature data, basic feature data and risk feature data, the basic feature data represent inherent attribute features of the objects to be predicted, and the risk feature data represent indexes for evaluating whether the N objects are at risk or not; the first generation module is used for training a preset model based on the characteristic data to generate a target model.
Optionally, the risk prediction apparatus further comprises: the device comprises a first determining module and a second generating module; the first determining module is used for determining an associated object corresponding to each object in the N objects based on the object information; the second generation module is used for generating an association relation table based on the N objects and the association objects.
Optionally, the risk prediction apparatus further comprises: a second determining module and a second extracting module; the second determining module is used for determining a relationship network based on the association relationship table, wherein the relationship network at least comprises a weight value representing the association relationship between each object and the associated object, and different association relationships correspond to different weight values; the second extraction module is used for extracting the characteristics of the relationship network to obtain structural characteristic data.
Optionally, the risk prediction device further comprises: and the third generation module is used for processing the object information based on the early warning rules contained in the preset risk early warning model and generating risk characteristic data.
Optionally, the risk prediction device further comprises: a fourth generation module, a fifth generation module and a sixth generation module; the fourth generation module is used for determining turnover funds required by the first type of object in a first preset period based on the goods turnover rate, the account turnover rate and the income information corresponding to the first type of object, and generating risk characteristic data when the turnover funds required in the first preset period are smaller than the outstanding loan amount of the first type of object, wherein the first type of object represents an object for selling goods; the fifth generation module is used for determining the total loan amount corresponding to the second type of object based on the unblended loan amount of the second type of object and the income amount of the second type of object in a second preset period, and generating risk characteristic data when the total loan amount is larger than the income amount of the second type of object in the second preset period, wherein the second type of object represents an object for trading real estate articles; the sixth generation module is used for determining loan information of the target shareholder based on the shareholder information of the target shareholder corresponding to the N objects and credit investigation data corresponding to the target shareholder, and generating risk characteristic data when the loan information of the target shareholder is abnormal, wherein the target shareholder is the shareholder with the largest share ratio among the M shareholders.
Optionally, the risk prediction apparatus further comprises: the third acquisition module, the third determination module and the seventh generation module; the third acquisition module is used for acquiring the characteristic data of the N objects in the third-party risk rating mechanism; the third determination module is used for obtaining risk grade sample data based on the characteristic data and the characteristic data of the third-party risk rating mechanism; and the seventh generating module is used for training the preset model based on the risk grade sample data to generate the target model.
Optionally, the risk prediction device further comprises: the device comprises an eighth generation module, a fourth determination module, a fifth determination module and a detection module; the eighth generation module is used for generating an early warning record based on the risk characteristic data; the fourth determining module is used for determining risk levels and object attributes corresponding to the N objects based on the early warning records, wherein different object attributes correspond to different classes of objects; the fifth determining module is used for determining a risk detection process based on the risk level and the object attribute, wherein different object attributes correspond to different risk detection processes; the detection module is used for detecting the early warning record based on the risk detection process to obtain a detection result.
Example 3
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned risk prediction method when running.
Example 4
According to another aspect of the embodiments of the present invention, there is also provided an electronic device, wherein fig. 4 is a schematic diagram of an alternative electronic device according to the embodiments of the present invention, as shown in fig. 4, the electronic device includes one or more processors; a memory for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement a method for predicting risk as described above, wherein the program is arranged to execute when executed.
Example 5
According to another aspect of embodiments of the present invention, there is also provided a computer program product comprising computer programs/instructions which, when executed by a processor, implement the above-mentioned risk prediction method.
The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages and disadvantages of the embodiments.
In the above embodiments of the present invention, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described in detail in a certain embodiment.
In the embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, a division of a unit may be a division of a logic function, and an actual implementation may have another division, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or may not be executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The above is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, a plurality of modifications and embellishments can be made without departing from the principle of the present invention, and these modifications and embellishments should also be regarded as the protection scope of the present invention.
Claims (12)
1. A method for predicting risk, comprising:
acquiring object information of an object to be predicted, wherein the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted;
processing object information of the object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to object information of N objects, the object information at least comprises structural feature data, and the structural feature data are data representing an association relation between the N objects.
2. The method of claim 1, wherein the object model is generated by:
acquiring object information of the N objects;
performing feature extraction on the object information of the N objects to obtain feature data, wherein the feature data at least comprise the structural feature data, basic feature data and risk feature data, the basic feature data represent inherent attribute features of the object to be predicted, and the risk feature data represent indexes for evaluating whether the N objects are at risk or not;
and training a preset model based on the characteristic data to generate the target model.
3. The method of claim 2, wherein after obtaining the object information for the N objects, the method further comprises:
determining an associated object corresponding to each object of the N objects based on the object information;
and generating an association relation table based on the N objects and the association objects.
4. The method of claim 3, wherein performing feature extraction on the object information of the N objects to obtain feature data comprises:
determining a relationship network based on the association relationship table, wherein the relationship network at least comprises a weight value for representing the association relationship between each object and the associated object, and different association relationships correspond to different weight values;
and extracting the characteristics of the relational network to obtain the structural characteristic data.
5. The method of claim 2, wherein performing feature extraction on the object information of the N objects to obtain feature data comprises:
and processing the object information based on an early warning rule contained in a preset risk early warning model to generate the risk characteristic data.
6. The method of claim 5, wherein the risk profile data is generated by at least one of:
determining turnover funds required by a first type of object in a first preset period based on the goods turnover rate, the account turnover rate and income information corresponding to the first type of object, and generating the risk characteristic data when the turnover funds required in the first preset period are smaller than the outstanding loan amount of the first type of object, wherein the first type of object represents an object for selling goods;
determining the sum of loans corresponding to the objects of a second type based on the amount of outstanding loans of the objects of the second type and the income amount of the objects of the second type in a second preset period, and generating the risk characteristic data when the sum of loans is greater than the income amount of the objects of the second type in the second preset period, wherein the objects of the second type represent the objects of trading real estate goods;
and determining loan information of the target shareholder based on the shareholder information of the target shareholder corresponding to the N objects and credit investigation data corresponding to the target shareholder, and generating the risk characteristic data when the loan information of the target shareholder is abnormal, wherein the target shareholder is the shareholder with the largest share ratio among the M shareholders.
7. The method of claim 2, wherein training a pre-set model based on the feature data to generate the target model comprises:
acquiring characteristic data of the N objects in a third-party risk rating mechanism;
obtaining risk grade sample data based on the characteristic data of the third-party risk rating mechanism and the characteristic data;
training the preset model based on the risk grade sample data to generate the target model.
8. The method of claim 6, wherein after processing the subject information based on the pre-set risk pre-warning rules contained in a pre-set risk pre-warning model to generate the risk characteristic data, the method further comprises:
generating an early warning record based on the risk characteristic data;
determining risk levels and object attributes corresponding to the N objects based on the early warning records, wherein different object attributes correspond to different classes of objects;
determining a risk detection process based on the risk level and the object attributes, wherein different object attributes correspond to different risk detection processes;
and detecting the early warning record based on the risk detection process to obtain a detection result.
9. A risk prediction apparatus, comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring object information of an object to be predicted, and the object information at least comprises credit investigation data, property data and transaction data of the object to be predicted;
the processing module is used for processing the object information of the object to be predicted based on a target model to obtain a prediction result, wherein the prediction result represents credit risk levels of the object to be predicted in a current time period and a time period next to the current time period, the target model is generated according to object information of N objects, the object information at least comprises structural feature data, and the structural feature data are data representing association relations among the N objects.
10. A computer-readable storage medium, in which a computer program is stored, wherein the computer program is arranged to, when executed, perform a method of risk prediction as claimed in any one of claims 1 to 8.
11. An electronic device, wherein the electronic device comprises one or more processors; memory for storing one or more programs which, when executed by the one or more processors, cause the one or more processors to implement a method for running a program, wherein the program is arranged to, when run, perform a method of predicting risk as claimed in any one of claims 1 to 8.
12. A computer program product comprising computer programs/instructions, characterized in that the computer programs/instructions, when executed by a processor, implement the method of risk prediction according to any of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210983442.7A CN115526700A (en) | 2022-08-16 | 2022-08-16 | Risk prediction method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210983442.7A CN115526700A (en) | 2022-08-16 | 2022-08-16 | Risk prediction method and device and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115526700A true CN115526700A (en) | 2022-12-27 |
Family
ID=84695795
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210983442.7A Pending CN115526700A (en) | 2022-08-16 | 2022-08-16 | Risk prediction method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115526700A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116681291A (en) * | 2023-08-02 | 2023-09-01 | 杭州小策科技有限公司 | Wind control prediction method and system based on integrated model |
| CN116934464A (en) * | 2023-07-26 | 2023-10-24 | 广东企企通科技有限公司 | Post-loan risk monitoring method, device, equipment and medium based on small micro-enterprises |
-
2022
- 2022-08-16 CN CN202210983442.7A patent/CN115526700A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116934464A (en) * | 2023-07-26 | 2023-10-24 | 广东企企通科技有限公司 | Post-loan risk monitoring method, device, equipment and medium based on small micro-enterprises |
| CN116681291A (en) * | 2023-08-02 | 2023-09-01 | 杭州小策科技有限公司 | Wind control prediction method and system based on integrated model |
| CN116681291B (en) * | 2023-08-02 | 2023-11-07 | 杭州小策科技有限公司 | Wind control prediction method and system based on integrated model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Altman | A fifty-year retrospective on credit risk models, the Altman Z-score family of models and their applications to financial markets and managerial strategies | |
| Lin et al. | Detecting the financial statement fraud: The analysis of the differences between data mining techniques and experts’ judgments | |
| Lismont et al. | Predicting tax avoidance by means of social network analytics | |
| Papik et al. | Detection models for unintentional financial restatements | |
| CN115526700A (en) | Risk prediction method and device and electronic equipment | |
| Torki et al. | The effects of electronic payment systems on the performance of the financial sector in selected Islamic Countries | |
| CN108492001A (en) | A method of being used for guaranteed loan network risk management | |
| Wong et al. | Financial accounting fraud detection using business intelligence | |
| Lei et al. | Development of an intelligent information system for financial analysis depend on supervised machine learning algorithms | |
| Chi et al. | Debt rating model based on default identification: Empirical evidence from Chinese small industrial enterprises | |
| Fu et al. | Property rights protection, financial access and corporate R&D: Evidence from a large representative sample of Chinese firms | |
| KR102139938B1 (en) | System for selection of companies subject to credit guarantees based on credit guarantees propensity analysis | |
| Zhou et al. | A two-stage credit scoring model based on random forest: Evidence from Chinese small firms | |
| Hu | Predicting and improving invoice-to-cash collection through machine learning | |
| Galeano et al. | Shifts in individual parameters of a GARCH model | |
| Marais et al. | Predicting financial statement manipulation in South Africa: A comparison of the Beneish and Dechow models | |
| Cheng et al. | A financial distress pre-warning study by fuzzy regression model of TSE-listed companies | |
| Biswas et al. | Automated credit assessment framework using ETL process and machine learning | |
| Asadi et al. | Evaluation of managerial overconfidence, cash holding, and investment efficiency in companies | |
| Minetti et al. | Institutions, bailout policies, and bank loan contracting: Evidence from Korean chaebols | |
| Li et al. | Construction of Credit Evaluation Index System for Two‐Stage Bayesian Discrimination: An Empirical Analysis of Small Chinese Enterprises | |
| Lee et al. | Application of machine learning in credit risk scorecard | |
| Li et al. | Research on Efficiency in Credit Risk Prediction Using Logistic‐SBM Model | |
| Babaei et al. | Studying the Information Content, Economic Value Added, Liquidity, and Activity in Market Value Added Determination. | |
| JABIN et al. | Bankruptcy risk and income smoothing tendency of NBFIs in Bangladesh |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |