CN116703155A - Risk identification method, risk identification equipment and computer readable storage medium - Google Patents

Abstract

The present application discloses a risk identification method, device and computer-readable storage medium. The method includes: obtaining target business data; using a risk identification model to divide the target business data to obtain risky business data and non-risk business data; using a risk classification model to determine The risk category of risky business data; based on the sample business data generated by the risky business data, the risk identification model is enhanced for training, and based on the sample data to be classified generated by the risky business data of the determined risk category, the risk classification model is Enhanced training, the above method, can improve the accuracy of risk identification.

Description

A risk identification method, device and computer-readable storage medium

technical field

The present application relates to the technical field of artificial intelligence, in particular to a risk identification method, device and computer-readable storage medium.

Background technique

At present, with the wide application of financial management systems, financial-related businesses can be carried out through financial management systems. In order to strengthen capital risk management and improve the security of financial business, financial risk control can be carried out based on the business data generated by the financial management system.

At present, there are ways to use risk identification models for risk management and control, but the accuracy of existing risk identification models is low, and the accuracy of risk identification is low. Therefore, how to improve the accuracy of risk identification has become an urgent problem to be solved.

Contents of the invention

The technical problem mainly solved by this application is to provide a risk identification method, device and computer-readable storage medium, which can improve the accuracy of risk identification.

In order to solve the above technical problems, a technical solution adopted by this application is to provide a risk identification method, which includes: obtaining target business data; using a risk identification model to divide target business data to obtain risky business data and non-risky business data; The risk classification model determines the risk category of the risk business data; based on the sample business data generated by the risk business data, the risk identification model is strengthened for training, and based on the sample data to be classified generated by the risk business data of which the risk category has been determined, Augmentation training of risk classification models.

Among them, using the risk identification model to divide the target business data to obtain risk business data and non-risk business data includes: using the risk identification model to sequentially identify direct risk data, indirect risk data and rare risk data from the target business data; Indirect risk data and rare risk data are used together as risky business data, and some business data in the target business data except direct risk data, indirect risk data and rare risk data are regarded as non-risk business data.

Among them, using the risk identification model to sequentially identify the direct risk data, indirect risk data and rare risk data from the target business data includes: obtaining the first probability that the target business data belongs to the direct risk data; determining from the target business data based on the first probability Direct risk data; obtain the second probability that part of the business data except the direct risk data in the target business data belongs to the indirect risk data; determine the indirect risk data from the part of the target business data except the direct risk data based on the second probability Risk data; obtain the third probability that part of the business data in the target business data except direct risk data and indirect risk data belongs to rare risk data; based on the third probability, exclude direct risk data and indirect risk data from the target business data Identify rare risk data in some of the business data.

Wherein, after using the risk identification model to divide the target business data to obtain the risk business data and non-risk business data, before using the risk classification model to determine the risk category of the risk business data, the method also includes: determining the detailed category of the risk business data, wherein the detailed The number of categories is greater than the number of risk categories; using the risk classification model to determine the risk category of risk business data includes: using the risk classification model to map the risk business data of the determined detailed category to the classification function dimension, and determine the corresponding separation of each risk category. plane; in the classification function dimension, based on the separation hyperplane, the risk business data of the determined detailed categories are divided into corresponding risk categories.

Wherein, the method further includes: generating sample business data by using the risk business data of the determined detailed category, wherein the sample business data is marked with a real risk label, and the real risk label indicates that the sample business data belongs to the risk data type and the non-risk data type. What; Based on the sample business data generated by the risk business data, the enhanced training of the risk identification model includes: using the risk identification model to divide the sample business data, obtaining the prediction result of the sample risk type, and determining the sample risk data based on the prediction result of the sample risk type and sample non-risk data, the sample risk prediction result represents which of the risk data type and non-risk data type the predicted sample business data belongs to; adjust the parameters of the risk identification model based on the first difference between the risk prediction result and the real risk label.

Wherein, the method further includes: using the risk business data of the determined risk category to generate sample data to be classified, wherein the sample data to be classified is marked with a real category label, and the real category label represents the risk category to which the sample to be classified data belongs; The enhanced training of the risk classification model includes: using the risk classification model to determine the risk category of the sample data to be classified, as a sample to predict the classification result; predicting the classification result based on the sample and the real The second difference between the class labels adjusts the parameters of the risk classification model.

Wherein, the risk category includes at least one of revenue and expenditure monitoring, budget monitoring, ongoing monitoring and financing monitoring.

Wherein, the enhanced training is performed every first preset time interval, and the target service data is acquired every second preset time interval.

Wherein, the method further includes: storing the risk-related data obtained based on the target business data into the database, and sending the data matching the data call request to the target object in response to the data call request, wherein the risk-related data includes risk business data and non-risk business data.

In order to solve the above technical problems, another technical solution adopted by this application is to provide an electronic device, including a memory and a processor coupled to each other, and the processor is used to execute the program instructions stored in the memory to realize any of the above risk identification method.

In order to solve the above technical problems, another technical solution adopted by the present application is to provide a computer-readable storage medium on which program instructions are stored, and when the program instructions are executed by a processor, any of the above risk identification methods is implemented.

In the above solution, the risk business data is identified from the target business data by using the risk identification model, which is used to generate sample business data for enhanced training of the risk identification model, and the risk classification model is used to determine the risk category of the risk business data, which is used to generate the sample to be Classification data is used to enhance the training of the risk classification model. On the one hand, it increases the number of samples for model training, increases the amount of information learned by the model, and improves the accuracy of the model, thereby improving the accuracy of risk identification. On the other hand, through the model to The processing of target business data excavates the information about financial risks in the target business data and uses it in model training, which increases the amount of information learned by the model and improves the accuracy of the model, thereby improving the accuracy of risk identification.

Description of drawings

Fig. 1 is a schematic flow chart of an embodiment of the risk identification method of the present application;

FIG. 2 is a schematic flow diagram of another embodiment of step S120 of the present application;

FIG. 3 is a schematic flow diagram of another embodiment of step S221 of the present application;

Fig. 4 is a schematic flow chart of another embodiment of the risk identification method of the present application;

Fig. 5 is a schematic flow chart of another embodiment of the risk identification method of the present application;

Fig. 6 is a schematic frame diagram of an embodiment of the electronic device of the present application;

Fig. 7 is a schematic diagram of an embodiment of a computer-readable storage medium of the present application.

Detailed ways

In order to make the purpose, technical solution and effect of the present application more clear and definite, the present application will be further described in detail below with reference to the accompanying drawings and examples. In the following description, for purposes of illustration rather than limitation, specific details, such as specific system architectures, interfaces, and techniques, are set forth in order to provide a thorough understanding of the present application.

The term "and/or" in this article is just an association relationship describing associated objects, which means that there can be three relationships, for example, A and/or B can mean: A exists alone, A and B exist simultaneously, and there exists alone B these three situations. In addition, the character "/" in this article generally indicates that the contextual objects are an "or" relationship. In addition, "many" herein means two or more than two. In addition, the term "at least one" herein means any one of a variety or any combination of at least two of the more, for example, including at least one of A, B, and C, which may mean including from A, Any one or more elements selected from the set formed by B and C.

The risk identification method described in this application can be executed by an electronic device, which can also be referred to as an execution device, and the electronic device can be any device with processing capabilities, such as a mobile phone, a computer, and the like.

Please refer to FIG. 1 . FIG. 1 is a schematic flowchart of an embodiment of a risk identification method of the present application.

Specifically, the method can include:

Step S110: Obtain target service data.

Wherein, the target business data may be business data related to finance and funds, for example, business data about travel reimbursement and the like. In addition, the target business data may include several pieces of business data.

Specifically, the target business data may be business data generated by the financial management system, and the executing device may obtain the target business data from the financial management system.

Step S120: Use the risk identification model to divide the target business data to obtain risky business data and non-risk business data.

It should be noted that the risk identification model can be a pre-trained model, which can be used to identify the risks of the business data input into the model, and determine whether the business data is a risk data type or a non-risk data type, thereby classifying the business data into risk data types. Business data and non-risk business data. Wherein, the risk data type indicates that the business data is judged to have capital risk, and the non-risk data type represents that the business data is judged to have no capital risk.

Specifically, the target business data can include several pieces of business data, the risky business data can include business data identified as risky data types in the target business data, and non-risk business data can include business data identified as non-risk data types in the target business data business data.

Wherein, the risk identification model may be obtained by pre-training using manually marked business data.

Step S130: Use the risk classification model to determine the risk category of the risk business data.

It should be noted that the risk classification model can be a pre-trained model, which can be used to classify the risks of the input business data, determine the risk categories of the business data, and divide the business data into corresponding risk categories. Wherein, the risk category may include several predetermined risk categories. In some embodiments, risk categories include revenue and expenditure monitoring, budget monitoring, ongoing monitoring, and financing monitoring.

Wherein, the risk classification model may be obtained through pre-training using manually labeled business data.

Of course, the risk categories that can be classified by the risk classification model are not limited to the categories in the above embodiments, and the risk categories that can be classified by the risk classification model can be adjusted according to actual application requirements.

Wherein, the risk business data may be output by the risk identification model and obtained by risk identification of the target business data by the risk identification model.

Step S140: Based on the sample business data generated from the risk business data, the risk identification model is enhanced for training, and the risk classification model is enhanced for training based on the sample data to be classified generated from the risk business data of the determined risk category.

Wherein, the risk business data is obtained by risk identification of the target business data by the risk identification model, and is used to generate sample business data, and the sample business data is used for enhanced training of the risk identification model. The risk business data of the determined risk category is obtained by risk classification of the risk business data by the risk classification model, and is used to generate sample data to be classified, and the sample data to be classified is used for enhanced training of the risk classification model.

In the above solution, the risk business data is identified from the target business data by using the risk identification model, which is used to generate sample business data for enhanced training of the risk identification model, and the risk classification model is used to determine the risk category of the risk business data, which is used to generate the sample to be Classification data is used to enhance the training of the risk classification model. On the one hand, it increases the number of samples for model training, increases the amount of information learned by the model, and improves the accuracy of the model, thereby improving the accuracy of risk identification. On the other hand, through the model to The processing of target business data mines the information about financial risks in the target business data from the chaotic target business data, and uses it in the training of the model, which increases the amount of information learned by the model, improves the accuracy of the model, and thus improves risk identification accuracy.

Please refer to FIG. 2 . FIG. 2 is a schematic flowchart of another embodiment of step S120 of the present application. Specifically, step S120 includes:

Step S221: Use the risk identification model to sequentially identify direct risk data, indirect risk data and rare risk data from the target business data.

Among them, the direct risk data refers to the business data directly generated by the financial management system and has financial risks; the indirect risk data refers to the business data indirectly generated by the financial management system and has financial risks; Risky business data.

In a specific application scenario, for suspected repeated payment monitoring, business data that meets the following conditions can be identified as direct risk data: 1. Take the latest transaction records of the current monitoring day as the monitoring object, and query the history within the past 30 days (Including 30 days) Expenditure records consistent with the name, account number, amount, and summary of the payee and payee, monitor the payment records with the number of repeated transactions > 0 (excluding the monitoring object's current transaction); 2. Transaction direction = expenditure; 3. Exclude transaction records that contain the words "handed in", "approved", "actually returned", "collected", and "fund pool member" in the payment summary.

In a specific application scenario, for the bill receivable budget execution business, the business data may include the execution deviation rate (x) of the bill receivable indirectly generated by the financial management system = | the scale of bills collected (actual execution number - secondary schedule number)|/number of secondary schedules*100%, (take 100% when it is greater than 100%; if the denominator is zero, the numerator is not zero, and the deviation rate is 100%; if the denominator is zero, the numerator is also zero, and the deviation rate is 0 %;), the business data can be identified as indirect risk data if it satisfies the risk judgment conditions.

In a specific application scenario, for the abnormal monitoring of account status, business data that meets the following conditions can be identified as rare risk data: 1. The bank account's "attachment status = unattached" and "whether the transaction cannot be attached" Filing = No"; 2. "Monitoring status = Unauthorized" of the bank account and "Whether the record cannot be monitored = No".

Step S222: Take direct risk data, indirect risk data and rare risk data together as risk business data, and use part of the business data in the target business data except direct risk data, indirect risk data and rare risk data as non-risk business data.

Among them, direct risk data, indirect risk data and rare risk data are all business data with financial risks, and they are regarded as risk business data together. Part of the business data in the target business data other than the above three, which means that it is judged not to belong to the above three risk data, and there is no capital risk, can be regarded as non-risk business data.

In some embodiments, during the risk identification process of the risk identification model, the types of risk data included in the risk business data may also be adjusted according to actual application needs. Exemplarily, direct risk data and indirect risk data are sequentially identified from target business data by using a risk identification model. Alternatively, the types of risk data included in the risk business data may also be increased.

Please refer to FIG. 3 , which is a schematic flowchart of another embodiment of step S221 of the present application. Specifically, step S221 includes:

Step S3211: Obtain the first probability that the target business data belongs to the direct risk data.

Specifically, different types of risk data are identified in different ways, and direct risk data, indirect risk data, and rare risk data are identified in different ways. First identify the direct risk data from the target business data, then identify the indirect risk data in the target business data except for the direct risk data, and finally identify the target business data except for the direct risk data and indirect risk data. Rare risk data is identified to distinguish target business data into risk business data and non-risk business data.

In some embodiments, the first probability that each piece of business data belongs to direct risk data can be calculated by the following formula:

Among them, Pr(S/W) indicates the first probability that this piece of business data belongs to direct risk data, Pr(S) indicates the overall probability that any business data is risk data, and Pr(W/S) indicates that "risk data in business data "The overall probability of appearing in risk data, Pr(H) represents the probability that any business data is not risk data, Pr(W/H) represents the probability of risk data appearing in business data, Pr(S), Pr(W/ S), Pr(H), and Pr(W/H) are calculated based on target business data.

Step S3212: Determine direct risk data from target business data based on the first probability.

Specifically, it may be determined whether the piece of business data belongs to direct risk data according to whether the first probability of the business data satisfies the first preset requirement.

In a specific application scenario, the first preset condition may be reaching a preset first probability threshold, so that business data whose first probability reaches the first probability threshold is taken as direct risk data. Of course, the first preset condition can be adjusted according to actual application needs.

Step S3213: Obtain the second probability that part of the business data in the target business data except the direct risk data belongs to the indirect risk data.

Wherein, business data may include several sub-data, and for a piece of business data, whether it belongs to indirect risk data may be associated with the risk probability of its sub-data.

In some embodiments, the second probability that each piece of business data in the target business data other than the direct risk data belongs to the indirect risk data can be calculated by the following formula:

Among them, p represents the second probability that the business data is indirect risk data, p ₁ represents the p(S/W ₁ ) probability that the first sub-data of the business data corresponds to the risk data, and p ₂ represents the second probability of the business data The two pieces of sub-data correspond to the p(S/W ₂ ) probability of the risk data, ..., p _n indicates that the nth piece of sub-data of the business data corresponds to the p(S/W _N ) probability of the risk data. p ₁ , p ₂ , ... p _n are obtained based on the sub-data corresponding to the piece of business data.

Step S3214: Based on the second probability, determine the indirect risk data from part of the target business data except the direct risk data.

Specifically, it may be determined whether the piece of business data belongs to indirect risk data according to whether the second probability of the business data meets the second preset requirement.

In a specific application scenario, the second preset condition may be reaching a preset second probability threshold, so that business data whose second probability reaches the second probability threshold is taken as indirect risk data. Of course, the second preset condition can be adjusted according to actual application needs.

Step S3215: Obtain the third probability that part of the business data in the target business data except direct risk data and indirect risk data belongs to rare risk data.

It should be noted that rare risk data refers to data with a relatively low frequency of occurrence and financial risk. The third probability of each item of business data in the target business data except direct risk data and indirect risk data can be calculated by the following formula:

Among other things, the above formula can be extended to the case where n is equal to zero (no junkness defined), and evaluates to PrS in this case.

Among them, P'r(S/W) represents the third probability that business data is rare risk data, and can also be understood as the correction probability of business data (that is, the probability of updating calculations based on the first probability), and s represents the relevant input The strength of background information about risk data, Pr(S) indicates the probability that any input data is risk data, n indicates the number of times this data appears in the learning phase, Pr(S/W) indicates that the business data belongs to direct risk data first probability.

In some embodiments, PrS can be equal to 0.5 again, so as to avoid too much suspicion of the input service data. 3 is a good value for s, meaning that the learned corpus must contain 3 more information than the business information, put more confidence in the risky value than in the default value.

Step S3216: Based on the third probability, determine rare risk data from part of the target business data except direct risk data and indirect risk data.

Specifically, it may be determined whether the piece of business data belongs to rare risk data according to whether the third probability of the business data satisfies the third preset requirement.

In a specific application scenario, the third preset condition may be reaching a preset third probability threshold, so that business data whose third probability reaches the third probability threshold is regarded as rare risk data. Of course, the third preset condition can be adjusted according to actual application needs.

Please refer to FIG. 4 . FIG. 4 is a schematic flowchart of another embodiment of the risk identification method of the present application.

Specifically, the method may include the following steps:

Step S410: Obtain target service data.

In some embodiments, the executing device may acquire target business data at intervals of a second preset time, and further, acquire target business data from the financial management system at intervals of a second preset time.

In a specific application scenario, the second preset duration may be 5 minutes. Of course, the second preset duration can also be evaluated and adjusted according to the actual service data volume, for example, the second preset duration can also be a preset duration.

In some embodiments, acquisition of target business data may also be triggered according to the amount of unprocessed business data generated in the financial management system. For example, when the amount of unprocessed business data reaches a first preset amount, execute The device acquires unprocessed business data as target business data.

In some embodiments, after the target business data is acquired, the target business data may also be preprocessed, and the preprocessing may include but not limited to desensitization, cleaning, and the like.

In some embodiments, after the target business data is acquired, the target business data may also be stored in a database in the executing device for subsequent calling.

By periodically acquiring business data generated by the financial management system for risk identification and risk classification, the target business data acquired each time may include business data of ongoing business and business that has been completed data. Exemplarily, for example, A's reimbursement is in the process of reimbursement, and B's reimbursement has been completed. Risk identification and risk classification are carried out on the business data of the completed business, which can realize the assessment of the capital risk that has occurred. The risk identification and risk classification of the business data of the ongoing business can realize the prediction of the upcoming capital risk. Therefore, through the above method, it is possible to identify the capital risk that has already occurred and the capital risk that is about to occur, so as to achieve full coverage of capital risk identification and accurately and comprehensively identify capital risk.

Step S420: Use the risk identification model to divide the target business data to obtain risky business data and non-risk business data.

In a specific application scenario, input the data set of target business data into the risk identification model, carry out risk identification on each piece of business data in the model, and obtain the risk type prediction result. The risk type prediction result indicates that the business data belongs to risk data Which of type and non-risk data type. Based on the risk type prediction results, the risky business data can be divided into two data sets: risky business data and non-risky business data.

In some embodiments, the execution device can store the two data sets of risk business data and non-risk business data processed by the risk identification model into the database for subsequent use, for example, for subsequent model calls.

Step S430: Determine the detailed category of risk business data.

Wherein, the risk business data includes several business data, specifically, the above step may be to determine which preset detailed category each piece of business data belongs to. The number of detail categories is greater than the number of risk categories.

In some embodiments, detailed categories may include but not limited to the following preset categories: inefficient account monitoring, over-standard account monitoring, unpaid account monitoring, off-book account monitoring, electricity account abnormal expenditure monitoring, account status abnormal monitoring, electricity account Unreconciled data monitoring, bank account validity monitoring, electronic payment monitoring, suspected double payment monitoring, reserve fund balance monitoring, external lending fund monitoring, abnormal payable balance monitoring, large-value account transfer monitoring, check refund monitoring, payment process return Refund monitoring, payment document transmission timeliness monitoring, large-amount private payment monitoring, "revenue and expenditure balance" monitoring, large-amount cash receipts and payments monitoring, large-amount cash balance monitoring, internal closed settlement monitoring, abnormal two-way transaction of funds monitoring, special payment Monitoring, MAC address duplicate interception and return, supplier blacklist interception and return, signature process mismatch interception and return, external independent account large-value payment interception, external supplier large-value payment interception, private large-value payment interception warning, abnormal time payment instruction Interception early warning, super large payment interception early warning, loan repayment timeliness monitoring, financing ledger normative monitoring, bill redemption timeliness monitoring, newly added bill receivable unit monitoring, bill ledger normative monitoring, bill receivable budget execution accuracy , Accuracy of bills payable budget execution, bill ledger - normative monitoring of bill date, acceptance bank normative monitoring, non-bank acceptance bill monitoring, non-electronic bill monitoring.

Among them, the risks indicated by each detailed category are different, as follows:

Low-efficiency account monitoring: monitor accounts with account transactions less than a certain standard under different account classifications; over-standard account monitoring: monitor the number of accounts with bank account openings exceeding the account classification and control standards under each account classification of each unit; unreached accounts Payment monitoring: monitoring the unpaid account balance of each unit at the end of last month; off-book account monitoring: monitoring the transaction records of fund income and expenditure transactions, the account name is the unit in the system, but the account number is not included in the monitoring scope according to the regulations; the electricity account is abnormal Expenditure monitoring: monitoring the transaction records of capital expenditure behaviors other than fund handover and handling fees in the electricity account; bank account validity monitoring: checking all accounts of the previous day through the balance of income and expenditure rules, analyzing and judging the status of bank accounts ( The balance at the beginning of the previous day + income - expenditure is compared with the balance at the end of the previous day. If they are inconsistent, the account difference will be displayed); the real-time monitoring of the three-party reconciliation of the electricity account: monitoring the result of the three-party reconciliation of the electricity account of the capital center, there are unclear accounts Records of at least one of the payment, unsettled accounts or outstanding accounts; electronic payment monitoring: monitor the electronic payment ratio and non-electronic payment status of each unit; suspected double payment monitoring: monitor the payment units and account numbers within the past 30 days, Payment records with the same payment amount and payment summary; reserve fund balance monitoring: monitor the details of the reserve fund stock of each unit; external lending funds monitoring: monitor the detailed payment records of external lending funds; abnormal balance of payables monitoring: monitor external lending Detailed payment records of funds; large-amount transfer account monitoring: monitor records of single transfer amount > 10 million yuan or monthly accumulative transfer amount > 20 million yuan from Zhongdiancai account of each unit to external bank account; check refund monitoring: monitor fund payment refund Payment process rollback monitoring: monitor the rollback of centralized payment business; payment document transmission timeliness monitoring: monitor the records that the centralized payment delivery center time and rolling schedule date are more than 3 days; large-amount private payment monitoring: monitor Private payment records with a single transaction amount of more than 5,000 yuan in the fund account; "receipt and expenditure balance" monitoring: monitor the income, expenditure and balance of the bank account reflected in the transaction details of the financial institution, and the income, expenditure and balance reflected in the accounting books. Records; monitoring of large cash receipts and expenditures: monitor units with a monthly accumulative inventory cash account with a total value of more than 5,000 yuan from debit and credit; monitoring of large cash balances: monitor units with large cash balances at the end of the month; internal closed settlement monitoring: monitor each The unit did not implement internal closed settlement; monitoring of abnormal two-way transaction of funds: monitoring of false transactions of each unit (temporary lending, bridge funds); monitoring of special payments: monitoring of transaction summaries containing "entertainment", "cigarette", and "wine" , "tea", "gift", "reception" and other characters of special payment records; MAC address duplicate interception and return: monitor MAC address duplicate interception and return; supplier blacklist interception and return: monitor supplier blacklist interception and return; signature Process mismatch interception and return: monitoring of signature process mismatch interception and return; external independent account large-value payment interception: monitoring of external independent account large-value payment interception and early warning; external supplier large-value payment interception: monitoring external supplier large-value payment interception Early warning of payment interception; early warning of interception of large-amount private payment: monitoring of early warning of interception of large-amount private payment; early warning of payment instruction interception at abnormal time: monitoring of early warning of payment instruction interception at abnormal time; early warning of interception of super-large payment: monitoring of super-large payment Implementation of the implementation of the processing situation; loan repayment timeliness monitoring: monitoring the existence of overdue loan repayment records; financing ledger normative monitoring: monitoring the differences between the ending balance and book balance of each unit's financing ledger; bill redemption timeliness monitoring: should Bills received unpaid (monitor records of bills that have expired before the cut-off date of inquiry and bills receivable have not been processed), bills payable have not been paid (monitored records of bills that have expired before the cut-off date of inquiry and bills payable have not been processed); new Add bill receivable unit monitoring: monitor the unit that has newly collected bills; add bill receivable unit monitoring: monitor the unit that has newly collected bills; bill ledger normative monitoring: monitor the bills receivable, payable There is a discrepancy between the face value of the accumulative ledger and the balance of the accumulative bills; the execution accuracy of the bills receivable budget: monitor the deviation between the monthly budget of the bills receivable by each unit at the end of last month and the actual execution amount; the execution accuracy of the bills payable budget: monitor The deviation between the monthly budget and actual execution of bills payable by each unit at the end of last month; standard monitoring of bill ledger-bill issuance date: monitor bill ledger-bill ledger registration information inconsistent with the bill issuance date; acceptance bank normative Monitoring: monitoring bill ledger - the acceptor does not include bill records in the fields of "Agricultural Bank of China", "Industrial and Commercial Bank of China", "Construction Bank" and "Bank of China"; monitoring of non-bank acceptance bills: monitoring bill ledger - monitoring bill type does not Bill records of "bank acceptance bills"; non-electronic bill monitoring: monitor bill ledger - bill records whose bill medium is not "electronic bills".

It should be noted that the risks suggested by different detailed categories are different, but the risks suggested by the same detailed category have their commonalities. The classification of detailed categories is to further mine the information about risk identification contained in the target business data on the basis of risk identification. The way. In some embodiments, other processing methods may be used to replace the classification of detailed categories to mine the information about risk identification contained in the target business data.

In some embodiments, the execution device may classify the risk business data according to the detailed category and store it as risk detailed data in the database for subsequent use.

Step S440: Using the risk classification model to determine the risk category of the risk business data whose detailed category has been determined.

For related descriptions of step S440, reference may be made to related content about step S130 in the foregoing embodiments.

Further, use the risk classification model to map the risk business data of the identified detailed categories to the classification function dimension, and determine the separation hyperplane corresponding to each risk category; in the classification function dimension, based on the separation hyperplane, the risks Business data is divided into corresponding risk categories. In this way, risk business data with detailed categories can be obtained.

In a specific application scenario, each item of business data in the risk business data whose detailed category has been determined has already determined the detailed category. The risk business data whose detailed categories have been determined can be further divided into several major risk categories. Exemplarily, detailed category inefficient account monitoring, over-standard account monitoring, outstanding account monitoring, off-book account monitoring, electricity account abnormal expenditure monitoring, account status abnormal monitoring, electricity account unreconciled data monitoring, bank account validity monitoring Real-time monitoring of tripartite reconciliation of electricity account can be identified as the risk category of account monitoring. Detailed category electronic payment monitoring, suspected double payment monitoring, reserve fund balance monitoring, external loan fund monitoring, payable balance abnormal monitoring, large-amount transfer account monitoring, check refund monitoring, payment process rollback monitoring, payment document transfer timeliness monitoring, Large-amount private payment monitoring, "receipt and expenditure surplus" monitoring, large-amount cash receipt and expenditure monitoring, large-amount cash balance monitoring, internal closed settlement monitoring, abnormal two-way transaction of funds monitoring, and special payment monitoring can be determined as the monitoring of income and expenditure. risk category. Detailed category MAC address duplicate interception and return, supplier blacklist interception and return, signature process mismatch interception and return, external independent account large-value payment interception, external supplier large-value payment interception, private large-value payment interception warning, abnormal time payment instruction Interception early warning and super large payment interception early warning can be identified as the risk category of in-process monitoring. Detailed category loan repayment timeliness monitoring, financing ledger normative monitoring, bill redemption timeliness monitoring, newly added bill receivable unit monitoring, bill ledger normative monitoring, bill receivable budget execution accuracy, bill payable budget execution accuracy , Bill ledger-regular monitoring of bill date, normative monitoring of acceptance banks, monitoring of non-bank acceptance bills, and monitoring of non-electronic bills can be identified as the risk category of financing monitoring.

Through the setting of the above detailed categories and risk categories, it is possible to identify possible capital risks in the whole process of capital operation and improve the accuracy and comprehensiveness of risk prediction.

In some embodiments, different feature conversion functions may be used for risky business data of different detailed categories, that is, different ways may be used to map risky business data to classification function dimensions.

Step S450: Based on the sample business data generated from the risky business data, perform enhanced training on the risk identification model, and based on the sample data to be classified generated from the risky business data with identified risk categories, perform enhanced training on the risk classification model.

It should be noted that the enhanced training of the risk identification model and the enhanced training of the risk classification model may be independent of each other. Augmentation training for both models can be performed multiple times.

In some embodiments, the executing device may execute the related steps of the enhanced training every first preset time interval.

In some embodiments, the execution device may also repeatedly execute the related steps of enhancing training of the risk identification model and enhancing training of the risk classification model at different time intervals.

By continuously obtaining the target business data, it is used to generate sample business data and sample data to be classified, so as to periodically carry out enhanced training of the two models, and realize the rolling update of the training data. On the one hand, the number of samples for model training is significantly expanded. Increase the amount of information learned by the model and improve the accuracy of the model, thereby improving the accuracy of risk identification. On the other hand, through the rolling update of data, the model is continuously strengthened to learn and improve the accuracy of the model, thereby improving the accuracy of risk identification sex.

Specifically, the execution device can also generate sample business data by using the risk business data of the determined detailed category, wherein the sample business data is marked with a real risk label, and the real risk label indicates which of the risk data types and non-risk data types the sample business data belongs to. Type, the sample business data is used to enhance the training of the risk identification model.

In a specific application scenario, since the sample business data is generated based on the risk business data of the specified category, the business data contained in the sample business data comes from the data set of risk business data, so the data from the risk business data The real risk label corresponding to the business data indicates that the business data belongs to the risk data type.

In some embodiments, it is also possible to generate sample business data by using the risk business data of the determined detailed category together with other business data. Exemplarily, other business data may be sample data previously used in the training process of the risk identification model.

Specifically, the execution device can use the risk identification model to divide the sample business data, obtain the sample risk type prediction result, and determine the sample risk data and sample non-risk data based on the sample risk type prediction result, wherein the sample risk prediction result indicates that the sample business data belongs to Which one of the risk data type and the non-risk data type, and then adjust the parameters of the risk identification model based on the first difference between the risk prediction result and the real risk label.

Wherein, for the sample risk type, the predicted result is the sample business data belonging to the risk data type as the sample risk data, and for the sample risk type, the predicted result is the sample business data belonging to the non-risk data type as the sample non-risk data. The prediction result of the sample risk type is the result of using the risk identification model to predict whether the sample business data belongs to the risk data type or the non-risk data type, while the real risk label is predetermined, whether the sample business data belongs to the risk data type or the non-risk data type The real label of the type, comparing the two to get the first difference can be used to adjust the parameters of the risk identification model, thereby improving the accuracy of the risk identification model.

In some embodiments, since the sample business data is generated based on the risk business data of the specified category, the real risk label can also contain information about the category, and the risk identification model can also be set to be able to learn the details contained in the real risk label. categories of information to improve the accuracy of risk identification models.

Specifically, the execution device can also generate sample data to be classified by using the risk business data of the determined risk category, wherein the sample data to be classified is marked with a real classification label, and the real classification label is used to represent the risk category described in the sample risk data.

In some embodiments, the risk business data of the determined risk category may also be used together with other business data to generate sample data to be classified. Exemplarily, other business data may be sample data previously used in the training process of the risk classification model.

Specifically, the execution device may use the risk classification model to determine the risk category of the sample to-be-classified data as a predicted classification result of the sample, and adjust the parameters of the risk classification model based on the second difference between the predicted classification result of the sample and the real category label.

Among them, the sample prediction classification result is the result of using the risk classification model to predict the risk category of the sample data to be classified, and the real category label is the real label of the risk category to which the sample data to be classified belongs. Differences can adjust the parameters of the risk classification model to improve the accuracy of the risk classification model.

In a specific application scenario, the training process of the risk classification model includes: first determine the following formula:

f(x)＝sign(w ^*T Φ(x)+b ^* )

Among them, Φ(x) represents the feature transformation function of the space, x represents the sample data to be classified, including n pieces of business data, and y represents the data set after risk classification corresponding to x, including n pieces of business data. Sign() represents the sign function. Secondly, α _i ^* (the optimal solution of α _i ) is solved based on the following formula:

α _i ≥0,i=1,2,…,n

After that, solve w*, b* (the optimal solution of w, b) based on the following formula:

Finally, the separation hyperplane is calculated as follows:

w ^* Φ(x)+b ^* ＝0

Then, the separation hyperplane can be used to obtain the sample data to be classified according to the predicted risk category, compare the result with the real category label labeled by the sample data to be classified, and use the difference between the two to adjust the parameters of the risk classification model.

In some embodiments, the execution device may also store the risk-related data obtained based on the target business data in the database, so as to respond to the data call request and send the data matching the data call request to the target object. Wherein, the data call request is sent by the target object, and the risk-related data includes risk business data and non-risk business data.

In some embodiments, the risk-related data may also include risk business data classified according to risk categories.

Please refer to FIG. 5 . FIG. 5 is a schematic flowchart of another embodiment of the risk identification method of the present application.

In this embodiment, the execution device can run a risk monitoring system, and the execution device includes two databases, wherein the first database can be used to store target business data obtained from the production environment (financial management system). The second database can be used to store risk-related data, as well as sample business data required for enhanced training, sample data to be classified, and the like.

The execution device can communicate with the devices in the production environment, obtain business data generated in the production environment, and store them in the first database, and then the risk monitoring system can obtain target business data from the first database. The risk monitoring system can identify the risky business data in the target business data and determine its risk category through the risk identification model and the risk classification model. The data processed by the risk identification model and the risk classification model can be stored in the second database as risk-related data for subsequent enhanced training. The DB data (data set) is used for enhanced training to complete the correction of the model. The terminal user can communicate with the execution device through the client running on the terminal device, and send a data call request to the execution device through the client, so that the execution device will feed back the data stored in the second database to the client, so that the client can Relevant data is displayed, for example, business data of each risk category is displayed.

Please refer to FIG. 6 . FIG. 6 is a schematic frame diagram of an embodiment of the electronic device of the present application.

In this embodiment, the electronic device 60 includes a memory 61 and a processor 62 , wherein the memory 61 is coupled to the processor 62 . Specifically, various components of the electronic device 60 may be coupled together through a bus, or the processor 62 of the electronic device 60 may be connected to other components one by one. The electronic device 60 may be any device with processing capability, such as a computer, a tablet computer, a mobile phone, and the like.

The memory 61 is used to store program instructions executed by the processor 62 and data during processing by the processor 62 . For example, risk business data, sample business data, etc. Wherein, the memory 61 includes a non-volatile storage part for storing the above-mentioned program instructions.

The processor 62 controls the operation of the electronic device 60, and the processor 62 may also be referred to as a CPU (Central Processing Unit, central processing unit). The processor 62 may be an integrated circuit chip with signal processing capability. The processor 62 can also be a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components . A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. In addition, the processor 62 may be jointly implemented by multiple circuit chips.

The processor 62 invokes the program instructions stored in the memory 61 to execute the instructions to implement any of the above risk identification methods.

Please refer to FIG. 7 . FIG. 7 is a schematic frame diagram of an embodiment of a computer-readable storage medium of the present application.

In this embodiment, the computer-readable storage medium 70 stores program instructions 71 executable by the processor, and the program instructions 71 can be executed to implement any of the risk identification methods described above.

The computer-readable storage medium 70 can specifically be a medium that can store program instructions, such as a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk. , or it can also be a server storing the program instructions, and the server can send the stored program instructions to other devices to run, or can also run the stored program instructions by itself.

In some embodiments, the computer-readable storage medium 70 may also be a memory as shown in FIG. 6 .

The above is only the implementation of the application, and does not limit the patent scope of the application. Any equivalent structure or equivalent process conversion made by using the specification and drawings of the application, or directly or indirectly used in other related technologies fields, are all included in the scope of patent protection of this application in the same way.

Claims (10)

1. A risk identification method, characterized in that the method comprises: Obtain target business data; Using a risk identification model to divide the target business data to obtain risk business data and non-risk business data; Using a risk classification model to determine the risk category of the risk business data; Perform enhanced training on the risk identification model based on the sample business data generated from the risk business data, and classify the risk based on the sample data to be classified generated from the risk business data for which the risk category has been determined The model is enhanced for training. 2. The method according to claim 1, wherein said dividing said target business data using a risk identification model to obtain risk business data and non-risk business data comprises: Using the risk identification model to sequentially identify direct risk data, indirect risk data and rare risk data from the target business data; taking the direct risk data, the indirect risk data and the rare risk data together as the risk business data, and dividing the target business data from the direct risk data, the indirect risk data and the Part of the business data other than the rare risk data is used as the non-risk business data. 3. The method according to claim 2, characterized in that, using the risk identification model to sequentially identify direct risk data, indirect risk data and rare risk data from the target business data comprises: Obtaining a first probability that the target business data belongs to the direct risk data; determining said immediate risk data from said target business data based on said first probability; Obtaining a second probability that part of the business data in the target business data other than the direct risk data belongs to the indirect risk data; determining the indirect risk data from part of the target business data except the direct risk data based on the second probability; Obtaining a third probability that part of the target business data other than the direct risk data and the indirect risk data belongs to the rare risk data; The rare risk data is determined from part of the target service data except the direct risk data and the indirect risk data based on the third probability. 4. The method according to claim 1, characterized in that, after said using a risk identification model to divide said target business data to obtain risky business data and non-risk business data, said using a risk classification model to determine said risky business data Before the risk category of , the method also includes: determining the detailed categories of the risk business data, wherein the number of detailed categories is greater than the number of risk categories; The determining the risk category of the risk business data by using the risk classification model includes: Using the risk classification model to map the risk business data of which the detailed category has been determined to a classification function dimension, and determine the separation hyperplane corresponding to each of the risk categories; In the classification function dimension, the risk business data of which the detailed categories have been determined are divided into corresponding risk categories based on the separating hyperplane. 5. method according to claim 4, is characterized in that, described method also comprises: The sample business data is generated by using the risk business data of which the detailed category has been determined, wherein the sample business data is marked with a real risk label, and the real risk label indicates that the sample business data belongs to the risk data type and non-risk data. Which of the risk data types; The performing enhanced training on the risk identification model based on the sample business data generated by the risk business data includes: Using the risk identification model to divide the sample business data to obtain a sample risk type prediction result, determine sample risk data and sample non-risk data based on the sample risk type prediction result, and the sample risk prediction result represents and predicts the sample business Which of the risk data type and non-risk data type does the data belong to; Adjusting parameters of the risk identification model based on a first difference between the risk prediction result and the true risk label. 6. The method according to claim 4, characterized in that the method further comprises: The sample data to be classified is generated by using the risk business data of which the risk category has been determined, wherein the sample data to be classified is marked with a real category label, and the real category label represents all the samples to which the sample data to be classified belongs. the above risk categories; The enhanced training of the risk classification model based on the sample data to be classified generated from the risk business data of which the risk category has been determined includes: Using the risk classification model to determine the risk category of the sample data to be classified as a sample prediction classification result; Adjusting parameters of the risk classification model based on a second difference between the sample predicted classification result and the true class label. 7. The method according to claim 1, wherein the risk category includes at least one of revenue and expenditure monitoring, budget monitoring, ongoing monitoring and financing monitoring; And/or, the enhanced training is performed every first preset time period, and the target service data is acquired every second preset time period. 8. The method according to claim 1, further comprising: storing the risk-related data obtained based on the target business data in a database, and sending data matching the data call request to the target object in response to the data call request, wherein the risk-related data includes the risk business data and the non-risk business data. 9. An electronic device, characterized in that it comprises a memory and a processor coupled to each other, and the processor is configured to execute the program instructions stored in the memory, so as to realize the method described in any one of claims 1 to 8. risk identification method. 10. A computer-readable storage medium, on which program instructions are stored, wherein, when the program instructions are executed by a processor, the risk identification method according to any one of claims 1 to 8 is implemented.