Credit line dynamic configuration method and device based on merchant cashier data, equipment and medium

Abstract

The invention discloses a credit line dynamic configuration method and device based on merchant cashier data, equipment and a medium, relates to the technical field of computers, is applied to a bank credit system, and comprises the following steps: collecting transaction data, inventory data, member data, industry data and credit investigation data of merchants in a software-as-a-service cashier system in real time; triggering a timestamp of the data and a device fingerprint recording operation; performing verification, desensitization, standardization processing and abnormal data backflow cleaning on the obtained merchant cashier data acquisition result based on a data governance strategy, and storing the governed data to a data lake; constructing and training a dynamic credit extension model based on the data lake, a plurality of predefined interpretable factors and a preset model self-training rule; and when a money use application initiated by the merchant is received, adjusting the original credit line of the merchant based on the current dynamic credit model, and outputting a target credit line. The financing availability of small and micro merchants can be greatly improved, and business cooperation among banks, merchants and cashier system providers is promoted.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a credit limit dynamic configuration method and device based on merchant cash register data, equipment and medium. BACKGROUND

[0002] At present, the availability of financing of small and micro merchants has been limited by the traditional credit granting mode for a long time. In this mode, banks mainly rely on static financial reports, collateral and manual due diligence to assess the credit risk of merchants, determine the credit limit, and there are problems such as high due diligence cost, low credit coverage, and lagging risk control, which cannot adapt to the rapid changes in the market and cannot meet the growing operating capital needs of small and micro merchants, limiting the in-depth development of inclusive finance.

[0003] At the same time, the homogenization of competition in the acquirer market is intensifying, and banks urgently need high-stickiness scenarios and real-time operating data to improve the comprehensive value of merchants and build a perfect transaction financial ecosystem, but the traditional credit granting mode has been difficult to adapt to this demand. On the other hand, although the merchant cash register SaaS (Software as a Service) service independently developed and promoted by commercial banks has been rapidly popularized and provides merchants with all-round services including transaction flow, inventory management, and member marketing, these high-dimensional and high-frequency data lack effective coupling with credit and risk control models and have not been fully utilized in the field of small and micro merchant financing. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a credit limit dynamic configuration method and device based on merchant cash register data, which can solve the problems of high due diligence cost, low credit coverage, and lagging risk control in the traditional credit granting mode, greatly improve the availability of financing of small and micro merchants, and promote business cooperation between banks, merchants, and cash register system providers, thereby realizing "data as credit" inclusive credit without collateral and manual due diligence, reducing due diligence cost, and realizing dynamic synchronization of credit granting, merchant operation, and risk control, thereby improving user experience. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a credit limit dynamic configuration method based on merchant cash register data, applied to a bank credit system, comprising:

[0006] Based on the multi-thread concurrent strategy and the standardized interface, the transaction data, inventory data, member data, industry data, and credit investigation data of the merchant in the Software as a Service cash register system are collected in real time, and the timestamp and device fingerprint recording operation of the data is triggered to obtain the merchant cash register data collection result;

[0007] The data in the merchant cash register data collection result is verified, desensitized, standardized, and abnormal data is backwashed based on a data governance strategy, and the corresponding governed data is stored in a data lake;

[0008] Based on the data lake, a plurality of interpretable factors and a preset model self-training rule, a dynamic credit model is constructed and periodically trained; the plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight, and a risk adjustment factor;

[0009] When receiving the cash application initiated by the merchant, the original credit limit corresponding to the merchant is adjusted based on the current dynamic credit model, the target credit limit corresponding to the cash application is determined and output, so as to complete the corresponding credit process based on the target credit limit.

[0010] Optionally, based on the multi-thread concurrent strategy and the standardized interface, the transaction data, inventory data, member data, industry data and credit data corresponding to the merchant in the software as a service cash register system are collected in real time, including:

[0011] Based on the standardized interface and the software as a service cash register system, a connection is established;

[0012] Based on the multi-thread concurrent strategy, the open authorization mechanism and the hypertext transfer protocol version 2, the transaction data, inventory data, member data, industry data and credit data corresponding to the merchant in the software as a service cash register system are collected in real time;

[0013] The transaction data includes the amount, time, payment method and location of the transaction; the inventory data includes the inventory quantity unit code, quantity, warehouse in-out time and turnover rate of the inventory; the member data includes the level, consumption frequency, amount and right usage of the member; the industry data includes the industry code and industry sentiment index; the credit data includes the credit score, historical overdue data and litigation data.

[0014] Optionally, the timestamp and device fingerprint record operation of the trigger data are recorded to obtain the merchant cash register data collection result, including:

[0015] For any item of data collected, based on the coordinated universal time format and the data generation time corresponding to the current item of data, the target timestamp corresponding to the current item of data is determined and recorded;

[0016] By collecting the hardware features and software features of the corresponding device, and combining a preset encryption algorithm, the device fingerprint information corresponding to the current item of data is determined and recorded; the hardware features include device model, media access control address and operating system version; the software features include browser fingerprint and application version number;

[0017] determine a target data collection result based on the collected data and the target time stamp and the device fingerprint information corresponding to the data;

[0018] In the process of returning the target data collection result, the target data collection result is compressed based on a preset compression technique, and the corresponding compressed data is transmitted to a ring buffer of a local data receiving end using a preset encryption protocol and a breakpoint resume mechanism, so that when the compressed data is checked in the ring buffer, the merchant cash register data collection result is determined.

[0019] Optionally, the data in the merchant cash register data collection result is checked, desensitized, standardized and cleaned up by abnormal data return based on the data governance strategy, including:

[0020] For different data types and different business scenarios in the merchant cash register data collection result, the rule engine is used for integrity check, consistency check, data format check, data type check and value range check respectively to determine the check result.

[0021] For the first data in the merchant cash register data collection result represented by the check result that passes the check, sensitive data recognition is performed to determine the sensitive data recognition result.

[0022] When the sensitive data recognition result indicates that there is preset sensitive data, the data sensitivity and data use scenario are analyzed based on the sensitive data recognition result, and the target desensitization strategy is determined using the corresponding analysis result; the data use scenario includes development test scenario and data analysis scenario.

[0023] Desensitization is performed based on the target desensitization strategy, the analysis result and the sensitive data recognition result to determine the desensitized data.

[0024] The desensitized data is standardized to determine the standardized data.

[0025] For the abnormal sensitive data in the merchant cash register data collection result or the second data represented by the check result that does not pass the check, corresponding data collection operation is performed again using the software as a service cash register system.

[0026] Optionally, the dynamic credit model is constructed and periodically trained based on the data lake, a plurality of predefined interpretable factors and a preset model self-training rule, including:

[0027] A plurality of predefined interpretable factors and target formulas corresponding to each of the interpretable factors are obtained.

[0028] The target formula, a preset machine learning algorithm, and the data lake are combined to construct a dynamic credit model, with each of the interpretable factors as an independent variable and the merchant credit limit as a dependent variable.

[0029] Based on a preset time interval, it is determined whether to trigger model training. When yes, based on a rolling window, the latest merchant POS data is obtained from the data lake, and the dynamic credit model is trained and verified using the latest merchant POS data, and the factor weights corresponding to each of the interpretable factors are updated to determine the updated dynamic credit model.

[0030] During the construction and training of the dynamic credit model, the transaction data is monitored through a preset fuse mechanism and a preset threshold, and an abnormal event is identified according to the data monitoring result to determine an abnormal event identification result.

[0031] When the abnormal event identification result indicates that there is a preset abnormal event, based on the target merchant, event type and severity corresponding to the preset abnormal event in the abnormal event identification result, a risk adjustment factor decrement operation is triggered to obtain a decrement result.

[0032] Based on the decrement result, it is determined whether to trigger a credit limit freezing operation corresponding to the target merchant.

[0033] Optionally, when the merchant initiates a cash application, the original credit limit corresponding to the merchant is adjusted based on the current dynamic credit model, including:

[0034] When the merchant initiates a cash application through the POS terminal of the software as a service POS system, it is determined whether the merchant belongs to the inventory of customers who have been granted credit to determine a merchant determination result.

[0035] If the merchant determination result is yes, the original credit limit corresponding to the merchant is retrieved.

[0036] If the merchant determination result is no, the basic credit limit corresponding to the merchant is determined based on a preset credit model, and the basic credit limit is taken as the original credit limit.

[0037] Based on the current dynamic credit model and the data lake, the additional credit amount corresponding to the merchant is determined.

[0038] Based on the additional credit amount, the original credit limit is increased, the target credit limit corresponding to the cash application is determined and output, and the corresponding credit process is completed based on the target credit limit.

[0039] Optionally, after the target credit limit is used to complete the corresponding credit process, the method further includes:

[0040] starting a post-loan monitoring mechanism;

[0041] collecting business data of the merchant based on the post-loan monitoring mechanism and the software-as-a-service cash register system to determine a data collection result; the business data includes the transaction data, the inventory data, the member data, the industry data, and the credit investigation data;

[0042] monitoring and analyzing data in the data collection result by using a preset big data analysis algorithm to determine a data analysis result;

[0043] based on the data analysis result, performing risk identification, and using a corresponding risk identification result to determine whether to trigger a risk control operation.

[0044] In a second aspect, the application provides a credit limit dynamic configuration device based on merchant cash register data, applied to a bank credit system, including:

[0045] a data acquisition module, configured to acquire transaction data, inventory data, member data, industry data, and credit investigation data of a merchant in a software-as-a-service cash register system in real time based on a multi-thread concurrent strategy and a standardized interface, trigger a timestamp and a device fingerprint recording operation, and obtain a merchant cash register data acquisition result;

[0046] a data governance module, configured to verify, desensitize, standardize, and clean up abnormal data of data in the merchant cash register data acquisition result based on a data governance strategy, and store the corresponding governed data to a data lake;

[0047] a model self-training module, configured to construct and periodically train a dynamic credit model based on the data lake, a plurality of predefined interpretable factors, and a preset model self-training rule; the plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight, and a risk adjustment factor;

[0048] a limit adjustment module, configured to, when receiving a cash application initiated by the merchant, adjust a original credit limit corresponding to the merchant based on the current dynamic credit model, determine and output a target credit limit corresponding to the cash application, so as to complete a corresponding credit process based on the target credit limit.

[0049] In a third aspect, the application provides an electronic device, including:

[0050] a memory, configured to save a computer program;

[0051] A processor is configured to execute the computer program to implement the steps of the credit limit dynamic configuration method based on merchant cash register data.

[0052] In a fourth aspect, the present application provides a computer readable storage medium for storing a computer program, which, when executed by a processor, implements the steps of the credit limit dynamic configuration method based on merchant cash register data.

[0053] It can be seen that, in the present application, the bank credit system comprises: based on a multi-thread concurrent strategy and a standardized interface, real-time collection of transaction data, inventory data, member data, industry data and credit investigation data corresponding to a merchant in a software-as-a-service cash register system, triggering of timestamp and device fingerprint recording operations of the data, and obtaining of a merchant cash register data collection result; based on a data governance strategy, verification, desensitization, standardization processing and abnormal data backflow cleaning of the data in the merchant cash register data collection result, and storage of the corresponding governed data to a data lake; based on the data lake, a plurality of predefined interpretable factors and preset model self-training rules, construction and periodic training of a dynamic credit model; the plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight and a risk adjustment factor; when receiving a cash application initiated by the merchant, adjusting a original credit limit corresponding to the merchant based on the current dynamic credit model, determining and outputting a target credit limit corresponding to the cash application, so as to complete a corresponding credit process based on the target credit limit. That is, in the present application, the bank credit system first collects multi-dimensional data of a merchant in a software-as-a-service cash register system based on a multi-thread concurrent strategy and a standardized interface, then verifies, desensitizes, standardizes and cleanses abnormal data backflow of the merchant cash register data collection result, and stores it to a data lake, and then constructs and periodically trains a dynamic credit model based on the data lake, a plurality of predefined interpretable factors and preset model self-training rules. When receiving a cash application initiated by the merchant, the original credit limit of the merchant is adjusted based on the current dynamic credit model to determine a target credit limit. In this way, the problems of high due diligence cost, low credit coverage and lagging risk control in the traditional credit mode can be solved, the financing availability of small and micro merchants can be greatly improved, and business cooperation between banks, merchants and cash register system providers can be promoted, thereby realizing "data as credit" unsecured and artificial due diligence inclusive credit, reducing due diligence cost, realizing dynamic synchronization of credit, merchant operation and risk control, and improving user experience. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the accompanying drawings in the following description only represent some embodiments of the present application, and all other drawings obtained by those of ordinary skill in the art without creative effort based on the provided drawings are within the scope of the present application.

[0055] Figure 1 A flow chart of a credit limit dynamic configuration method based on merchant cash register data is provided for the present application.

[0056] Figure 2 A flow chart of a specific credit limit dynamic configuration method based on merchant cash register data is provided for the present application.

[0057] Figure 3 A specific data governance process schematic diagram is provided for the present application.

[0058] Figure 4 A merchant loan application process schematic diagram is provided for the present application.

[0059] Figure 5 A structure schematic diagram of a credit limit dynamic configuration device based on merchant cash register data is provided for the present application.

[0060] Figure 6 A structure diagram of an electronic device is provided for the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments only represent some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of the present application.

[0062] Currently, the availability of financing for small and micro merchants has been limited by traditional credit granting modes for a long time. In this mode, there are problems such as high due diligence cost, low credit coverage rate, and risk control lag, which cannot adapt to the rapid market changes and cannot meet the growing operating capital needs of small and micro merchants, limiting the in-depth development of inclusive finance.

[0063] Moreover, the traditional credit granting mode has been difficult to adapt to the demand for building a perfect transaction financial ecosystem. On the other hand, although the merchant cash register SaaS service independently developed and promoted by commercial banks has been rapidly popularized, these high-dimensional and high-frequency data lack effective coupling with credit and risk control models and have not been fully utilized in the field of small and micro merchant financing.

[0064] To this end, the application provides a credit limit dynamic configuration scheme based on merchant cash register data, which can solve the problems of high due diligence cost, low credit coverage, and risk control lag in traditional credit granting mode, greatly improve the financing availability of small and micro merchants, and promote business cooperation between banks, merchants, and cash register system providers, thereby realizing "data as credit" unsecured and artificial due diligence inclusive credit, reducing due diligence cost, and realizing dynamic synchronization of credit granting, merchant operation, and risk control, thereby improving user experience.

[0065] Referring to Figure 1 The embodiment of the application discloses a credit limit dynamic configuration method based on merchant cash register data, applied to a bank credit system, comprising:

[0066] In step S11, based on a multi-thread concurrent strategy and a standardized interface, the transaction data, inventory data, member data, industry data, and credit investigation data corresponding to the merchant in the software-as-a-service cash register system are collected in real time, and a timestamp and a device fingerprint record operation of the data are triggered to obtain a merchant cash register data collection result.

[0067] In this embodiment, in combination with Figure 2 As shown in the figure, in the real-time multi-dimensional data collection process, the system establishes a long connection with the cash register system through a standardized API (Application Programming Interface, API) and collects data based on the standardized interface, that is, establishes a connection with the software-as-a-service cash register system based on the standardized interface; based on a multi-thread concurrent strategy, an open authorization mechanism, and a hypertext transfer protocol version 2, the transaction data, inventory data, member data, industry data, and credit investigation data corresponding to the merchant in the software-as-a-service cash register system are collected in real time; wherein the transaction data includes the amount, time, payment method, and location of the transaction; the inventory data includes the inventory quantity unit code, quantity, warehouse entry and exit time, and turnover rate of the inventory; the member data includes the level, consumption frequency, amount, and rights and interests usage of the member; the industry data includes the industry code and industry sentiment index; and the credit investigation data includes the credit score, historical overdue data, and litigation data. It can be understood that based on the HTTP / 2 protocol (Hypertext Transfer Protocol version 2, HTTP / 2), the connection supports multiplexing and server push, effectively reducing data transmission delay and bandwidth consumption. The system uses the OAuth2.0 authorization mechanism (Open Authorization 2.0, OAuth2.0) to ensure the security and legality of data transmission. That is, in the data collection stage, the system uses multi-thread concurrent technology to capture five-dimensional data of transaction, inventory, member, industry, and credit investigation.

[0068] It needs to be understood that the collected transaction data includes amount, time, payment method, location, etc.; the inventory data covers SKU code, quantity, warehouse-in and warehouse-out time, turnover rate, etc.; the member data involves level, consumption frequency, amount, benefit usage, etc.; the industry data is obtained through industry code and business index; and the credit investigation data contains credit score, historical overdue, legal lawsuit, etc.

[0069] Further, in the embodiment, to ensure the accuracy and traceability of the data, the system adds a time stamp and a device fingerprint to each record, that is, for any data collected, based on the coordinated universal time format and the data generation time corresponding to the current item of data, a target time stamp corresponding to the current item of data is determined and recorded; by collecting the hardware features and software features of the corresponding device and combining a preset encryption algorithm, the device fingerprint information corresponding to the current item of data is determined and recorded; the hardware features include device model, media access control address and operating system version; the software features include browser fingerprint and application version number; based on the collected data and the target time stamp and the device fingerprint information corresponding to the data, a target data collection result is determined; in the process of returning the target data collection result, the target data collection result is compressed based on a preset compression technology, and the corresponding compressed data is transmitted to the ring buffer of the local data receiving end by using a preset encryption protocol and a breakpoint resume mechanism, so as to determine the merchant cash register data collection result when the compressed data is checked in the ring buffer. Wherein, the time stamp can record the data generation time with nanosecond-level precision in UTC format (Coordinated Universal Time); the device fingerprint is generated by collecting device hardware features (such as MAC address (Media Access Control Address), device model, operating system version) and software features (such as browser fingerprint, application version number), and is encrypted by SHA-256 (Secure Hash Algorithm 256-bit).

[0070] It can be understood that during data transmission, the system adopts Gzip or Zstandard compression technology to reduce transmission volume, and uses SSL / TLS (Secure Sockets Layer / Transport Layer Security) encryption transmission to prevent data leakage and man-in-the-middle attacks. Moreover, at the data receiving end, the system sets up a ring buffer to ensure efficient storage and reading of data. After receiving the data, integrity, format correctness and consistency checks are immediately performed, and the breakpoint resume mechanism is used to ensure complete data reception. Finally, the collected merchant POS data is transmitted to the data processing module through the data bus, providing support for subsequent data processing. The entire process uses load balancing technology to reasonably allocate data processing tasks.

[0071] In step S12, the data in the merchant POS data collection result is verified, desensitized, standardized, and abnormal data is cleaned, and the corresponding data after governance is stored in the data lake.

[0072] In combination with Figure 2 As shown in the figure, in this embodiment, after obtaining the result of real-time multi-dimensional data collection, the collected data is subjected to closed-loop data governance. That is, for different data types and different business scenarios in the merchant POS data collection result, based on the rule engine, integrity check, consistency check, data format check, data type check and value range check are performed respectively to determine the check result; for the first data in the merchant POS data collection result represented by the check result that passes the check, sensitive data recognition is performed to determine the sensitive data recognition result; when the sensitive data recognition result indicates that there is preset sensitive data, data sensitivity and data usage scenario analysis are performed based on the sensitive data recognition result, and the target desensitization strategy is determined using the corresponding analysis result; the data usage scenario includes development and test scenario and data analysis scenario; desensitization processing is performed based on the target desensitization strategy, the analysis result and the sensitive data recognition result to determine the desensitized data; the desensitized data is standardized to determine the standardized data; for the abnormal sensitive data in the merchant POS data collection result, or the second data represented by the check result that fails the check, corresponding data collection operations are performed again using the software as a service POS system. That is, this embodiment performs automatic verification, desensitization, standardization, lake entry and metadata registration on the merchant POS data collected in the foregoing steps; abnormal data triggers backflow cleaning to form a "collection-feedback-re-collection" closed loop to ensure data freshness.

[0073] In combination with Figure 3 As shown in the figure, the closed-loop data governance process specifically includes the following steps:

[0074] (a) Automatic verification: The system verifies the completeness, accuracy, and consistency of the collected merchant POS data according to pre-set data verification rules. It uses a rule engine to check data format, data type, value range, etc. For example, it checks whether the transaction amount is within a reasonable range, whether the inventory quantity is non-negative, whether the member card number meets the established coding rules, etc. It customizes verification rules for different data types and business scenarios to ensure that the data meets business requirements. For data that fails the verification, the system records detailed error information for subsequent processing. For example, if the transaction amount in the record is negative or exceeds the upper limit of the conventional transaction amount in the merchant's industry, the system determines that the data is abnormal and records the specific error content.

[0075] (b) Desensitization processing: The system automatically identifies sensitive data such as merchant bank account information and customer contact information. According to the sensitivity of the data and the use scenario, it selects the appropriate desensitization method. For development and testing scenarios, it uses techniques such as replacement and deformation to ensure that the desensitized data retains the format and statistical properties of the original data. For data analysis scenarios, it values the relevance between data and uses techniques such as generalization and homogenization to avoid affecting the accuracy of data analysis results due to desensitization. The desensitization process is implemented through the configuration of desensitization rules and algorithms to ensure that the desensitized data cannot be reversed and does not affect the normal use of the data. For example, the merchant's bank account is replaced with stars, and the middle four digits of the customer's phone number are replaced with stars to protect the privacy information of the merchant and customer.

[0076] (c) Standardization processing: The system standardizes the data to ensure consistency and comparability. It converts different sources of transaction time fields to a standard time format and unifies the currency unit of transaction amounts. For example, it converts different inventory units of goods such as "pieces", "individuals", and "boxes" to "pieces". It establishes an enterprise-level data dictionary to uniformly define the business meaning, format, and measurement unit of data to ensure consistent understanding of data by different departments.

[0077] (d) Data into lake: After verification, desensitization, and standardization processing, the data is stored in the data lake according to the pre-set data organization structure and storage strategy. The data lake provides massive and highly scalable storage capabilities, supporting the storage of structured, semi-structured, and unstructured data. For example, it stores structured data such as member consumption records and level information in a relational database and stores unstructured data such as member consumption preferences in the form of text files in object storage. At the same time, the system optimizes partition storage and indexing of data to improve data query and retrieval efficiency.

[0078] (e) Metadata registration: The system automatically registers and manages the metadata of data. Metadata includes information such as the source of data, data type, data format, data quality indicators, and data blood relationship. Taking transaction data as an example, record the metadata information such as which merchant and which cash register equipment the transaction data comes from, and the collection time. By establishing a data blood map, the flow path of key data can be visualized, and the responsible subject and flow process of the data can be clarified. Metadata management tools are used to track the whole process of data from collection to storage, processing, and analysis, providing support for data traceability and problem troubleshooting.

[0079] (f) Abnormal data backflow cleaning: The system monitors data quality in real time, and for sensitive data that fails to pass verification or has abnormalities, triggers a backflow cleaning process. Abnormal data is sent back to the data collection end or data cleaning module for targeted repair or re-collection. During the backflow cleaning process, the system records the abnormal situation and processing of the data, forming a closed-loop management to ensure continuous improvement of data quality. For example, if it is found that the transaction time format of a certain transaction data is incorrect, the system will send the data back to the cash register system to re-collect the correct transaction time information.

[0080] Step S13, constructing and periodically training the dynamic credit model based on the data lake, a plurality of predefined interpretable factors, and a preset model self-training rule; the plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight, and a risk adjustment factor.

[0081] In combination Figure 3As shown, in this embodiment, a dynamic credit granting model is constructed using multiple predefined interpretable factors and data from the data lake. The model is then periodically trained using preset model self-training rules. Specifically, multiple predefined interpretable factors and their corresponding target formulas are obtained. Each interpretable factor is used as an independent variable, and the merchant's credit limit is used as a dependent variable. The dynamic credit granting model is constructed by combining the target formula, a preset machine learning algorithm, and the data lake. Model training is triggered based on a preset time interval. If triggered, the latest merchant payment data is retrieved from the data lake using a scrolling window, and this latest payment data is used to train the dynamic credit granting model. The dynamic credit granting model is trained and validated, and the factor weights corresponding to each interpretable factor are updated to determine the updated model. During the construction and training of the model, transaction data is monitored using a preset circuit breaker mechanism and a preset threshold. Anomaly identification is performed based on the data monitoring results to determine the anomaly identification results. When the anomaly identification results indicate the existence of a preset anomaly, a risk adjustment factor reduction operation is triggered based on the target merchant, event type, and severity of the preset anomaly in the anomaly identification results to obtain a reduction result. Based on the reduction result, it is determined whether to trigger a credit limit freeze operation corresponding to the target merchant. It is understood that the dynamic credit granting model uses the Trading Stability Index (TSI), Inventory Health Index (IHI), Member Viscosity Coefficient (MVC), Industry Prosperity Weight (IIW), and Risk Adjustment Factor (RAF) as core variables, combined with the Shapley value algorithm to output a readable explanation of a single credit granting transaction.

[0082] It is important to understand the calculation and output of the explanatory factor of the model's independent variables, as shown below:

[0083] (1) The stability of merchant cashier transactions has a positive effect on credit limit increases. For the Transaction Stability Index (TSI) calculation, the system calculates the ratio of the standard deviation of the daily transaction amount to the average transaction amount within a certain time window (e.g., the past 30 trading days) based on the transaction flow information in the merchant cashier data. The formula is as follows:

[0084] .

[0085] In the formula, This represents the standard deviation of transaction amounts over the past 30 trading days. represents the average value of transaction amount in the past 30 trading days. The standard deviation of transaction amount reflects the transaction fluctuation, and the average transaction amount represents the average business scale of the merchant. The ratio of the standard deviation to the average value can quantify the stability of the transaction. The value range of TSI is (0, 1), and the larger the TSI value, the more stable the merchant transaction, and the relatively lower the operating risk. Example calculation: assuming that the average transaction amount of a certain merchant is 1000 yuan, the standard deviation of transaction amount is 200 yuan, then TSI = 1 / [1+(200 / 1000)]≈0.833.

[0086] (2) The healthier the inventory data of the merchant cash register, the more positive effect on the credit limit increase. For the calculation of inventory health index (IHI), the system analyzes the inventory data in the cash register system of the merchant, including average inventory quantity, annual sales quantity, non-moving sales ratio, total inventory quantity, obsolete inventory amount, and total inventory amount. The calculation formula is:

[0087]

[0088] In the formula, DIO (Days Inventory Outstanding) represents the turnover days, and the turnover days = (average inventory quantity x 365) / annual sales quantity. The shorter the turnover days, the faster the inventory turnover, and the healthier the inventory. NMR (Non Moving Sales Ratio) represents the non-moving sales ratio, that is, the inventory with no record of out-of-stock in the past 30 days accounts for the proportion of total inventory, and the non-moving sales ratio = non-moving sales quantity / total inventory quantity. The lower the non-moving sales ratio, the healthier the inventory. OR (Obsolescence Rate) represents the obsolescence rate, that is, the inventory with an age of more than 90 days accounts for the proportion of total inventory, and the obsolescence rate = (obsolescence inventory amount / total inventory amount) x 100%. The lower the obsolescence rate, the healthier the inventory. The value range of IHI is (0, 1), and the larger the IHI value, the healthier the inventory data of the merchant cash register, and the relatively lower the operating risk. Example calculation: assuming that the average inventory quantity of a certain merchant is 1000, the annual sales quantity is 3650, the non-moving sales ratio is 0.1, and the obsolescence rate is 0.05, then IHI = 1 / (1+0.274+0.1+0.05)≈0.702.

[0089] (3) For the calculation of member stickiness coefficient (MVC), the system uses member data, including member level, consumption frequency, and consumption amount. The member repeat purchase rate is calculated, that is, the number of members who have consumed at least twice in the statistical period (such as the past 6 months) accounts for the proportion of total number of members. The calculation formula is:

[0090] ​​​​

[0091] In the formula, RPR (Repeat Purchase Rate) represents the member repeat purchase rate, i.e. the proportion of the number of members who have consumed at least twice to the total number of members within a statistical period (such as the past 6 months). Member repeat purchase rate = number of members who have consumed at least twice / total number of members. The higher the member repeat purchase rate, the higher the loyalty of the members to the merchant. AMT (Average Consumption Interval Time) represents the average consumption interval time, i.e. the average time interval between two consumptions of a member, reflecting the consumption activity of the member. Average consumption interval time = total member consumption interval days / member consumption times. The shorter the average consumption interval time, the higher the consumption frequency of the member and the stronger the stickiness. The MVC value ranges from (0, 1), and the larger the MVC value, the better the stickiness of the merchant member customers and the relatively lower the operating risk. Example calculation: assuming that the total number of members of a certain merchant is 100, the number of members who have consumed at least twice is 40, the total member consumption interval days is 365 days, and the member consumption times is 10, then MVC = 1 / (1+2.5+36.5)≈0.0256, indicating that the member stickiness is at a low level and the sustainability risk of the merchant experience is high.

[0092] (4) For the determination of the industry sentiment index (IIW), the system obtains industry index data of multiple authoritative institutions through external data sources, including retail industry sentiment index, retail industry sentiment index monthly analysis report, Wind information provides entrepreneur confidence index of wholesale and retail industry, and other related institutions publish retail industry sentiment index report, retail industry sentiment index analysis report, etc. The specific data source is more actual or adjusted. The system uniformly normalizes the external index data, then multiplies the normalized sentiment index by the corresponding weight, sums up, and obtains the final IIW value.

[0093] EDS represents a certain external index data source, and the normalized external index data source NEDS is calculated according to the following formula:

[0094] ;

[0095] In the formula, n represents the nth external index data source. The calculation formula of the industry sentiment index (IIW) is as follows:

[0096] ;

[0097] Wherein, NEDS represents the normalized external data source industry sentiment index, represents the weight. The IIW value ranges from (0, 1), and the larger the IIW value, the relatively more prosperous the industry to which the merchant belongs, and the relatively lower the operating risk.

[0098] (5) Regarding the calculation of the risk adjustment factor (RAF), the system uses non-fixed machine learning algorithms such as logistic regression, decision tree, random forest, and gradient boosting tree (such as XGBoost and LightGBM) to calculate the default probability p. It selects feature variables related to default risk and uses credit features such as credit score, number of historical delinquencies, and credit limit utilization rate to complete the calculation of the default probability p. The formula is as follows:

[0099] .

[0100] in, , , … These are the coefficients of the model, obtained by fitting the training data. , … These are selected feature variables, such as credit score, number of historical delinquencies, and credit limit utilization rate.

[0101] The formula for calculating RAF is as follows:

[0102] .

[0103] in, This is the risk adjustment factor, with a value range of (0,1), and a default value of 1. Figure 2 The dynamic adjustment is shown in step S104. The value of RAF is within the range of (0,1]. The closer the risk adjustment factor RAF is to 0, the higher the risk, and the greater the adjustment range of the credit limit, thereby achieving effective risk control. Example calculation: Assuming a merchant has a default probability p of 0.1 calculated by machine learning, and a risk adjustment coefficient w of 0.8, then RAF = (1−0.2)×(1-0.1) = 0.72.

[0104] (6) Finally, the system uses five factors—TSI (Transaction Stability Index), IHI (Inventory Health Index), MVC (Member Stickiness Coefficient), IIW (Industry Prosperity Weight), and RAF (Risk Adjustment Factor)—as independent variables, and credit limit as the dependent variable. It employs machine learning algorithms such as multiple linear regression and logistic regression to construct a credit granting model, and dynamically maintains the factor weights (β). During model construction, the model is optimized using methods such as cross-validation and grid search to ensure its accuracy and generalization ability. Simultaneously, the system combines the Shapley value algorithm to calculate the contribution of each factor to the credit limit, outputting the contribution value of each factor and its direction of influence on the credit limit, enabling merchants to clearly understand the model's decision-making basis and achieving interpretability of credit granting decisions. Furthermore, to reduce data noise and improve credit limit effectiveness, a minimum credit limit threshold is set for credit limit increases. , if the calculated extra credit amount is less than a, then the extra credit amount is equal to 0; if the extra credit amount is greater than or equal to , the data is valid.

[0105] The formula for calculating the extra credit amount (ECA) is as follows:

[0106] .

[0107] wherein, is the coefficient of the model, is the error term. If the calculated extra credit amount , then ECA = 0; if the calculated extra credit amount ≥ , then ECA = the calculated extra credit amount.

[0108] Example calculation: Assume is 10000, is 5000, is 3000, is 2000, is 4000, is 6000, TSI is 0.8, IHI is 0.7, MVC is 0.6, IIW is 0.9, RAF is 0.8, is 0, is 20000. Then ECA = 10000 + 5000 x 0.8 + 3000 x 0.7 + 2000 x 0.6 + 4000 x 0.9 + 6000 x 0.8 = 25700 yuan. 25700 ≥ 20000, so the extra credit amount (ECA) = 25700 yuan.

[0109] Further, as shown in Figure 2 , the present embodiment will also perform model dynamic self-evolution and RAF melting. In terms of model dynamic self-evolution mechanism. First, the system will incrementally retrain the weight every month, using the latest cash register data of merchants, retraining the machine learning model every month to adjust the weight to improve accuracy and adapt to market changes. Second, the system has a periodic update and rolling window mechanism for data, i.e. the model training data uses a rolling window mechanism and is updated every month to ensure the timeliness and representativeness of the data. In addition, the system will perform model optimization and verification, using techniques such as cross-validation and grid search to find the optimal parameter combination and improve the performance of the model. Finally, the system dynamically adjusts the factor weight w, and the weight of each factor in the credit model can be dynamically adjusted to reflect the changes in its impact on the default risk. For example, when an abnormal event (such as a transaction drop of ≥ 30%) is detected, the risk adjustment factor is lowered from the value 1 to 0.85.

[0110] Regarding the RAF circuit breaker mechanism, the first step is real-time monitoring and anomaly detection. The system monitors merchant transaction data in real time and identifies abnormal events, such as sudden drops in transaction volume or overdue payments, by setting thresholds and rules. Then there's the dynamic RAF reduction logic, which triggers RAF reduction rules based on the type and severity of the abnormal event. For example, if the transaction volume drops by 30%, the risk adjustment coefficient is adjusted. This will cause the RAF value to decrease by 20%; frequent delinquencies will result in a 30% decrease. Furthermore, circuit breaker conditions and credit limit freezes are core mechanisms. The system sets a lower limit for the RAF value or a risk indicator threshold as the circuit breaker condition. If the RAF falls below 0.3 or the default probability exceeds 0.2, the system automatically freezes the credit limit and suspends credit limit increases until the risk is resolved. Finally, in terms of processing, the system performs risk event feedback and model optimization, feeding abnormal events and RAF adjustments back to the model as retraining data to optimize risk identification and response.

[0111] Step S14: When a loan application initiated by the merchant is received, the original credit limit corresponding to the merchant is adjusted based on the current dynamic credit granting model, and a target credit limit corresponding to the loan application is determined and output so as to complete the corresponding credit granting process based on the target credit limit.

[0112] In this embodiment, merchants can base their actions on... Figure 4 The process shown initiates a loan application. Afterwards, the system adjusts the merchant's original credit limit based on the application. Specifically, when a loan application is received from the merchant through the POS terminal of the Software as a Service (SaaS) POS system, it determines whether the merchant is an existing customer with existing credit to determine the merchant's credit limit. If the merchant's credit limit is yes, the system retrieves the merchant's original credit limit. If the merchant's credit limit is no, the system determines the merchant's basic credit limit based on a preset credit limit model and uses this basic credit limit as the original credit limit. Based on the current dynamic credit model and the data lake, the system determines the merchant's additional credit amount. Based on this additional credit amount, the system increases the original credit limit, determines and outputs the target credit limit corresponding to the loan application, so that the corresponding credit granting process can be completed based on the target credit limit.

[0113] Combination Figure 4The one-key loan and post-loan monitoring process is shown. After the merchant initiates the loan application in the "financial services" panel of the SaaS cash register terminal, the system automatically starts the automatic processing process. The system determines whether it is a customer with a credit line from the bank. If so, the original credit line is directly called. If not, the traditional credit line model of the small and micro customer group is called. This model mainly focuses on the income, profit, assets and credit of the customer. Through the calculation of the merchant's net profit, operating income, real estate value and other methods, the basic credit line is obtained, which is considered as the basic line. The system calculates the dynamic credit line based on the five-factor model (i.e. dynamic credit model), superimposes the credit line, determines the final credit line and interest rate. After completing the credit, the system synchronously starts the post-loan monitoring mechanism, collects the operating data of the merchant in real time, and uses big data analysis technology to monitor and analyze the data. Once the risk signal exceeds the preset threshold, the system will automatically push the post-loan task to the bank credit core system to ensure that the risk can be controlled in time and effectively.

[0114] That is, after completing the corresponding credit process based on the target credit line, it also includes: synchronously starting a post-loan monitoring mechanism; collecting operating data of the merchant based on the post-loan monitoring mechanism and the software as a service cash register system to determine a data collection result; the operating data includes the transaction data, the inventory data, the member data, the industry data, and the credit data; monitoring and analyzing data in the data collection result using a preset big data analysis algorithm to determine a data analysis result; identifying risks based on the data analysis result, and determining whether to trigger a risk control operation using a corresponding risk identification result.

[0115] In addition, in combination with Figure 2As shown, the present embodiment can also realize bank non-inductive direct authorization. The system adopts a "pre-authorization token pool" mechanism (OAuth2.0+ two-way TLS), and the merchant completes one-time authorization when signing up for the first time. Subsequent credit, lending, and repayment processes are all called bank APIs by the cash register system background in silence, with zero interaction and zero perception on the merchant side. Specifically, the system pre-generates and stores a batch of authorization tokens to form a token pool. When the merchant signs up for the first time, an authorization token is allocated from the token pool to complete one-time authorization. In subsequent business processes such as credit, lending, and repayment, the cash register system background obtains the corresponding access token from the token pool according to business needs, communicates through two-way TLS encryption, silently calls the bank's API interface, and sends business requests. After receiving the request, the bank's API interface verifies the access token, executes the corresponding operation if the verification is passed, and returns the result. If the access token expires, the cash register system uses the refresh token to obtain a new access token and refresh token, and updates the token pool. For example, when a small and micro merchant uses a certain cash register SaaS service, it needs to apply for a bank credit limit. When signing up for the cash register SaaS service for the first time, the system guides the merchant to complete the bank non-inductive direct authorization. After the merchant completes one-time authorization through simple operation, the cash register system background automatically obtains the merchant's cash register data, and dynamically assesses the merchant's credit status and credit limit according to the five-factor explainable credit model. In the subsequent business process, such as the merchant applying for an increase in the credit limit, the cash register system background silently calls the bank API, and the merchant does not need to input identity information for verification again, thus completing the increase operation and realizing non-inductive authorization. This process not only improves authorization efficiency, but also enhances user experience, while ensuring data security and promoting business cooperation between banks, merchants, and cash register system providers.

[0116] In summary, the embodiment discloses a real-time collection-cleaning-modeling-credit granting-risk early warning whole process system architecture based on cash register data, taking five-dimensional data of cash register real-time transaction flow, SKU inventory turnover, member value, industry sentiment and external credit investigation as input, constructing a dynamic credit granting model through five interpretable factors of TSI, IHI, MVC, IIW and RAF, realizing minute-level non-collateral credit line calculation of "data is credit" and automatic daily update, and triggering the adjustment of the credit line through risk events. Among them, through real-time collection of transaction, inventory, member, industry and credit investigation multi-dimensional data of the cash register system, the dynamic credit granting model is used to complete the credit line calculation at the minute level, realize the non-collateral and non-artificial credit granting of "data is credit", significantly reduce the bank's investigation cost by more than 70%, and control the non-performing rate within 1%; the model uses machine learning to continuously learn default samples, and the credit line is updated daily and adjusted in real time by risk events, truly realizing the dynamic synchronization of credit granting and operation; the system directly connects the bank's credit core through standard API, and the merchant side can view the credit line and withdraw money with one key, the funds are credited in the fastest 2 hours, and the availability of small and micro merchant financing is greatly improved; at the same time, the bank can obtain customers in batches through the cash register scene and deposit transaction closed-loop data, laying an ecological foundation for cross-selling of subsequent deposit, wealth management and supply chain financial products, forming a closed loop of cash register, credit granting, risk control and operation, and having an industry-leading advantage in three aspects of inclusive finance, accurate risk control and ecological operation.

[0117] As can be seen, in the present application, first, through the bank credit system, based on multi-thread concurrent strategy and standardized interface, the multi-dimensional data of the merchant of the software as a service cash register system is collected in real time, then the merchant cash register data collection results are verified, desensitized, standardized and abnormal data backflow cleaned, and stored in the data lake, and then based on the data lake, a plurality of interpretable factors and preset model self-training rules, a dynamic credit granting model is constructed and periodically trained, and when receiving a money application initiated by the merchant, the original credit line of the merchant is adjusted based on the current dynamic credit granting model to determine the target credit line. In this way, the problems of high investigation cost, low credit coverage and lagging risk control in the traditional credit granting mode can be solved, the availability of small and micro merchant financing is greatly improved, and business cooperation between banks, merchants and cash register system providers is promoted, thereby realizing the non-collateral and non-artificial inclusive credit granting of "data is credit", reducing the investigation cost, and realizing the dynamic synchronization of credit granting, merchant operation and risk control, thereby improving the user experience.

[0118] Referring to Figure 5 The embodiment of the present application also discloses a credit line dynamic configuration device based on merchant cash register data, applied to a bank credit system, comprising:

[0119] The data collection module 11 is configured to collect transaction data, inventory data, member data, industry data, and credit investigation data of a merchant in a software-as-a-service (SaaS) cash register system in real time based on a multi-thread concurrent strategy and a standardized interface, trigger timestamp and device fingerprint recording operations, and obtain a merchant cash register data collection result.

[0120] The data governance module 12 is configured to verify, desensitize, standardize, and clean up abnormal data of the merchant cash register data collection result based on a data governance strategy, and store the corresponding governed data to a data lake.

[0121] The model self-training module 13 is configured to construct and periodically train a dynamic credit model based on the data lake, a plurality of predefined interpretable factors, and a preset model self-training rule. The plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight, and a risk adjustment factor.

[0122] The credit limit adjustment module 14 is configured to adjust a current dynamic credit model when receiving a cash application initiated by the merchant, adjust the original credit limit corresponding to the merchant based on the dynamic credit model, determine and output a target credit limit corresponding to the cash application, and complete a corresponding credit process based on the target credit limit.

[0123] In some embodiments, the data collection module 11 can be configured to establish a connection with a SaaS cash register system based on a standardized interface, collect transaction data, inventory data, member data, industry data, and credit investigation data of a merchant in a SaaS cash register system in real time based on a multi-thread concurrent strategy, an open authorization mechanism, and a hypertext transfer protocol version 2 (HTTP / 2), and the transaction data includes transaction amount, time, payment method, and location. The inventory data includes inventory quantity unit code, quantity, inventory time, and turnover rate. The member data includes member level, consumption frequency, amount, and rights usage. The industry data includes industry code and industry sentiment index. The credit investigation data includes credit score, historical overdue data, and litigation data.

[0124] In some embodiments, the data collection module 11 can be specifically configured to: for any collected data, determine and record a target timestamp corresponding to the current data based on the coordinated universal time format and the data generation time corresponding to the current data; determine and record the device fingerprint information corresponding to the current data by collecting the hardware features and software features of the corresponding device and combining a preset encryption algorithm; the hardware features include device model, media access control address and operating system version; the software features include browser fingerprint and application version number; determine the target data collection result based on the collected data and the target timestamp and the device fingerprint information corresponding to the data; in the process of returning the target data collection result, compress the target data collection result based on a preset compression technology, and use a preset encryption protocol and a breakpoint resume mechanism to transmit the corresponding compressed data to the ring buffer of the local data receiving end, so as to determine the merchant cash register data collection result when the compressed data is verified in the ring buffer.

[0125] In some embodiments, the data management module 12 can be specifically configured to: for data of different data types and different business scenarios in the merchant cash register data collection result, respectively perform integrity checking, consistency checking, data format checking, data type checking and value range checking based on a rule engine to determine a checking result; for the first data in the merchant cash register data collection result represented by the checking result that passes the checking, perform sensitive data identification to determine a sensitive data identification result; when the sensitive data identification result indicates that there is preset sensitive data, perform analysis of data sensitivity and data use scenarios based on the sensitive data identification result, and determine a target desensitization strategy using the corresponding analysis result; the data use scenarios include development test scenarios and data analysis scenarios; perform desensitization processing based on the target desensitization strategy, the analysis result and the sensitive data identification result to determine desensitized data; perform standardized processing on the desensitized data to determine standardized data; for the abnormal sensitive data in the merchant cash register data collection result or the second data represented by the checking result that fails the checking, perform corresponding data collection operations again using the software as a service cash register system.

[0126] In some embodiments, the model self-training module 13 can be specifically configured to: obtain a plurality of predefined interpretable factors and target formulas corresponding to each of the interpretable factors; take each of the interpretable factors as an independent variable, take the merchant credit line as a dependent variable, and combine the target formulas, a preset machine learning algorithm, and the data lake to construct a dynamic credit model; determine whether to trigger model training based on a preset time interval, and when yes, obtain the latest merchant POS data from the data lake based on a rolling window, and train and verify the dynamic credit model using the latest merchant POS data, and update the factor weights corresponding to each of the interpretable factors to determine an updated dynamic credit model; in the process of constructing and training the dynamic credit model, monitor the transaction data through a preset fuse mechanism and a preset threshold, and identify abnormal events according to the data monitoring result to determine an abnormal event identification result; when the abnormal event identification result indicates that there is a preset abnormal event, trigger a risk adjustment factor decrement operation based on the target merchant, event type, and severity corresponding to the preset abnormal event in the abnormal event identification result to obtain a decrement result; and determine whether to trigger a credit line freezing operation corresponding to the target merchant based on the decrement result.

[0127] In some embodiments, the credit adjustment module 14 can be specifically configured to: when receiving a withdrawal application initiated by the merchant through the software-as-a-service POS system, determine whether the merchant is a customer with a granted credit line to determine a merchant determination result; if the merchant determination result is yes, retrieve the original credit line corresponding to the merchant; if the merchant determination result is no, determine a basic credit line corresponding to the merchant based on a preset credit model, and take the basic credit line as the original credit line; determine an additional credit amount corresponding to the merchant based on the current dynamic credit model and the data lake; and increase the original credit line based on the additional credit amount to determine and output a target credit line corresponding to the withdrawal application, so as to complete a corresponding credit process based on the target credit line.

[0128] In some embodiments, the credit line dynamic configuration apparatus based on merchant POS data can also be configured to: simultaneously start a post-loan monitoring mechanism; collect operating data of the merchant based on the post-loan monitoring mechanism and the software-as-a-service POS system to determine a data collection result; the operating data includes the transaction data, the inventory data, the member data, the industry data, and the credit investigation data; monitor and analyze the data in the data collection result using a preset big data analysis algorithm to determine a data analysis result; perform risk identification based on the data analysis result, and determine whether to trigger a risk control operation using the corresponding risk identification result.

[0129] Further, the embodiment of the present application further discloses an electronic device, Figure 6 is an electronic device 20 structure diagram shown according to an exemplary embodiment, the contents in the figure cannot be considered as any limitation to the use range of the present application.

[0130] Figure 6 An electronic device 20 structure diagram is provided for the embodiment of the present application. The electronic device 20 can specifically include at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25 and a communication bus 26. Wherein, the memory 22 is used for storing a computer program, the computer program is loaded and executed by the processor 21, to realize the related steps in the credit limit dynamic configuration method based on merchant cash register data disclosed in any of the preceding embodiments. In addition, the electronic device 20 in the embodiment can be an electronic computer.

[0131] In the embodiment, the power supply 23 is used for providing working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol followed by the communication interface 24 can be any communication protocol applicable to the technical solution of the present application, which is not limited here; the input / output interface 25 is used for obtaining external input data or outputting data to the outside world, and the specific interface type can be selected according to the specific application needs, which is not limited here.

[0132] In addition, the memory 22 as a resource storage carrier can be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage mode can be temporary storage or permanent storage.

[0133] Wherein, the operating system 221 is used for managing and controlling each hardware device on the electronic device 20 and the computer program 222, which can be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program capable of completing the credit limit dynamic configuration method based on merchant cash register data executed by the electronic device 20 disclosed in any of the preceding embodiments, the computer program 222 can further include a computer program capable of completing other specific work.

[0134] Further, the present application further discloses a computer readable storage medium for storing a computer program; wherein the computer program is executed by the processor to realize the credit limit dynamic configuration method based on merchant cash register data disclosed in the preceding embodiment. The specific steps of the method can refer to the corresponding contents disclosed in the preceding embodiment, which will not be repeated here.

[0135] The various embodiments described in this specification are presented by way of example, and each embodiment is not necessarily composed of all features described with respect to other embodiments. Each embodiment described in this specification can be implemented in conjunction with one or more other embodiments described in this specification, and / or in combination with one or more features described with respect to other embodiments described in this specification. For example, some embodiments can include, in combination, elements of one or more other embodiments described in this specification, and / or elements of one or more other embodiments described in this specification. The scope of the application is not limited to the features described in connection with any one embodiment described in this specification.

[0136] Those skilled in the art will further appreciate that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various examples disclosed herein are described generally in terms of their functionality, without reference to the corresponding

[0137] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), flash memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is tangible.

[0138] Finally, it should be noted that the terms "first", "second", and the like, herein do not denote any order, quantity, combination, or arrangement, but are used to distinguish one element from another. Also, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0139] The technical solutions provided by the present application are described in detail above, and the principles and implementation manners of the present application are described by applying specific examples. The above description of the examples is only applicable to helping understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application ranges will have changes, and the above description of the content of the specification should not be understood as a limitation on the present application.

Claims

1. A method for dynamically configuring a credit limit based on merchant cash register data, the method comprising: receiving a plurality of merchant cash register data; determining a plurality of merchant cash register data trends; and dynamically configuring a credit limit based on the plurality of merchant cash register data trends. The application is applied to a bank credit system, comprising: Based on multi-thread concurrent strategy and standardized interface, real-time collection of transaction data, inventory data, member data, industry data and credit investigation data corresponding to a merchant in a software-as-a-service POS system, triggering timestamp and device fingerprint recording operation of the data, and obtaining merchant POS data collection results; Based on data governance strategy, the data in the merchant POS data collection results are verified, desensitized, standardized and abnormal data backflow cleaned, and the corresponding governed data is stored in a data lake; Based on the data lake, a plurality of interpretable factors and a preset model self-training rule, a dynamic credit model is constructed and periodically trained; the plurality of interpretable factors include transaction stability index, inventory health index, member stickiness coefficient, industry sentiment weight and risk adjustment factor; When receiving a cash application initiated by the merchant, the current dynamic credit model is used to adjust the original credit limit corresponding to the merchant, determine and output the target credit limit corresponding to the cash application, and complete the corresponding credit process based on the target credit limit. 2.The method of claim 1, wherein, The real-time collection of transaction data, inventory data, member data, industry data and credit investigation data corresponding to a merchant in a software-as-a-service POS system based on multi-thread concurrent strategy and standardized interface, comprising: Based on the standardized interface, the software-as-a-service POS system is connected; Based on multi-thread concurrent strategy, open authorization mechanism and hypertext transfer protocol version 2, real-time collection of transaction data, inventory data, member data, industry data and credit investigation data corresponding to a merchant in a software-as-a-service POS system; The transaction data includes the amount, time, payment method and location of the transaction; the inventory data includes the inventory quantity unit code, quantity, warehouse in-out time and turnover rate; the member data includes the level, consumption frequency, amount and rights and interests usage of the member; the industry data includes the industry code and industry sentiment index; the credit investigation data includes the credit score, historical overdue data and litigation data. 3.The method of claim 1, wherein, The triggering of the timestamp and device fingerprint recording operation of the data to obtain the merchant POS data collection results, comprising: For any item of data collected, based on the coordinated universal time format and the data generation time corresponding to the current item of data, a target timestamp corresponding to the current item of data is determined and recorded; By collecting the hardware features and software features of the corresponding device, and combining a preset encryption algorithm, the device fingerprint information corresponding to the current item of data is determined and recorded; the hardware features include device model, media access control address and operating system version; the software features include browser fingerprint and application version number; Based on the collected data and the target timestamp and device fingerprint information corresponding to the data, a target data collection result is determined; In the process of returning the target data collection result, the target data collection result is compressed based on a preset compression technology, and the compressed data is transmitted to a ring buffer of a local data receiving end using a preset encryption protocol and a breakpoint resume mechanism, so as to determine the merchant cash register data collection result when the compressed data is verified in the ring buffer. 4.The method of claim 1, wherein, The data in the merchant cash register data collection result is checked, desensitized, standardized, and abnormal data is backwashed based on the data governance strategy, including: For different data types and different business scenarios in the merchant cash register data collection result, the rule engine is used for integrity check, consistency check, data format check, data type check, and value range check to determine the check result; For the first data in the merchant cash register data collection result represented by the check result, sensitive data recognition is performed to determine the sensitive data recognition result; When the sensitive data recognition result indicates that there is preset sensitive data, the data sensitivity and data usage scenario are analyzed based on the sensitive data recognition result, and the target desensitization strategy is determined using the corresponding analysis result; the data usage scenario includes development test scenario and data analysis scenario; Desensitization is performed based on the target desensitization strategy, the analysis result, and the sensitive data recognition result to determine the desensitized data; The desensitized data is standardized to determine the standardized data. For the abnormal sensitive data in the merchant cash register data collection result, or the second data represented by the check result, corresponding data collection operations are performed again using the software as a service cash register system. 5.The method of claim 1, wherein, The construction and periodic training of the dynamic credit model based on the data lake, the pre-defined plurality of interpretable factors, and the preset model self-training rules, including: Obtain a plurality of pre-defined interpretable factors and target formulas corresponding to each interpretable factor; Each of the interpretable factors is used as an independent variable, the merchant credit limit is used as a dependent variable, and the target formula, a preset machine learning algorithm, and the data lake are combined to construct a dynamic credit model; Based on a preset time interval, it is determined whether to trigger model training, and when it is, based on a rolling window, the latest merchant cash register data is obtained from the data lake, and the latest merchant cash register data is used to train and verify the dynamic credit model, and the factor weights corresponding to each of the interpretable factors are updated to determine the updated dynamic credit model. In the process of constructing and training the dynamic credit model, the transaction data is monitored through a preset fuse mechanism and a preset threshold, and abnormal event recognition is performed according to the data monitoring result to determine an abnormal event recognition result; When the abnormal event recognition result indicates that there is a preset abnormal event, a risk adjustment factor decrement operation is triggered based on the target merchant, event type, and severity corresponding to the preset abnormal event in the abnormal event recognition result to obtain a decrement result; Determine whether the credit limit freezing operation corresponding to the target merchant is triggered based on the decreasing result. 6.The method of claim 1, wherein, The original credit limit of the merchant is adjusted based on the current dynamic credit model when the cash application initiated by the merchant is received. Determine whether the merchant is a customer with a granted credit based on the cash application initiated by the merchant through the software-as-a-service cash register terminal, to determine a merchant determination result. If the merchant determination result is yes, the original credit limit corresponding to the merchant is retrieved. If the merchant determination result is no, the basic credit limit corresponding to the merchant is determined based on a preset credit model, and the basic credit limit is used as the original credit limit. An additional credit amount corresponding to the merchant is determined based on the current dynamic credit model and the data lake. The original credit limit is increased based on the additional credit amount, and a target credit limit corresponding to the cash application is determined and output, so that the corresponding credit process is completed based on the target credit limit.

7. The method of claim 1 to 6, wherein, After the corresponding credit process is completed based on the target credit limit, the following steps are further included: Synchronously start the post-loan monitoring mechanism. Collect the operating data of the merchant based on the post-loan monitoring mechanism and the software-as-a-service cash register system to determine a data collection result; the operating data includes the transaction data, the inventory data, the member data, the industry data, and the credit investigation data. Monitor and analyze the data in the data collection result using a preset big data analysis algorithm to determine a data analysis result. Risk identification is performed based on the data analysis result, and a corresponding risk identification result is used to determine whether to trigger a risk control operation.

8. A device for dynamically configuring credit limits based on merchant POS data, characterized in that, Applied to a bank credit system, comprising: A data acquisition module is configured to acquire transaction data, inventory data, member data, industry data, and credit investigation data corresponding to a merchant in a software-as-a-service cash register system in real time based on a multi-thread concurrent strategy and a standardized interface, trigger a timestamp and a device fingerprint record operation, and obtain a merchant cash register data acquisition result. A data governance module is configured to verify, desensitize, standardize, and clean up abnormal data of the data in the merchant cash register data acquisition result based on a data governance strategy, and store the corresponding governed data in a data lake. A model self-training module is configured to construct and periodically train a dynamic credit model based on the data lake, a plurality of predefined interpretable factors, and a preset model self-training rule; the plurality of predefined interpretable factors include a transaction stability index, an inventory health index, a member stickiness coefficient, an industry sentiment weight, and a risk adjustment factor. An amount adjustment module is configured to adjust the original credit limit corresponding to the merchant based on the current dynamic credit model when the cash application initiated by the merchant is received, determine and output a target credit limit corresponding to the cash application, and complete the corresponding credit process based on the target credit limit.

9. An electronic device, comprising: Comprising: A memory for saving a computer program; A processor configured to execute the computer program to implement the method of claim 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program product comprising a computer readable medium storing a computer program which, when executed by a processor, implements the method of claim 1 to 7.

Images