Credit line dynamic calculation method and device based on fund system, and storage medium

Abstract

The invention provides a credit line dynamic calculation method and device based on a fund system and a storage medium, and relates to the technical field of financial science and technology, and the method comprises the steps: carrying out the multi-source data fusion processing, and collecting the structured data and unstructured data of a financial system, a credit investigation platform and a business system in real time through a dynamic data engine; converting the unstructured data into quantitative values through semantic analysis; the mixed model carries out cooperative calculation and evaluation on the model, dynamically adjusts index weights according to real-time economic cycle parameters, calculates weighted scores for quantifiable indexes and non-quantitative indexes after semantic conversion, and generates credit rating; the calculation model is used for executing formula operation based on credit rating and generating a basic quota value; the statistical model is used for analyzing a supply chain behavior sequence and transaction time sequence characteristics through a graph neural network and outputting a risk correction coefficient; event-triggered risk control response: monitoring market fluctuation events and user transaction behaviors in real time; and compliance self-adaptive output is carried out, and a credit granting result is displayed through a visual billboard.

Description

Technical Field

[0001] This invention relates to the field of financial technology, and more specifically, to credit assessment and risk management technology. Background Technology

[0002] Because customer credit data in the financial sector is scattered across different systems such as finance, transactions, and credit reporting, existing technologies cannot correlate and analyze the data globally. Traditional models rely on fixed-period updates to credit limits, which are out of touch with dynamic needs and cannot respond in real time to market fluctuations or changes in user behavior. Furthermore, they rely on expert experience to adjust rules, which is highly subjective and difficult to standardize. Summary of the Invention

[0003] The purpose of this invention is to provide a method, device, and storage medium for dynamically calculating credit limits based on a funding system, in order to solve the problems of data silos, information fragmentation, and disconnection from dynamic needs in the existing technology.

[0004] To address the aforementioned issues, this invention first provides a dynamic credit limit calculation method based on a funding system, comprising the following steps: multi-source data fusion processing, using a dynamic data engine to collect structured and unstructured data from financial systems, credit reporting platforms, and business systems in real time; converting unstructured data into quantitative values ​​through semantic analysis and automatically mapping relative time parameters to absolute time windows; hybrid model collaborative calculation, evaluating the model, dynamically adjusting indicator weights based on real-time economic cycle parameters, calculating weighted scores for quantifiable indicators and semantically converted unquantifiable indicators, and generating a credit rating; calculation model: performing formula calculations based on the credit rating to generate a basic credit limit value; statistical model: analyzing supply chain behavior sequences and transaction time sequence characteristics through graph neural networks, outputting risk correction coefficients; event-triggered risk control response, real-time monitoring of market fluctuation events and user transaction behavior; if a predefined risk threshold is triggered, calling distributed computing nodes to recalculate the credit limit and automatically freezing high-risk accounts; compliance adaptive output, displaying credit granting results and SHAP value analysis reports through a visual dashboard; using natural language processing to parse regulatory new rules text, matching compliance rules from a strategy template library, and updating model parameters.

[0005] Furthermore, the dynamic data engine is characterized by comprising: an unstructured data processing unit that maps qualitative evaluations into numerical indicators through text semantic analysis and extracts the temporal characteristics of transaction behavior; and a time dimension adaptive unit that automatically parses relative time parameters based on the current time and generates data slices within a dynamic time window.

[0006] Furthermore, the event-triggered risk control response mechanism includes: real-time transaction requests are diverted based on Kafka message queues, supporting tens of millions of concurrent processing requests; the rule engine predefines event types, including: market fluctuations exceeding thresholds, abnormal daily transaction frequency, and risk transmission from supply chain stakeholders.

[0007] This invention also provides a dynamic credit limit calculation device based on a funding system, used to implement the dynamic credit limit calculation method based on a funding system as described in any of the above technical solutions, comprising: a data input layer, including a multi-source data acquisition interface and a dynamic data engine for data acquisition and fusion; a core calculation layer, integrating an evaluation model unit, a calculation model unit, and a statistical model unit, respectively performing rating calculation, formula calculation, and risk correction functions; a dynamic risk control layer, including a real-time monitoring module and a distributed computing node cluster, which implements credit limit recalculation and account freezing based on an event triggering mechanism; and an output layer, including a visualization dashboard, a compliance report generation module, and a business system callback interface, used to output credit granting results and risk analysis data.

[0008] Furthermore, the core computing layer also includes an indicator grouping manager, which classifies and manages financial indicators, transaction behavior indicators, and credit indicators, and configures independent weight adjustment strategies for different indicator types.

[0009] The present invention also provides an electronic device, comprising: a processor, and a memory connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the dynamic credit limit calculation method based on the financial system as described in any of the above technical solutions.

[0010] The present invention also provides a computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the dynamic credit limit calculation method based on the financial system as described in any of the above technical solutions.

[0011] The present invention provides a method, device, and storage medium for dynamic calculation of credit limits based on a funding system. By breaking down data silos from multiple sources through multiple dynamic data engines, it achieves real-time fusion of structured and unstructured data and automatic mapping of time dimensions, thus constructing a comprehensive risk perception foundation. The combination of dynamic weight adjustment mechanism and graph neural network modeling improves the accuracy and timeliness of credit assessment. Event-triggered distributed risk control achieves second-level response and automatically blocks high-risk transactions. The driving strategy template library enables intelligent adaptation to new regulatory rules, ultimately forming an intelligent credit granting closed-loop system that integrates data fusion, dynamic modeling, real-time risk control, and compliance self-iteration. Detailed Implementation

[0012] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0013] Example 1

[0014] This embodiment provides a method for dynamically calculating credit limits based on a funding system. The application scenario is a commercial bank approving and dynamically managing credit limits for corporate clients. The specific steps are as follows:

[0015] I. Multi-source data fusion processing operation

[0016] 1. Data Acquisition and Access:

[0017] By using the API interface of the dynamic data engine, structured data such as the balance sheet and profit and loss statement of the enterprise for the past three years can be obtained from the bank's core financial system;

[0018] Connect to the central bank's credit reporting platform to collect corporate credit reports, including loan records and default history;

[0019] Real-time monitoring of business system transaction flow, obtaining unstructured data such as account transaction records and transfer records for the past 6 months, such as transaction logs in JSON format;

[0020] Manually fill in supplementary information, such as the qualitative evaluation of the enterprise's qualifications as an "industry leader".

[0021] 2. Quantization processing of unstructured data

[0022] The unstructured data processing unit of the dynamic data engine performs semantic analysis on "industry leading enterprises" and matches them with a preset mapping rule library. For example, "industry leading enterprise" corresponds to 90 points and is converted into a quantitative value.

[0023] Extract time-series features from transaction logs, such as daily transaction time distribution and transfer interval frequency, and generate time-series feature vectors, such as 80% of transactions occurring between 9:00 and 11:00 on weekdays.

[0024] 3. Relative time parameter mapping

[0025] When a company applies for "transaction records for the past 12 months", the time-dimensional adaptive unit automatically parses "the past 12 months" as June 2024 to May 2025 based on the current system time, such as June 2025, and generates data slices for the corresponding time window.

[0026] Store data uniformly in a data platform and establish a standardized data model indexed by enterprise ID, such as uniformly naming the "annual income" field as "annual income".

[0027] II. Collaborative Computation Operations in Hybrid Models

[0028] 1. Evaluation Model Execution Process

[0029] Obtain real-time economic cycle parameters, such as the current GDP growth rate of 5.2% and the medium risk index of the industry, and dynamically adjust the weights of indicators: the debt ratio weight is increased from the default 30% to 35% to strengthen the assessment of debt repayment capacity during economic downturns; the income growth rate weight is reduced from 25% to 20%.

[0030] For quantitative indicators, such as a debt ratio of 40% corresponding to a score of 70 points and a revenue growth rate of 8% corresponding to a score of 85 points, and semantically transformed indicators, such as enterprise qualification (90 points), the weighted score is calculated as follows:

[0031] Total score = 70 × 35% + 85 × 20% + 90 × 45% = 82 points, generating a credit rating of "AA".

[0032] 2. Calculation of model formulas

[0033] The basic credit limit is calculated based on a credit rating of "AA". The basic credit limit is calculated as follows: Basic credit limit = annual income of the enterprise × industry coefficient × rating coefficient; Assuming an annual income of 10 million yuan, an industry coefficient of 1.2, and an AA rating coefficient of 0.9, the basic credit limit is 10 million × 1.2 × 0.9 = 10.8 million yuan.

[0034] Handling abnormal scenarios: If the company's annual revenue is 0, return the default limit of 0 yuan through the ternary logic.

[0035] 3. Statistical model risk correction

[0036] Graph neural network analysis of enterprise supply chain behavior sequences, such as discovering that its upstream core supplier has three recent records of delayed delivery, constructs an "enterprise-supplier" relationship graph;

[0037] Analyze the timing characteristics of transactions, such as a sudden increase in the frequency of daily transactions by enterprises in the past month, and an increase in the proportion of nighttime transactions from 10% to 40%;

[0038] The output risk correction coefficient is 0.8. In high-risk scenarios, the credit limit is reduced by 20%. The final credit limit is 1080 × 0.8 = 8.64 million yuan.

[0039] III. Event-triggered risk control response operations

[0040] 1. Real-time event monitoring:

[0041] Kafka message queues offload real-time transaction requests from enterprises, such as the 10,000th transaction of the day. The rules engine listens for the following events:

[0042] Market volatility: The industry index fell by more than 7% in a single day, triggering the threshold;

[0043] Transaction anomaly: The company made 25 transfers in a single day, compared to a historical average of 5.

[0044] Related party risk: Core enterprises in the supply chain are included in the list of dishonest entities.

[0045] 2. Risk event handling:

[0046] When a company makes 25 transfers in a single day, a predefined threshold (≥20 times) is triggered. The distributed computing node cluster calls the hybrid model in parallel to recalculate the credit limit: the evaluation model recalculates the score (the weight of the transaction anomaly indicator is increased to 30%); the calculation model generates a new basic credit limit; and the statistical model outputs a correction coefficient of 0.5 (high risk).

[0047] The recalculated amount was 5 million yuan, a 42% reduction from the original limit of 8.64 million yuan. The system automatically froze the account and sent an anomaly report (including transaction details) to the account manager.

[0048] IV. Compliant Adaptive Output Operation

[0049] 1. Visualization of credit granting results

[0050] Visual dashboard display: final credit line of 5 million yuan; SHAP value analysis report (transaction anomaly indicator contribution of 45%, debt ratio contribution of 30%); risk warning details (reasons for abnormal transaction frequency).

[0051] 2. Adaptation to new regulatory rules

[0052] For example, the People's Bank of China issued new anti-money laundering regulations requiring that "single transactions exceeding 500,000 yuan must be reported in real time";

[0053] Natural language processing was used to analyze the new regulations text and extract the keywords "single transaction of 500,000 yuan" and "real-time reporting";

[0054] Match anti-money laundering rule templates from the strategy template library, automatically update the transaction limit parameters in the calculation model (the original threshold of 1 million yuan is adjusted to 500,000 yuan), and synchronize them to the limit calculation logic of all relevant customers.

[0055] Example 2

[0056] This embodiment provides a dynamic credit limit calculation device based on a funding system, which can implement the method provided in Embodiment 1. Taking an internet finance platform as an example, the device architecture includes:

[0057] 1. Data Input Layer

[0058] The multi-source data acquisition interface simultaneously accesses e-commerce platform transaction data, third-party payment records, and enterprise business registration information;

[0059] The dynamic data engine automatically maps "last month" to May 2025, maps the qualitative evaluation of "stable business conditions" to 75 points, and generates standardized data platform records.

[0060] 2. Core Computing Layer

[0061] The indicator group manager categorizes e-commerce transaction frequency as a "transaction behavior indicator" and sets a real-time weight of +15%.

[0062] The evaluation model reduces the debt ratio weight to 25% and increases the transaction activity weight to 30% based on the current consumer credit policy (easing period).

[0063] The statistical model uses a graph neural network to discover that the actual controller of the enterprise has financial transactions with high-risk accounts, outputs a correction coefficient of 0.6, and finally reduces the limit by 40%.

[0064] 3. Dynamic Risk Control Layer

[0065] The real-time monitoring module detected that a corporate account transferred 600,000 yuan to an unknown account (exceeding the new regulation threshold of 500,000 yuan), triggering Kafka message routing. The distributed nodes completed the recalculation of the limit and froze the account within 100ms.

[0066] 4. Output layer

[0067] The compliance report generation module automatically generates anti-money laundering filing documents, which are then redirected to the business system to trigger a manual review process. A visual dashboard simultaneously displays the credit limit adjustment trajectory and compliance basis.

[0068] Example 3

[0069] This embodiment provides an electronic device, including: a processor and a memory connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the dynamic credit limit calculation method based on the funding system provided in Embodiment 1.

[0070] Example 4

[0071] This embodiment provides a computer-readable storage medium storing computer-executable instructions. When executed, these instructions are used to implement the dynamic credit limit calculation method based on a funding system provided in this embodiment.

[0072] This invention achieves multi-dimensional technological breakthroughs by utilizing core technologies such as dynamic data fusion, hybrid model calculation, real-time risk control response, and compliance adaptation to address the shortcomings of traditional credit assessment. Specific technical effects are as follows:

[0073] I. Improved Data Integration Efficiency and Utilization

[0074] Breaking down data silos and achieving full-dimensional data integration; real-time collection of multi-source data such as financial, credit, and transaction flows through a dynamic data engine; unified modeling through a data platform to transform unstructured data into quantitative indicators and automatically parse relative time into absolute time windows; improving data integration efficiency and utilization of unstructured data, and discovering hidden risk signals ignored by traditional models, such as abnormal transaction times and defaults by supply chain stakeholders.

[0075] Unified data standards improve data usability; the data standardization module maps multi-source heterogeneous data to a unified model, resolving inconsistencies in field naming. Data cleaning costs are reduced, cross-system correlation analysis efficiency is improved, and a standardized data foundation is provided for real-time credit assessment.

[0076] II. Breakthroughs in the Accuracy and Real-Timeliness of Credit Assessment

[0077] Dynamic weight adjustments adapt to real-time scenario changes; the evaluation model dynamically adjusts indicator weights based on real-time economic cycle parameters (such as GDP growth rate and industry risk index) (e.g., increasing the debt ratio weight by 10% during economic downturns), and combines SHAP values ​​to analyze indicator contribution. The credit assessment model's AUC value increases, the false judgment rate decreases, and it accurately captures sudden industry risks, such as automatically reducing credit limits for customers in a particular industry when policies are adjusted.

[0078] The hybrid model works in synergy to cover multi-dimensional assessment needs; evaluation, calculation, and statistical models work together, using graph neural networks to analyze supply chain transaction sequences and identify complex risks such as joint guarantee fraud. The expanded dimensions of credit limit calculation indicators enhance the ability to identify hidden risks, such as uncovering a company's covert asset transfer activities through related parties.

[0079] III. Risk Control Response Speed ​​and Automation Upgrade

[0080] Event-triggered real-time risk control enables second-level risk response. It utilizes a Kafka queue to distribute millions of concurrent transaction requests, and a rules engine predefines risk events. Upon triggering, the distributed node cluster recalculates the credit limit and freezes the account. Risk response is shortened from traditional daily report alerts to real-time responses, improving the interception rate of high-risk transactions. For example, if a customer's daily transfers exceed a threshold, the system automatically freezes the account to prevent financial losses.

[0081] The distributed computing architecture supports high-concurrency scenarios; it employs a distributed node cluster to process transaction stream data in parallel, and a Kafka message queue to distribute requests. This improves system performance, enabling credit limit calculations for tens of millions of customers simultaneously online, such as a bank's real-time risk control with zero latency during peak periods, handling 100,000 transactions per second.

[0082] IV. Compliance Adaptation Efficiency and Cost Optimization

[0083] Automating compliance significantly shortens the adaptation cycle; natural language processing parses new regulatory rules, such as adjustments to anti-money laundering limits, automatically generating compliance rules and updating model parameters from the strategy template library, eliminating the need for manual code modification. Compliance strategy adaptation time is reduced, and operational costs are lowered. For example, after the People's Bank of China released new anti-money laundering regulations, the system quickly updated the limit calculation rules for all customers.

[0084] Visualized compliance reports improve audit efficiency; the output layer generates SHAP value analysis reports and compliance reports, clarifying the legal basis for quota adjustments. Internal audit time is shortened, meeting regulatory requirements for "decision explainability" and reducing compliance risks.

[0085] V. Overall Business Value Enhancement

[0086] Credit approval efficiency is improved, and business processes are accelerated; the time taken from data collection to credit limit generation is shortened, supporting banks' "second-level credit granting" business scenarios.

[0087] Enhanced business value and competitiveness; reduced bad debt rate through dynamic credit limit management, while increasing credit limits for high-quality customers in real time, promoting loan disbursement growth, and helping financial institutions achieve a balance between returns and risks while complying with regulations.

[0088] This invention achieves a leap from post-event assessment to real-time risk control, and from manual rules to intelligent adaptation, comprehensively solving the problems of efficiency, accuracy, and compliance in credit assessment in the financial field, and possesses significant technological advancement and commercial application value.

[0089] In the description of this embodiment, it should be noted that those skilled in the art will understand that all or part of the processes in the above-described embodiments can be implemented by a computer program instructing a control device. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the above-described method embodiments. The storage medium can be a memory, a disk, an optical disk, etc.

[0090] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

[0091] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0092] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0093] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for dynamically calculating credit limits based on a funding system, characterized in that, The following steps are involved: Multi-source data fusion processing uses a dynamic data engine to collect structured and unstructured data from financial systems, credit reporting platforms, and business systems in real time; unstructured data is converted into quantitative values ​​through semantic analysis and automatically mapped relative time parameters to absolute time windows; The hybrid model performs collaborative calculations and evaluates the model. It dynamically adjusts the indicator weights based on real-time economic cycle parameters, calculates weighted scores for quantifiable indicators and semantically transformed non-quantifiable indicators, and generates credit ratings. Computational model: Based on credit rating, a formula is executed to generate a basic credit limit value; Statistical model: Through graph neural network analysis of supply chain behavior sequences and transaction time sequence characteristics, a risk correction coefficient is output. Event-triggered risk control response, real-time monitoring of market fluctuations and user trading behavior; If a predefined risk threshold is triggered, the distributed computing node will be invoked to recalculate the credit limit and high-risk accounts will be automatically frozen. Compliance-adaptive output, displaying credit granting results and SHAP value analysis reports through a visual dashboard; Natural language processing is used to parse the text of new regulatory rules, match compliance rules from a policy template library, and update model parameters.

2. The method for dynamically calculating credit limits based on a funding system according to claim 1, characterized in that, The dynamic data engine includes: The unstructured data processing unit maps qualitative evaluations into numerical indicators and extracts the time-series features of transaction behavior through text semantic analysis. The time-dimensional adaptive unit automatically parses relative time parameters based on the current time and generates data slices within a dynamic time window.

3. The method for dynamically calculating credit limits based on a funding system according to claim 1, characterized in that, The event-triggered risk control response mechanism includes: Real-time transaction requests are distributed based on Kafka message queues, supporting tens of millions of concurrent processing requests; The rules engine predefines event types, including: market volatility exceeding thresholds, abnormal daily transaction frequency, and risk transmission from supply chain stakeholders.

4. A dynamic credit limit calculation device based on a funding system, characterized in that, The method for dynamically calculating credit limits based on a funding system as described in any one of claims 1-3 includes: The data input layer includes multi-source data acquisition interfaces and a dynamic data engine for data acquisition and fusion. The core computing layer integrates evaluation model units, calculation model units, and statistical model units, which respectively perform rating calculations, formula calculations, and risk correction functions. The dynamic risk control layer includes a real-time monitoring module and a distributed computing node cluster, which implements credit limit recalculation and account freezing based on an event triggering mechanism; The output layer includes a visual dashboard, a compliance report generation module, and business system callback interfaces, used to output credit granting results and risk analysis data.

5. The dynamic credit limit calculation device based on the capital system according to claim 4, characterized in that, The core computing layer also includes: The indicator group manager categorizes and manages financial indicators, transaction behavior indicators, and credit indicators, and configures independent weight adjustment strategies for different indicator types.

6. An electronic device, characterized in that, include: A processor, and a memory connected to the processor; The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory to implement the dynamic credit limit calculation method based on the funding system as described in any one of claims 1-3.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed, are used to implement the dynamic calculation method for credit limits based on a funding system as described in any one of claims 1-3.