Credit strategy intelligent optimization method and device based on digital twin driving

Abstract

The invention relates to an intelligent credit strategy optimization method and device based on digital twinborn driving, and the method comprises the steps: integrating internal financial data and external associated data based on a multi-source data fusion technology, obtaining multi-source data, and constructing a digital twinborn body based on the multi-source data and a knowledge graph technology; inputting the real-time variable data through a preset credit evaluation model and a digital twinborn simulation technology based on the digital twinborn body, and obtaining a risk evaluation result matched with the business scene; constructing a risk scene library based on a risk assessment result and the digital twin, simulating implementation effects of different credit strategies, and generating an optimal strategy combination; based on the optimal strategy combination and a preset credit target, utilizing a particle swarm optimization algorithm to generate a dynamic adjustment scheme including a credit line, interest rate pricing, guarantee conditions and a post-loan management strategy; and collecting feedback data of service execution, synchronizing the feedback data to the digital twin, and correcting the deviation between the virtual model and the physical service.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of risk control, in particular to a credit strategy intelligent optimization method and device based on digital twin driving. BACKGROUND

[0002] With the development of social economy, more and more small and micro enterprises are established, and the loan requests of small and micro enterprises to banks also increase. At present, when facing the loan requests of small and micro enterprises, the bank makes credit decisions based on reference indexes such as national policies, enterprise transaction invoices and supply and demand sides of upstream and downstream enterprises of the enterprise.

[0003] Traditional credit risk assessment mainly relies on limited structured data such as enterprise financial statements and personal credit reports, and a fixed quantitative model is used to statically calculate the debt repayment ability and credit level of a customer, and then a credit manager supplements unstructured information (such as industry reputation of the enterprise owner, related transaction background, sudden operating events, etc.) based on experience to complete the final audit.

[0004] The above credit assessment model is difficult to capture the dynamic influence of external variables such as customer behavior, industry environment and supply chain, and excessively relies on manual audit, and is not suitable for enterprise loans with increasing scale. SUMMARY

[0005] In order to improve the problem that the credit assessment model is difficult to capture the dynamic influence of external variables such as customer behavior, industry environment and supply chain, excessively relies on manual audit, and is not suitable for enterprise loans with increasing scale, the application provides a credit strategy intelligent optimization method and device based on digital twin driving.

[0006] In a first aspect, the application provides a credit strategy intelligent optimization method based on digital twin driving, which adopts the following technical solution: comprising: Based on multi-source data fusion technology, internal financial data and external associated data are integrated to obtain multi-source data, and a digital twin body is constructed based on the multi-source data and knowledge graph technology; Based on the digital twin body, through a preset credit assessment model and digital twin simulation technology, real-time variable data is input to obtain a risk assessment result adapted to a business scenario; Based on the risk assessment result and the digital twin body, a risk scenario library is constructed to simulate the implementation effect of different credit strategies and generate an optimal strategy combination; Based on the optimal strategy combination and a preset credit target, a particle swarm optimization algorithm is used to generate a dynamic adjustment scheme including credit limit, interest rate pricing, guarantee conditions and post-loan management strategies; Collecting feedback data of the business execution, synchronizing the feedback data to the digital twin, and correcting the deviation between the virtual model and the physical business.

[0007] Preferably, the multi-source data fusion technology integrates internal financial data and external associated data to obtain multi-source data, and constructs a digital twin based on the multi-source data and knowledge graph technology, including: The internal financial data and the external associated data are subjected to data cleaning and standardization processing, including missing value filling, abnormal value denoising, unit unification and data normalization, to form a structured data set and obtain the multi-source data. Based on the knowledge graph technology, the customer basic information, risk characteristics, industry influence information and business scenario information in the multi-source data are subjected to semantic association modeling to construct the digital twin, and the digital twin realizes dynamic data interaction through the association relationship between the knowledge graph nodes and edges.

[0008] Preferably, based on the digital twin, real-time variable data is input through a preset credit evaluation model and digital twin simulation technology to obtain a risk evaluation result adapted to the business scenario, including: Based on the digital twin, the credit evaluation model and the digital twin simulation technology are fused to generate a dynamic evaluation framework. Based on the three-dimensional mapping characteristics of the digital twin, the real-time variable data is input into the dynamic evaluation framework, and an initial risk evolution path is simulated and output through simulation calculation, and whether the dynamic evaluation framework is preset with a risk logic is verified. If the dynamic evaluation framework conforms to the risk logic, the business scenario characteristics are obtained through the digital twin, the real-time feedback data and the initial risk evolution path are combined, and the model parameters in the dynamic evaluation framework are adaptively adjusted. The adjusted dynamic evaluation framework is subjected to simulation verification, and an ultimate risk evolution path is output, and the matching degree between the ultimate risk evolution path and the risk characteristics in the business scenario characteristics is calculated. If the matching degree is greater than a preset threshold, the risk evaluation result adapted to the business scenario is output.

[0009] Preferably, the matching degree between the ultimate risk evolution path and the risk characteristics in the business scenario characteristics is calculated, including: The Euclidean distance is used to calculate the difference value between the default probability distribution of the adjusted path and the historical actual default probability distribution of the business scenario. The coverage rate of the nodes in the ultimate risk evolution path to a preset core node is calculated. screening a plurality of target samples most similar to the business scenario characteristics in the digital twin; integrating historical default data, real-time behavior data and industry data corresponding to the target samples according to the digital twin, and extracting key features; determining an actual risk level according to preset risk level quantification rules and the key features; simulating a predicted default probability of the business scenario characteristics according to the dynamic evaluation framework, mapping the predicted default probability to a predicted risk level, and calculating a level difference between the actual risk level and the predicted risk level; if the difference value, the coverage rate and the level difference meet preset matching conditions, outputting the risk assessment result adapted to the business scenario.

[0010] Preferably, a risk scenario library is constructed based on the risk assessment result and the digital twin, the implementation effect of different credit strategies is simulated, and an optimal strategy combination is generated, including: a risk scenario library covering multi-dimensional risk features is constructed according to the risk assessment result, the historical default data, the real-time behavior data and the industry data; a plurality of candidate credit strategies are defined based on credit business targets, the credit strategies at least including guarantee pricing, credit limit, interest rate pricing and term structure; each group of the credit strategies is input into the risk scenario library for different risk level scenario stress testing based on the simulation capability of the digital twin, and a simulation result is output; the simulation result is comprehensively evaluated based on preset credit conditions, the optimal strategy combination is screened, and the optimal strategy combination is stored in association with the business scenario characteristics, forming associated data that can be dynamically invoked.

[0011] Preferably, a dynamic adjustment scheme including credit limit, interest rate pricing, guarantee condition and post-loan management strategy is generated based on the optimal strategy combination and preset credit targets by using a particle swarm optimization algorithm, including: initial parameter ranges of a particle swarm are defined based on the optimal strategy combination; all initial parameter combinations are input into the risk scenario library to simulate performances under different risk scenarios, risk indicators and income indicators of each scheme are calculated, and an evaluation score is generated; the best scheme in the history of each particle is recorded, and the optimal scheme with the best performance among all particles is recorded; parameters of each particle are adjusted according to the best scheme of the individual and the optimal scheme of the global; When the preset round of adjustment is repeated or a more optimal scheme does not appear in a plurality of consecutive rounds, the optimization is stopped, and the current global optimal scheme is set as the dynamic adjustment scheme.

[0012] Preferably, the feedback data collected by the business execution is synchronized to the digital twin, and the deviation between the virtual model and the physical business is corrected, including: After the feedback data is collected and standardized, it is synchronized to the database of the digital twin; According to the simulation prediction value in the digital twin and the actual observation value of the feedback data, a deviation index is calculated, including: risk deviation, return deviation and risk path deviation; According to the deviation index, the feature weight of the virtual model of the digital twin is adjusted.

[0013] In a second aspect, the application provides a credit strategy intelligent optimization device based on digital twin driving, which adopts the following technical scheme, including: A data acquisition module is configured to integrate internal financial data and external associated data based on multi-source data fusion technology to obtain multi-source data, and construct a digital twin based on the multi-source data and knowledge graph technology; A risk assessment module is configured to input real-time variable data based on the digital twin through a preset credit evaluation model and digital twin simulation technology to obtain a risk assessment result adapted to a business scenario; A strategy combination module is configured to construct a risk scenario library based on the risk assessment result and the digital twin, simulate the implementation effect of different credit strategies, and generate an optimal strategy combination; A dynamic adjustment module is configured to generate a dynamic adjustment scheme including credit limit, interest rate pricing, guarantee conditions and post-loan management strategies based on the optimal strategy combination and a preset credit target using a particle swarm optimization algorithm; A physical correction module is configured to collect feedback data of business execution, synchronize the feedback data to the digital twin, and correct the deviation between the virtual model and the physical business.

[0014] In a third aspect, the application also provides a control device, which includes: The device includes a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform the above-mentioned credit strategy intelligent optimization method based on digital twin driving.

[0015] In a fourth aspect, the application also provides a computer readable storage medium storing a computer program capable of being loaded and executed by a processor to perform the above-mentioned credit strategy intelligent optimization method based on digital twin driving.

[0016] In summary, in this application, first, the internal and external data are integrated by multi-source data fusion and knowledge graph technology to construct a digital twin (data cleaning, standardization and semantic association modeling), realizing data structuring and dynamic interaction; then, the credit evaluation model and simulation technology are fused by the digital twin to input real-time variable data to generate risk assessment results suitable for the business scene (dynamic assessment framework adjustment, risk evolution path matching degree calculation); then, based on the risk assessment results, a risk scene library covering multi-dimensional risk characteristics is constructed to simulate the implementation effect of different credit strategies and select the optimal strategy combination; then, based on the optimal strategy, the particle swarm algorithm is used to optimize the generation of dynamic adjustment schemes such as credit limit and interest rate; finally, the business feedback data is collected and synchronized to the digital twin, and the virtual model parameters are corrected by calculating the risk, return and other deviation indicators to ensure that the model is synchronized with the physical business. Thus, it is convenient to capture the dynamic influence of external variables such as customer behavior, industry environment and supply chain, reduce the dependence on manual review, and further improve the intelligence, accuracy and adaptability of credit strategies, so as to efficiently manage credit business. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 FIG. 1 is a flow diagram of a credit strategy intelligent optimization method based on digital twin driving.

[0018] Figure 2 FIG. 2 is a structural block diagram of a credit strategy intelligent optimization device based on digital twin driving. DETAILED DESCRIPTION

[0019] The following Figure 1 - Figure 2 The application will be further described in detail.

[0020] With the acceleration of digital transformation of the financial industry, credit business puts forward higher requirements for the accuracy of risk control and the dynamic nature of strategy adjustment. Traditional credit strategy relies on historical data modeling, which has three pain points: first, risk assessment is mostly based on static indicators, which is difficult to capture the complex influence of real-time variables such as customer behavior and industry policy; second, strategy optimization relies on human experience, which lacks coverage of multi-dimensional risk scenarios, and is prone to imbalance between returns and risks; third, there is no effective feedback mechanism after strategy implementation, and the deviation between virtual model and actual business is difficult to correct in time, which may reduce the adaptability of the strategy in the long run. Digital twin technology provides a new path for dynamic simulation, prediction and optimization by constructing a virtual mapping body of physical business.

[0021] Referring to Figure 1 , the embodiments of the application at least include steps S10 to S50.

[0022] S10, based on multi-source data fusion technology, integrating internal financial data and external associated data to obtain multi-source data, and based on multi-source data and knowledge graph technology to construct a digital twin.

[0023] S20, based on the digital twin, inputting real-time variable data through a preset credit evaluation model and digital twin simulation technology to obtain a risk assessment result adapted to the business scenario.

[0024] S30, based on the risk assessment result and the digital twin, constructing a risk scenario library to simulate the implementation effect of different credit strategies and generating an optimal strategy combination.

[0025] S40, based on the optimal strategy combination and the preset credit target, using a particle swarm optimization algorithm to generate a dynamic adjustment scheme including credit limit, interest rate pricing, guarantee conditions and post-loan management strategies.

[0026] S50, collecting feedback data of business execution, synchronizing the feedback data to the digital twin, and correcting the deviation between the virtual model and the physical business.

[0027] Specifically, first, integrate internal and external multi-source data and use knowledge graph to construct a digital twin to form a virtual mapping of physical business; then, based on the digital twin, integrate the credit evaluation model and the simulation technology, dynamically evaluate the business scenario risk combined with real-time variable data; then, rely on the risk assessment result and the digital twin simulation capability to construct a risk scenario library, simulate the implementation effect of different credit strategies (such as credit limit, interest rate pricing, etc.) and select the optimal combination; then, based on the optimal strategy, generate a dynamic adjustment scheme including credit limit, interest rate, guarantee conditions, etc. through particle swarm algorithm optimization; finally, collect feedback data of business execution and synchronize to the digital twin to correct the deviation between the virtual model and the actual business. The whole forms a complete chain of "data fusion-risk assessment-strategy optimization-dynamic adjustment-closed loop correction", realizes the precise, dynamic and intelligent management of credit strategy, and improves the risk control and income balance ability.

[0028] In some embodiments, step S10 specifically includes the following steps: data cleaning and standardization processing of internal financial data and external associated data, the standardization processing includes missing value filling, abnormal value denoising, unit unification and data normalization, forming a structured data set to obtain multi-source data; based on knowledge graph technology, model the semantic association of customer basic information, risk characteristics, industry influence information and business scenario information in multi-source data to construct a digital twin, and realize dynamic data interaction through the association relationship of knowledge graph nodes and edges in the digital twin.

[0029] Specifically, through data cleaning and standardization processing and knowledge graph modeling, the basic construction of digital twin is completed: first, the internal financial data and external associated data are subjected to standardization processing such as missing value filling, abnormal value denoising, unit unification and data normalization, so as to eliminate the noise and inconsistency of the original data, form a structured multi-source data set, and ensure the data quality and availability; then, based on the knowledge graph technology, the customer basic information, risk characteristics, industry influence information and business scene information in the multi-source data are modeled through semantic association, a dynamic association is established in the form of nodes (entities) and edges (relationships), and a digital twin reflecting the real state of physical business is constructed. The role is to improve the data reliability through data cleaning and standardization, realize the dynamic interaction of multi-dimensional information through the semantic association of the knowledge graph, and provide accurate virtual mapping basis for subsequent risk assessment, strategy optimization and the like based on the digital twin.

[0030] In some embodiments, step S20 specifically comprises the following steps: based on the digital twin, the credit assessment model is fused with the digital twin simulation technology to generate a dynamic assessment framework; based on the three-dimensional mapping characteristics of the digital twin, real-time variable data is input into the dynamic assessment framework, an initial risk evolution path is simulated and output through simulation calculation, and whether the dynamic assessment framework conforms to the preset risk logic is verified; if the dynamic assessment framework conforms to the risk logic, the business scene characteristics are obtained through the digital twin, the real-time feedback data and the initial risk evolution path are combined, and the model parameters in the dynamic assessment framework are adaptively adjusted; the adjusted dynamic assessment framework is simulated and verified, an ultimate risk evolution path is output, and the matching degree between the ultimate risk evolution path and the risk characteristics in the business scene characteristics is calculated; if the matching degree is greater than a preset threshold, a risk assessment result adapted to the business scene is output.

[0031] Specifically, first, the credit assessment model is combined with the simulation technology based on the digital twin to generate a dynamically adjustable assessment framework; real-time variable data is input into the framework using the three-dimensional mapping characteristics of the digital twin, an initial risk evolution path is simulated through simulation calculation, and whether the framework conforms to the preset risk logic is verified; if the logic is met, the model parameters in the framework are adaptively adjusted in combination with the business scene characteristics obtained through the digital twin, the real-time feedback data and the initial path; the adjusted framework is simulated and verified again, an ultimate risk evolution path is output, and the matching degree between the ultimate risk evolution path and the risk characteristics of the business scene is calculated; when the matching degree exceeds the threshold, a risk assessment result adapted to the business scene is output. The role is to ensure that the risk assessment result conforms to the preset logic and can dynamically adapt to the changes in the business scene through dynamic framework construction, real-time data input, parameter adaptive adjustment and multiple verifications, so as to improve the accuracy and scene adaptability of risk assessment.

[0032] In some embodiments, step S20 further comprises the following steps: calculating a difference value between the adjusted path's default probability distribution and the historical actual default probability distribution of the business scenario using the Euclidean distance calculation; calculating the coverage rate of the nodes in the final risk evolution path to the preset core nodes; selecting a number of target samples most similar to the business scenario characteristics from the digital twin; integrating the historical default data, real-time behavior data, and industry data corresponding to the target samples according to the digital twin to extract key features; determining the actual risk level according to the preset risk level quantification rules and the key features; simulating the predicted default probability of the business scenario characteristics according to the dynamic evaluation framework, and mapping the predicted default probability to a predicted risk level, and calculating the level difference between the actual risk level and the predicted risk level; if the difference value, the coverage rate, and the level difference meet the preset matching conditions, output the risk assessment result adapted to the business scenario.

[0033] Specifically, first, the difference value between the adjusted path's default probability distribution and the historical actual distribution is calculated using the Euclidean distance calculation, quantifying the matching degree at the probability level; then, the coverage rate of the nodes in the final risk evolution path to the preset core nodes is calculated to ensure that the path covers key risk points; subsequently, the target samples most similar to the current business scenario are selected from the digital twin, and their historical default data, real-time behavior data, and industry data are integrated to extract key features; the actual risk level is determined based on the risk level quantification rules and the key features, and the predicted default probability is simulated through the dynamic evaluation framework and mapped to a predicted risk level, and the level difference between the two is calculated; if the difference value, the coverage rate, and the level difference all meet the preset matching conditions, the risk assessment result adapted to the business scenario is output. Its role is to ensure that the risk assessment result is highly consistent with the actual business scenario through multi-dimensional (probability distribution, node coverage, sample comparison, level verification) quantitative verification, and to improve the scientificity and accuracy of the evaluation.

[0034] In some embodiments, step S30 specifically comprises the following steps: constructing a risk scenario library covering multi-dimensional risk features according to the risk assessment result, historical default data, real-time behavior data, and industry data; defining multiple groups of candidate credit strategies based on the credit business target, the credit strategies at least including: guarantee pricing, credit limit, interest rate pricing, and term structure; inputting each group of credit strategies into the risk scenario library for different risk level scenario stress testing based on the simulation capability of the digital twin, and outputting the simulation results; comprehensively evaluating the simulation results based on the preset credit conditions, selecting the optimal strategy combination, and storing the optimal strategy combination in association with the business scenario characteristics to form associated data that can be dynamically invoked.

[0035] Specifically, based on the risk assessment results, historical default data, real-time behavior data and industry data, a risk scenario library covering multi-dimensional risk characteristics is constructed; then according to the credit business target, a plurality of candidate strategies including guarantee pricing, credit limit, interest rate pricing, term structure, etc. are defined; by using the simulation capability of the digital twin, each group of strategies is input into the risk scenario library for stress testing of different risk levels, and the simulation effect is output; and based on the preset credit conditions, the simulation results are comprehensively evaluated to select the optimal strategy combination, which is stored in association with the business scenario characteristics to form associated data that can be dynamically called. Its role is to simulate the implementation effect of different strategies through multi-dimensional risk scenario, accurately select the optimal strategy combination that adapts to the business scenario, improve the scientificity and dynamic adaptability of credit strategy formulation, and provide data support for actual business decision-making.

[0036] In some embodiments, step S40 specifically includes the following steps: based on the optimal strategy combination, defining the initial parameter range of the particle swarm; inputting all initial parameter combinations into the risk scenario library to simulate the performance under different risk scenarios, calculating the risk indicators and benefit indicators of each scheme, and generating an evaluation score; recording the best scheme of each particle in history, and recording the optimal scheme with the best performance among all particles; adjusting the parameters of each particle according to the individual best scheme and the global optimal scheme; stopping optimization when a preset number of iterations is repeated or no better scheme appears for a plurality of consecutive rounds, and setting the current global optimal scheme as the dynamic adjustment scheme.

[0037] Specifically, first, the initial parameter range of the particle swarm is defined based on the optimal strategy combination (such as the value interval of parameters such as credit limit and interest rate); the initial parameter combinations are input into the risk scenario library to simulate the performance under different risk scenarios, the risk indicators (such as default probability) and benefit indicators of each scheme are calculated, and a comprehensive evaluation score is generated; the "best scheme" of each particle in history and the "optimal scheme" among all particles globally are recorded; the particle parameters are adjusted according to the individual best and global optimal schemes to guide the search to approach better solutions; when a preset number of iterations is reached or no better scheme appears for a plurality of consecutive rounds, optimization is stopped, and the current global optimal scheme is set as the dynamic adjustment scheme. Its role is to efficiently select a dynamic adjustment scheme that takes into account risk control and benefit targets through swarm intelligence algorithm combined with risk scenario simulation, improving the accuracy and adaptability of credit strategies.

[0038] In some embodiments, step S50 specifically includes the following steps: after collecting feedback data and performing standardization processing, synchronizing to the database of the digital twin; according to the simulation prediction value in the digital twin and the actual observation value of the feedback data, calculating deviation indicators including risk deviation, benefit deviation and risk path deviation; and adjusting the feature weights of the virtual model of the digital twin according to the deviation indicators.

[0039] Specifically, feedback data from business execution is collected and standardized, then synchronized to the digital twin's database. The digital twin's simulated predictions (such as risk level and return) are then compared with the actual observed values ​​of the feedback data. Multi-dimensional deviation indicators, such as risk deviation, return deviation, and risk path deviation, are calculated to quantify the differences between the virtual model and the physical business. Finally, the feature weights of the virtual model are adjusted based on the deviation indicators (i.e., the influence of each risk / return factor in the model is corrected). This approach aims to dynamically calibrate the digital twin's virtual model through real-time feedback data, narrowing the gap between virtual and reality, ensuring the accuracy and reliability of subsequent risk assessments, strategy optimization, and other processes, and making the model more relevant to actual business scenarios.

[0040] The implementation principle of a digital twin-driven intelligent credit strategy optimization method in this application embodiment is as follows: first, a digital twin is constructed by integrating internal and external data through multi-source data fusion and knowledge graph technology (data cleaning, standardization, and semantic association modeling), achieving data structuring and dynamic interaction. Next, the digital twin is integrated with a credit assessment model and simulation technology, inputting real-time variable data to generate risk assessment results adapted to business scenarios (dynamic assessment framework adjustment and risk evolution path matching calculation). Based on the risk assessment results, a risk scenario library covering multi-dimensional risk characteristics is constructed to simulate the implementation effects of different credit strategies and screen the optimal strategy combination. Based on the optimal strategy, a particle swarm algorithm is used to optimize and generate dynamic adjustment plans for credit limits, interest rates, and other factors. Finally, business feedback data is collected and synchronized to the digital twin. The virtual model parameters are corrected by calculating deviation indicators such as risk and return to ensure the model is synchronized with the physical business. This facilitates capturing the dynamic impact of external variables such as customer behavior, industry environment, and supply chain, reduces reliance on manual review, and thereby improves the intelligence, accuracy, and adaptability of credit strategies, facilitating efficient credit business management.

[0041] Figure 1 FIG1 is a flow chart of a method for intelligent optimization of credit strategy based on digital twin drive in one embodiment. It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows; unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps may be executed in other orders; and Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0042] Based on the same technical concept, referring to Figure 2 The embodiment of the application also provides a credit strategy intelligent optimization device based on digital twin driving, which adopts the following technical scheme, the device comprises: A data acquisition module is configured to integrate internal financial data and external associated data based on a multi-source data fusion technology, obtain multi-source data, and construct a digital twin based on the multi-source data and a knowledge graph technology; A risk assessment module is configured to input real-time variable data based on the digital twin, a preset credit assessment model and a digital twin simulation technology, and obtain a risk assessment result adapted to a business scenario; A strategy combination module is configured to construct a risk scenario library based on the risk assessment result and the digital twin, simulate implementation effects of different credit strategies, and generate an optimal strategy combination; A dynamic adjustment module is configured to generate a dynamic adjustment scheme including a credit limit, an interest rate pricing, a guarantee condition and a post-loan management strategy by using a particle swarm optimization algorithm based on the optimal strategy combination and a preset credit target; A physical correction module is configured to collect feedback data of business execution, synchronize the feedback data to the digital twin, and correct deviations between a virtual model and physical business.

[0043] In some embodiments, the data acquisition module is specifically configured to perform data cleaning and standardization processing on the internal financial data and the external associated data, the standardization processing includes missing value filling, abnormal value denoising, unit unification and data normalization, a structured data set is formed, and multi-source data is obtained; Based on the knowledge graph technology, customer basic information, risk characteristics, industry influence information and business scenario information in the multi-source data are modeled by semantic association, a digital twin is constructed, and dynamic data interaction is realized in the digital twin through an association relationship between knowledge graph nodes and edges.

[0044] In some embodiments, the risk assessment module is specifically configured to fuse a credit assessment model and a digital twin simulation technology based on the digital twin, and generate a dynamic assessment framework; Based on a three-dimensional mapping characteristic of the digital twin, real-time variable data is input into the dynamic assessment framework, an initial risk evolution path is simulated and output through simulation calculation, and whether the dynamic assessment framework is preset with a risk logic is verified; If the dynamic assessment framework meets the risk logic, business scenario characteristics are obtained through the digital twin, real-time feedback data and the initial risk evolution path are combined, and model parameters in the dynamic assessment framework are adaptively adjusted; The adjusted dynamic assessment framework is verified through simulation, an ultimate risk evolution path is output, and a matching degree between the ultimate risk evolution path and risk characteristics in the business scenario characteristics is calculated; If the matching degree is greater than a preset threshold, a risk assessment result adapted to the business scenario is output.

[0045] In some embodiments, the risk assessment module is further configured to calculate a difference value between the default probability distribution of the adjusted path and the historical actual default probability distribution of the business scenario using the Euclidean distance; Calculate the coverage rate of the preset core node in the final risk evolution path; According to the digital twin, a number of target samples most similar to the characteristics of the business scenario are screened; According to the digital twin, the historical default data, real-time behavior data and industry data corresponding to the target samples are integrated to extract key features; According to the preset risk level quantification rule and the key features, the actual risk level is determined; According to the dynamic evaluation framework, the predicted default probability of the business scenario characteristics is simulated, and the predicted default probability is mapped to a predicted risk level, and a level difference value between the actual risk level and the predicted risk level is calculated; If the difference value, the coverage rate and the level difference value meet the preset matching condition, a risk assessment result adapted to the business scenario is output.

[0046] In some embodiments, the strategy combination module is specifically configured to construct a risk scenario library covering multi-dimensional risk characteristics according to the risk assessment result, the historical default data, the real-time behavior data and the industry data; Based on the credit business target, a plurality of groups of candidate credit strategies are defined, and the credit strategies at least include: guarantee pricing, credit limit, interest rate pricing and term structure; Based on the simulation capability of the digital twin, each group of credit strategies is input into the risk scenario library for scenario stress testing of different risk levels, and a simulation result is output; Based on the preset credit conditions, the simulation result is comprehensively evaluated, the optimal strategy combination is selected, and the optimal strategy combination is stored in association with the characteristics of the business scenario to form associated data that can be dynamically called.

[0047] In some embodiments, the dynamic adjustment module is specifically configured to define an initial parameter range of a particle swarm based on the optimal strategy combination; All initial parameter combinations are input into the risk scenario library to simulate performance under different risk scenarios, calculate each scheme risk indicator and income indicator, and generate an evaluation score; Record the best scheme that each particle has performed best in history, and record the optimal scheme that performs best among all particles; According to the individual's best scheme and the global optimal scheme, the parameters of each particle are adjusted; When the preset round of adjustment is repeated or a better scheme does not appear for a plurality of rounds, the optimization is stopped, and the current global optimal scheme is set as the dynamic adjustment scheme.

[0048] In some embodiments, the physical correction module is specifically configured to collect feedback data and perform standardization processing, and then synchronize to the database of the digital twin; According to the simulation prediction value in the digital twin and the actual observation value of the feedback data, a deviation index is calculated, and the deviation index includes: risk deviation, income deviation, and risk path deviation. According to the deviation index, the feature weight of the virtual model of the digital twin is adjusted.

[0049] Embodiments of the present application also disclose a control device.

[0050] Specifically, the control device includes a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform the above-mentioned credit strategy intelligent optimization method based on digital twin driving.

[0051] Embodiments of the present application also disclose a computer readable storage medium.

[0052] Specifically, the computer readable storage medium stores a computer program capable of being loaded and executed by the processor to perform the above-mentioned credit strategy intelligent optimization method based on digital twin driving, and the computer readable storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage program codes.

[0053] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, therefore: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A credit strategy intelligent optimization method based on digital twin drive, characterized by: include: Based on multi-source data fusion technology, internal financial data and external related data are integrated to obtain multi-source data, and a digital twin is constructed based on the multi-source data and knowledge graph technology. The multi-source data includes: basic customer information, risk characteristics, industry impact information, and business scenario information; Based on the digital twin, through the preset credit assessment model and digital twin simulation technology, real-time variable data is input to obtain risk assessment results adapted to the business scenario; Building a risk scenario library based on the risk assessment results and the digital twin, simulating the implementation effects of different credit strategies, and generating an optimal strategy combination; Based on the optimal strategy combination and the preset credit target, a particle swarm optimization algorithm is used to generate a dynamic adjustment plan including credit limit, interest rate pricing, guarantee conditions and post-loan management strategy; Collect feedback data from business execution, synchronize the feedback data to the digital twin, and correct the deviation between the virtual model and the physical business; The digital twin is based on the preset credit assessment model and digital twin simulation technology, and real-time variable data is input to obtain risk assessment results adapted to the business scenario, including: Based on the digital twin, the credit assessment model is integrated with the digital twin simulation technology to generate a dynamic assessment framework; Based on the three-dimensional mapping characteristics of the digital twin, the real-time variable data is input into the dynamic assessment framework, the initial risk evolution path is output through simulation calculation, and the risk logic preset in the dynamic assessment framework is verified; If the dynamic assessment framework conforms to the risk logic, the business scenario characteristics are obtained through the digital twin, and the model parameters in the dynamic assessment framework are adaptively adjusted in combination with the real-time feedback data and the initial risk evolution path; The adjusted dynamic assessment framework is simulated and verified, a final risk evolution path is output, and a matching degree between the final risk evolution path and the risk characteristics in the business scenario characteristics is calculated; If the matching degree is greater than a preset threshold, the risk assessment result adapted to the business scenario is output.

2. The method for intelligent optimization of credit strategies based on digital twins according to claim 1, characterized in that: The multi-source data fusion technology is used to integrate internal financial data and external related data to obtain multi-source data, and a digital twin is constructed based on the multi-source data and knowledge graph technology, including: Performing data cleaning and standardization processing on the internal financial data and the external related data, wherein the standardization processing includes missing value filling, outlier denoising, unit unification, and data normalization to form a structured data set to obtain the multi-source data; Based on knowledge graph technology, the customer basic information, the risk characteristics, industry impact information and business scenario information in the multi-source data are semantically associated with each other to construct the digital twin. In the digital twin, dynamic data interaction is achieved through the association relationship between knowledge graph nodes and edges.

3. The method for intelligent optimization of credit strategies based on digital twin drive according to claim 2, characterized in that: The calculating the matching degree between the final risk evolution path and the risk feature in the business scenario feature includes: Use Euclidean distance to calculate the difference between the default probability distribution of the adjusted path and the actual default probability distribution of the business scenario history; Calculating the coverage of the nodes in the final risk evolution path to the preset core nodes; Selecting several target samples that are closest to the business scenario characteristics from the digital twin; Integrate historical default data, real-time behavior data, and industry data corresponding to the target sample according to the digital twin to extract key features; Determine the actual risk level based on the preset risk level quantification rules and the key characteristics; Simulating the predicted default probability of the business scenario characteristics according to the dynamic assessment framework, mapping the predicted default probability into a predicted risk level, and calculating the difference between the actual risk level and the predicted risk level; If the difference value, the coverage rate, and the level difference value meet the preset matching conditions, the risk assessment result adapted to the business scenario is output.

4. The method for intelligent optimization of credit strategies based on digital twin drive according to claim 3, characterized in that: The risk scenario library is constructed based on the risk assessment results and the digital twin, the implementation effects of different credit strategies are simulated, and the optimal strategy combination is generated, including: Constructing a risk scenario library covering multi-dimensional risk characteristics based on the risk assessment results, the historical default data, the real-time behavior data, and the industry data; Based on the credit business objectives, define multiple sets of candidate credit strategies, wherein the credit strategies include at least: guarantee pricing, credit limit, interest rate pricing, and term structure; Based on the simulation capability of the digital twin, each group of credit strategies is input into the risk scenario library to perform scenario stress testing at different risk levels, and the simulation results are output; The simulation results are comprehensively evaluated based on preset credit conditions, the optimal strategy combination is screened, and the optimal strategy combination is associated with the business scenario characteristics and stored to form dynamically callable associated data.

5. The method for intelligent optimization of credit strategies based on digital twin drive according to claim 4, characterized in that: Based on the optimal strategy combination and the preset credit target, a particle swarm optimization algorithm is used to generate a dynamic adjustment plan including credit limit, interest rate pricing, guarantee conditions and post-loan management strategy, including: Based on the optimal strategy combination, defining the initial parameter range of the particle swarm; Input all initial parameter combinations into the risk scenario library, simulate the performance under different risk scenarios, calculate the risk index and benefit index of each plan, and generate an evaluation score; Record the best solution for each particle in history, and also record the best solution for all particles; Adjust the parameters of each particle according to the individual optimal solution and the global optimal solution; When the preset rounds of adjustments are repeated or no better solution appears for multiple consecutive rounds, the optimization is stopped, and the current global optimal solution is set as the dynamic adjustment solution.

6. The method for intelligent optimization of credit strategies based on digital twins according to claim 1, characterized in that: The collecting feedback data of the business execution, synchronizing the feedback data to the digital twin, and correcting the deviation between the virtual model and the physical business include: After collecting and standardizing the feedback data, the data is synchronized to the database of the digital twin; Calculating deviation indicators based on the simulated predicted values ​​in the digital twin and the actual observed values ​​of the feedback data, wherein the deviation indicators include: risk deviation, return deviation, and risk path deviation; According to the deviation index, the feature weight of the virtual model of the digital twin is adjusted.

7. A credit strategy intelligent optimization device driven by digital twins, characterized in that: The device comprises: A data acquisition module is used to integrate internal financial data and external related data based on multi-source data fusion technology to obtain multi-source data, and to build a digital twin based on the multi-source data and knowledge graph technology. The multi-source data includes: basic customer information, risk characteristics, industry impact information, and business scenario information; The risk assessment module, based on the digital twin, uses a preset credit assessment model and digital twin simulation technology to input real-time variable data to obtain risk assessment results adapted to the business scenario; A strategy combination module, configured to construct a risk scenario library based on the risk assessment results and the digital twin, simulate the implementation effects of different credit strategies, and generate an optimal strategy combination; A dynamic adjustment module, based on the optimal strategy combination and the preset credit target, uses a particle swarm optimization algorithm to generate a dynamic adjustment plan including credit limit, interest rate pricing, guarantee conditions and post-loan management strategy; A physical correction module collects feedback data from business execution, synchronizes the feedback data to the digital twin, and corrects the deviation between the virtual model and the physical business; The risk assessment module is specifically used to integrate the credit assessment model with the digital twin simulation technology based on the digital twin to generate a dynamic assessment framework; Based on the three-dimensional mapping characteristics of the digital twin, the real-time variable data is input into the dynamic assessment framework, the initial risk evolution path is output through simulation calculation, and the risk logic preset in the dynamic assessment framework is verified; If the dynamic assessment framework conforms to the risk logic, the business scenario characteristics are obtained through the digital twin, and the model parameters in the dynamic assessment framework are adaptively adjusted in combination with the real-time feedback data and the initial risk evolution path; The adjusted dynamic assessment framework is simulated and verified, a final risk evolution path is output, and a matching degree between the final risk evolution path and the risk characteristics in the business scenario characteristics is calculated; If the matching degree is greater than a preset threshold, the risk assessment result adapted to the business scenario is output.

8. A control device, characterized in that: The device comprises: The method comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 6.

Images