Multi-dimensional data integration-based financial risk intelligent risk control middle platform system and method

Abstract

The application discloses a financial risk intelligent risk control middle platform system and method based on multi-dimensional data integration, and belongs to the field of financial risk intelligent risk control. The financial risk intelligent risk control middle platform system based on multi-dimensional data integration comprises a data processing module, a model construction module, a decision construction module and a scheme formulation module. The application solves the problems of poor system integration, high maintenance cost, low availability, insufficient expansibility, low information transparency and great management difficulty in the prior art, greatly shortens the strategy iteration cycle, meets the demand of high-concurrency business scenarios, improves the accuracy of fraud identification and credit evaluation, reduces the misjudgment rate, can adapt to market changes, reduces repeated development through unified variable management, improves the data asset reuse rate, improves business autonomy, flexibly outputs the risk control capability to support personalized product design, and realizes the best balance between risk and benefit through the deep cooperation of risk control and business.

Description

A Financial Risk Intelligent Risk Control Platform System and Method Based on Multi-Dimensional Data Integration Technical Field

[0001] This invention relates to the field of intelligent risk control technology for financial risks, specifically to an intelligent risk control platform system and method for financial risks based on multi-dimensional data integration. Background Technology

[0002] Traditional risk control decision-making systems rely primarily on static rules and limited data dimensions, making it difficult to cope with rapid changes in the market environment, the diversification of risk types, and the dynamic evolution of individual loan user needs. Especially in the digital economy era, the speed of risk transmission has accelerated, fraud methods are constantly evolving, and individual loan users are increasingly demanding higher service efficiency and accuracy. This has made the limitations of the original risk control system increasingly apparent, including problems such as delayed response, insufficient coverage, and high misjudgment rates. These issues not only affect the company's risk control capabilities but may also hinder business innovation and the improvement of the individual loan user experience. Traditional risk control decision-making systems have the following drawbacks:

[0003] 1. Poor system integration and high maintenance costs: Traditional risk control decision systems typically employ a modular architecture, with core components (such as decision engines, model management, and variable calculation) operating in isolation. Business personnel must rely on multiple independent systems working together to configure risk control strategies. For example, variables must be processed in the variable system before being integrated into the decision engine by the development team. This fragmented architecture not only increases development and management costs but also leads to long strategy iteration cycles, making it unable to adapt to rapidly changing market demands.

[0004] 2. Low availability and insufficient scalability: Traditional risk control systems mostly adopt a process-oriented architecture, with complex and difficult-to-maintain decision-making processes and poor system scalability. In high-concurrency or complex strategy scenarios, the computing power and data request response capabilities of the decision engine are limited, resulting in performance bottlenecks and affecting the real-time performance of risk control decisions. In addition, due to the lack of flexible component design, the system is unable to support changing business needs, such as personalized risk control strategies and dynamic pricing, which restricts business innovation.

[0005] 3. Low information transparency and high management difficulty: The variable processing logic of traditional risk control systems is scattered across different modules, lacking unified management. This makes variable assets difficult to reuse and optimize. Furthermore, a large amount of key information (such as parameter assembly, variable acquisition, and intermediate result transmission) remains a "black box," making it difficult for business personnel to intuitively understand the strategy logic, increasing the difficulty of strategy optimization and management. In addition, due to the lack of visualization tools, the adjustment and optimization of risk control strategies heavily rely on the technical team, resulting in low business autonomy.

[0006] Therefore, it does not meet the existing needs. In response, we propose a financial risk intelligent risk control platform system and method based on multi-dimensional data integration. Summary of the Invention

[0007] The purpose of this invention is to provide a financial risk intelligent risk control platform system and method based on multi-dimensional data integration. Through atomic risk control components and visualized strategy orchestration, business personnel can independently maintain strategies, significantly shortening the strategy iteration cycle. Real-time computing capabilities are greatly enhanced, meeting the needs of high-concurrency business scenarios. Multi-dimensional data fusion and intelligent algorithm optimization improve the accuracy of fraud identification and credit assessment, reduce false judgment rates, and increase data asset reuse rates. Policy transparency reduces management complexity and enhances business autonomy. Flexible risk control capabilities support personalized product design, achieving the optimal balance between risk and return, thus solving the problems mentioned in the background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a financial risk intelligent risk control platform system based on multi-dimensional data integration, comprising:

[0009] The data processing module is used to acquire and integrate the credit data, transaction data, and behavioral data of individual credit users to obtain integrated individual credit user characteristic data variables.

[0010] The model building module is used to build scoring models, pricing models, and fraud detection models based on the fused personal credit user characteristic data variables, and to train, validate, and optimize them.

[0011] The decision-making module is used to build dedicated decision flows for different risk control scenarios for personal credit users using the business modules, and to manage them using the management module.

[0012] The scheme formulation module is used to select risk control decision flows that match the characteristics of the product and individual credit users from the resource pool;

[0013] The financial risk control decision-making module is used to make corresponding financial risk control decisions based on a risk control decision flow that conforms to the characteristics of products and individual credit users. Based on the implementation effect of the financial risk control decision results, the module dynamically feeds back and optimizes the risk control decision flow. Specifically, this includes: scoring the implementation effect of the financial risk control decision results, detecting decision gaps, formulating optimization and compensation strategies, and dynamically optimizing the decision flow. Based on the optimized and compensated risk control decision flow, the module then makes corresponding financial risk control decisions again.

[0014] Preferably, the data processing module includes:

[0015] The data acquisition unit is used to acquire credit data, transaction data, and behavioral data of individual credit users.

[0016] The data integration unit is used to integrate the acquired personal credit user credit data, transaction data, and behavioral data to build a unified data asset center;

[0017] The variable expansion unit is used to expand the risk control feature dimensions of personal credit users by utilizing feature engineering and variable derivation.

[0018] Preferably, the variable expansion unit specifically includes:

[0019] Use the IQR rule to correct outliers in personal credit user credit data, transaction data, and behavioral data;

[0020] Personal credit user credit data, transaction data, and behavioral data are classified into categorical variables, date variables, and continuous variables according to data type, and date variables are parsed into time interval features.

[0021] Label Encoding is used to convert categorical variables into numerical data and then standardize the numerical data.

[0022] By using statistical methods to derive new features from numerical data based on time interval characteristics, we can obtain the fused personal credit user feature data variables.

[0023] Preferably, the model building module includes:

[0024] The building unit is used to construct a personal credit user scoring model, a personal credit user pricing model, and a personal credit user fraud detection model using personal credit user characteristic data variables, respectively.

[0025] The training and optimization unit is used to train, validate, and optimize the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model.

[0026] Preferably, the model building module specifically includes:

[0027] Select the feature data variables that are highly correlated with the target variable for personal credit user characteristics, and remove irrelevant and redundant feature data variables;

[0028] The feature data variables are divided into different model building data, specifically including: data related to scoring model building, data related to pricing model building, and data related to fraud detection model building;

[0029] The data for building different models are divided into training sets and test sets for the relevant models. 70% of the samples are randomly selected as the training set and the remaining 30% as the test set.

[0030] Using convolutional neural networks as the basic framework, a personal credit user scoring model is constructed by leveraging relevant data and a scoring model.

[0031] Using recurrent neural networks as the basic framework, a pricing model for personal credit users is constructed using relevant data and a pricing model.

[0032] Using recurrent neural networks as the basic framework, a fraud detection model for personal credit users is constructed using relevant data. Recurrent neural networks can capture patterns in sequential data and are suitable for fraud detection.

[0033] The personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model were trained using the corresponding training sets.

[0034] After training, the performance of the personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model were verified using the corresponding test sets.

[0035] Based on the verification results, the performance of the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model were optimized respectively.

[0036] The personal credit user characteristic data variables are input into the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model, respectively, and the personal credit user scoring, personal credit user pricing, and personal credit user fraud detection results are output.

[0037] Preferably, the decision-making construction module includes:

[0038] The business segment includes a process engine, a decision engine, and a data base. Based on different risk control scenarios, the sample time window is determined, and the decision engine is used to process Boolean logic judgments, perform real-time feature calculations, and perform dynamic formula parsing to obtain a rule package.

[0039] Use rule packages to store variables in a structured way, and configure loop rules to process collection objects;

[0040] It accesses personal loan user credit data, transaction data, and behavioral data, and integrates scoring models, pricing models, and fraud detection models to build dedicated decision flows for different risk control scenarios for personal loan users.

[0041] Preferably, the scheme formulation module includes:

[0042] The decision-making unit is used to determine the risk control decision flow for individual credit users based on their credit scores, pricing, and fraud identification results.

[0043] The decision selection unit is used to select risk control decision flows from the resource pool that match the characteristics of the product and the individual credit user based on the risk control decision flow of the individual credit user.

[0044] Preferably, the implementation effect based on the financial risk control decision results is used to dynamically optimize the risk control decision flow through feedback, specifically by performing the following operations:

[0045] The effectiveness of financial risk control decisions can be preliminarily scored using the following formula:

[0046]

[0047] in, The scoring results represent the effectiveness of the implementation of financial risk control decisions. To assess the effectiveness of financial risk control decision-making outcomes in the implementation phase Scores under each scoring indicator For the first The first weight corresponding to each scoring indicator The total number of scoring indicators;

[0048] When the score of the implementation effect of financial risk control decision results is lower than the threshold, based on the pre-trained decision gap detection model, multiple decision gaps in the risk control decision flow are detected according to the implementation effect of financial risk control decision results.

[0049] The priority value for optimizing the filling of each decision gap is calculated using the following formula:

[0050]

[0051] in, For the first Priority values ​​for optimizing and filling decision gaps , The total number of decision gaps. For the first The severity of the decision-making gap The second weight corresponds to the severity. For the first The degree of significance of optimizing and filling the decision-making gap To optimize the third weight corresponding to the degree of compensation significance, For the first The reproducibility of each decision-making gap The fourth weight corresponds to the degree of reproducibility;

[0052] Iterate through each decision gap in descending order of optimization and compensation priority value;

[0053] During each iteration, the optimal optimization strategy is matched to the decision gap encountered.

[0054] Based on the optimal optimization and compensation strategy, the decision gaps encountered in the risk control decision flow are optimized and compensated. Based on the optimized and compensated risk control decision flow, the corresponding financial risk control decisions are re-made.

[0055] Preferably, the step of matching the optimal optimization and compensation strategy to the traversed decision gaps includes:

[0056] Analyze the optimization and compensation needs of the identified decision gaps to determine the set of optimization and compensation needs;

[0057] The optimization and compensation requirement set is characterized to obtain a feature description vector;

[0058] Obtain multiple sets of one-to-one corresponding standard feature description vectors and candidate optimization and compensation strategies;

[0059] The feature description vector is matched with any standard feature description vector to obtain the matching degree;

[0060] The candidate optimization strategy corresponding to the standard feature description vector with the highest matching degree with the feature description vector is taken as the optimal optimization strategy, and it is matched with the decision gaps encountered.

[0061] The financial risk intelligent risk control platform method based on multi-dimensional data integration, applied in a financial risk intelligent risk control platform system based on multi-dimensional data integration, includes the following steps:

[0062] S1: Acquire and integrate the credit data, transaction data, and behavioral data of individual credit users to obtain the integrated individual credit user characteristic data variables;

[0063] S2: New features are derived from numerical data using statistical methods based on time interval characteristics to obtain fused personal credit user characteristic data variables;

[0064] S3: Construct a scoring model, a pricing model, and a fraud detection model based on the fused personal credit user characteristic data variables, and then train, validate, and optimize them;

[0065] S4: Using the personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model, output the personal credit user scoring, personal credit user pricing, and personal credit user fraud detection results.

[0066] S5: Utilize business segments to build dedicated decision flows for different risk control scenarios for individual loan users;

[0067] S6: Determine the risk control decision flow for individual credit users based on their personal credit user scores, pricing, and fraud identification results.

[0068] S7: Select risk control decision flows from the resource pool that match the characteristics of the product and the individual credit user based on the risk control decision flow of the individual credit user. If there is no risk control decision flow that matches the characteristics of the individual credit user in the resource pool, then formulate a risk control decision flow product that matches the characteristics of the individual credit user based on the risk control decision flow that matches the characteristics of the individual credit user.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] 1. This invention enables business personnel to maintain strategies independently through atomic risk control components and visualized strategy orchestration, significantly shortening the strategy iteration cycle;

[0071] 2. This invention significantly enhances real-time computing capabilities to meet the needs of high-concurrency business scenarios. Multi-dimensional data fusion and intelligent algorithm optimization improve the accuracy of fraud identification and credit assessment, reduce the false judgment rate, and the rapid iteration and update mechanism of the model ensures the continuous evolution of risk control capabilities to adapt to market changes. Unified variable management reduces redundant development and improves the reuse rate of data assets.

[0072] 3. This invention reduces management complexity and increases business autonomy by making strategies transparent. Flexible risk control capabilities support personalized product design and help business growth. Through deep collaboration between risk control and business, it achieves the best balance between risk and return.

[0073] 4. Continuously improve decision-making quality during financial risk control through dynamic feedback and optimization mechanisms. By scoring implementation effectiveness, detecting decision gaps, formulating optimization and compensation strategies, and dynamically optimizing the decision flow, the accuracy and effectiveness of financial risk control decisions can be improved, thereby reducing risk and enhancing the security and sustainability of credit products.

[0074] 5. By analyzing the needs of decision gaps, matching feature vectors, and selecting the optimal optimization strategy, the accuracy of the financial risk control decision-making system is continuously improved. Through real-time dynamic feedback optimization, the system can maintain efficient and accurate risk control decision-making capabilities under the influence of external factors such as changes in the market environment and user behavior. Attached Figure Description

[0075] Figure 1 is a schematic diagram of the financial risk intelligent risk control platform system module based on multi-dimensional data integration of the present invention;

[0076] Figure 2 is a schematic diagram of the intelligent risk control platform method for financial risk based on multi-dimensional data integration of the present invention;

[0077] Figure 3 is a schematic diagram of the decision-making construction module of the financial risk intelligent risk control platform system based on multi-dimensional data integration of the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] To address the problems of poor system integration, high maintenance costs, low availability, insufficient scalability, low information transparency, and high management difficulty in existing technologies, please refer to Figures 1, 2, and 3. This embodiment provides the following technical solution:

[0080] A financial risk intelligent risk control platform system based on multi-dimensional data integration includes:

[0081] The data processing module is used to acquire and integrate credit data, transaction data, and behavioral data of individual credit users to obtain integrated individual credit user characteristic data variables. By integrating credit data, transaction data, and behavioral data, a unified data asset center is constructed. Through feature engineering and variable derivation, the risk control feature dimensions are expanded, the marginal effect of data is improved, and a variable center serving the entire risk control process is formed to support the decision-making of the entire risk control process. A combination of real-time calculation and batch processing is used to ensure the timeliness and accuracy of the data.

[0082] The model building module is used to build scoring models, pricing models, and fraud detection models based on the fused personal credit user characteristic data variables, and to train, validate, and optimize them. The model is continuously adjusted and optimized according to the actual running effect, so as to support continuous model updates, ensure iterative optimization of the model, and adapt to market changes.

[0083] The decision building module is used to build dedicated decision flows for different risk control scenarios for personal credit users using business modules. It supports the visual orchestration of strategies, and business personnel can quickly adjust the risk control logic by dragging and dropping, reducing technical dependence and providing strategy simulation testing capabilities to ensure the stability and effectiveness of new strategies before they go live.

[0084] The solution development module is used to select risk control decision flows that match the characteristics of products and personal credit users from the resource pool. It supports the customization of personalized risk control solutions, generates market-oriented and flexible risk control products, realizes flexible output of risk control capabilities, supports continuous monitoring and dynamic adjustment, and ensures that risk control strategies evolve in sync with business needs.

[0085] The data processing module includes:

[0086] The data acquisition unit is used to acquire credit data, transaction data, and behavioral data of individual credit users.

[0087] The data integration unit is used to integrate the acquired personal credit user credit data, transaction data, and behavioral data to build a unified data asset center;

[0088] The variable expansion unit is used to expand the risk control feature dimensions of personal credit users by utilizing feature engineering and variable derivation.

[0089] Variable expansion units, specifically including:

[0090] Use the IQR rule to correct outliers in personal credit user credit data, transaction data, and behavioral data;

[0091] Personal credit user credit data, transaction data, and behavioral data are classified into categorical variables, date variables, and continuous variables according to data type, and date variables are parsed into time interval features.

[0092] Label Encoding is used to convert categorical variables into numerical data and then standardize the numerical data.

[0093] By using statistical methods to derive new features from numerical data based on time interval characteristics, these new features become the integrated personal credit user characteristic data variables. These new features include: basic features, time-series features, cross-combinations, and complex features. For multiple categorical features, feature cross-combinations can be performed through multiplication, division, and other methods. Basic feature derivation includes statistical features, such as the average number of transactions in the past 3 months and the maximum number of historical overdue payments; ratio features, such as the debt-to-income ratio and credit card overdraft rate; time-series feature derivation includes sliding statistics based on time windows, such as the standard deviation of consumption amount in the past 6 months and the trend of loan application frequency changes in the past year; cross-combination derivation includes multi-field interaction and multinomial features; and complex feature mining includes graph features and text features.

[0094] The model building module includes:

[0095] The building unit is used to construct a personal credit user scoring model, a personal credit user pricing model, and a personal credit user fraud detection model using personal credit user characteristic data variables, respectively.

[0096] The training and optimization unit is used to train, validate, and optimize the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model.

[0097] The model building module specifically includes:

[0098] Select the feature data variables that are highly correlated with the target variable for personal credit user characteristics, and remove irrelevant and redundant feature data variables;

[0099] The feature data variables are divided into different model building data, specifically including: data related to scoring model building, data related to pricing model building, and data related to fraud detection model building;

[0100] The data for building different models are divided into training and test sets for the relevant models. 70% of the samples are randomly selected as the training set, and the model is developed on the training set based on the maximum separation algorithm. The remaining 30% is the test set, and the model effect is verified on the test set.

[0101] Using convolutional neural networks as the basic framework, a personal credit user scoring model is constructed by leveraging relevant data from a scoring model. Convolutional neural networks can handle complex nonlinear relationships and are suitable for constructing scoring models.

[0102] Using recurrent neural networks as the basic framework, a pricing model for personal credit users is constructed using relevant data from a pricing model. Recurrent neural networks can process time series data and are suitable for constructing pricing models.

[0103] Using recurrent neural networks as the basic framework, a fraud detection model for personal credit users is constructed using relevant data. Recurrent neural networks can capture patterns in sequential data and are suitable for fraud detection.

[0104] The personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model were trained using the corresponding training sets to ensure that the models could learn the complex relationships in the data.

[0105] After training, the performance of the personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model were verified using the corresponding test sets.

[0106] Based on the verification results, the performance of the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model were optimized respectively.

[0107] The personal credit user characteristic data variables are input into the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model, respectively, and the personal credit user scoring, personal credit user pricing, and personal credit user fraud detection results are output.

[0108] The decision-making construction module includes:

[0109] The business segment includes a process engine, a decision engine, and a data base. Based on different risk control scenarios, the sample time window is determined, and the decision engine is used to process Boolean logic judgments, perform real-time feature calculations, and perform dynamic formula parsing to obtain a rule package.

[0110] Use rule packages to store variables in a structured way, and configure loop rules to process collection objects;

[0111] It accesses personal loan user credit data, transaction data, and behavioral data, and integrates scoring models, pricing models, and fraud detection models to build dedicated decision flows for different risk control scenarios for personal loan users.

[0112] The business segments include a process engine, a decision engine, and a data foundation. The process engine includes decision flow orchestration, decision flow scheduling, input parameter standardization, blacklists and graylists, manual review, and traffic distribution. The decision engine comprises an application layer, a component layer, and a variable layer. The application layer includes strategy orchestration, component scheduling, traffic distribution, version management, simulation verification, rule package configuration, and expression configuration. Rule package configuration includes rule sets and priority allocation, while expression configuration includes Java-like processing and custom functions. The component layer includes routing nodes, rule nodes, assignment nodes, termination nodes, and the rule engine. The rule engine includes a decision table. The system comprises decision trees, scorecards, and mathematical formulas. The variable layer includes upstream system supply variables, variable center supply variables, and system-specific variables. Upstream system supply variables include standardized parameters of product elements and channel characteristics. Variable center supply variables include model results, credit variables, external variables, and derived variables. System-specific variables include custom variables and cached computational variables. The data foundation includes the variable center and the FlowCube engine. The management module includes release management, access control, and alert management. The release management process sequentially includes: draft, testing, batch verification, canary release, deployment, and champion challenge. Access control includes roles. The system also includes decision flows and rule packages. Alert management includes alerts for cross-component infinite loops, pass rates, custom error codes, cross-quantity, response time, and volatility.

[0113] The solution development module includes:

[0114] The decision-making unit is used to determine the risk control decision flow for individual credit users based on their credit scores, pricing, and fraud identification results.

[0115] The decision selection unit is used to select risk control decision flows from the resource pool that match the characteristics of the product and the individual credit users based on the risk control decision flows of individual credit users. It supports the customization of personalized risk control solutions and generates market-oriented, flexible and adaptable risk control products.

[0116] The system also includes:

[0117] The financial risk control decision-making module is used to make corresponding financial risk control decisions based on the risk control decision flow that conforms to the characteristics of products and personal credit users, and to dynamically optimize the risk control decision flow based on the implementation effect of the financial risk control decision results.

[0118] Specifically, the implementation effect based on the financial risk control decision-making results is used to dynamically optimize the risk control decision flow through feedback, and the following operations are performed:

[0119] The effectiveness of financial risk control decisions can be preliminarily scored using the following formula:

[0120]

[0121] in, The scoring results represent the effectiveness of the implementation of financial risk control decisions. To assess the effectiveness of financial risk control decision-making outcomes in the implementation phase Scores under each scoring indicator For the first The first weight corresponding to each scoring indicator The total number of scoring indicators;

[0122] When the score of the implementation effect of financial risk control decision results is lower than the threshold, based on the pre-trained decision gap detection model, multiple decision gaps in the risk control decision flow are detected according to the implementation effect of financial risk control decision results.

[0123] The priority value for optimizing the filling of each decision gap is calculated using the following formula:

[0124]

[0125] in, For the first Priority values ​​for optimizing and filling decision gaps , The total number of decision gaps. For the first The severity of the decision-making gap The second weight corresponds to the severity. For the first The degree of significance of optimizing and filling the decision-making gap To optimize the third weight corresponding to the degree of compensation significance, For the first The reproducibility of each decision-making gap The fourth weight corresponds to the degree of reproducibility;

[0126] Iterate through each decision gap in descending order of optimization and compensation priority value;

[0127] During each iteration, the optimal optimization strategy is matched to the decision gap encountered.

[0128] Based on the optimal optimization and compensation strategy, the decision gaps encountered in the risk control decision flow are optimized and compensated. Based on the optimized and compensated risk control decision flow, the corresponding financial risk control decisions are re-made.

[0129] The working principle and beneficial effects of the above technical solution are as follows:

[0130] The financial risk control decision-making module is used to make risk control decisions for credit users based on specific rules and standards. Risk control decision-making refers to assessing the credit risk of loan applicants and deciding whether to approve, reject, or adjust credit limits when granting loans or credit lines. Scoring indicators are standards used to evaluate the effectiveness of risk control decision implementation, including default rate, repayment rate, and delinquency rate. The first weight of each scoring indicator indicates its importance in the overall effectiveness evaluation. For example, if the default rate has a significant impact on the decision-making effect, then the default rate may have a higher weight. A comprehensive score is calculated by combining the scores and corresponding weights of the implementation effectiveness of financial risk control decisions under multiple scoring indicators; this score represents the actual effectiveness of the risk control decisions.

[0131] When the score is below a threshold, it indicates that the risk control decision-making effect is poor and needs optimization. A decision gap refers to the difference between the actual and expected results after implementing risk control decisions. For example, the system might find that the default rate of certain credit user groups is higher than expected; this situation can be considered a decision gap. The decision gap detection model is pre-trained using machine learning to obtain data from a large number of historical decision gaps encountered in making corresponding financial risk control decisions based on risk control decision flows.

[0132] The severity of a decision gap refers to its degree of importance; for example, if a decision gap may lead to a significant risk of default, its severity is high. The significance of mitigation indicates the value and impact of resolving the gap; for example, resolving the gap may significantly reduce risk or improve decision-making efficiency, thus its significance is high. Reproducibility refers to the degree to which the decision gap has historically recurred in different scenarios; a high reproducibility rate means it needs to be addressed first. A weighted calculation is performed on the severity, significance of mitigation, and reproducibility to obtain the priority value for mitigation. The second, third, and fourth weights are pre-set based on their respective impact on the priority value for mitigation.

[0133] The higher the priority value of optimization and compensation, the more the corresponding decision gap needs to be optimized and compensated. Therefore, it will be traversed first. When it is traversed, the optimal optimization and compensation strategy is matched for the traversed decision gap. Based on the optimal optimization and compensation strategy, the risk control decision flow is optimized and compensated for the traversed decision gap. Based on the risk control decision flow after optimization and compensation of the traversed decision gap, the corresponding financial risk control decision is re-made.

[0134] The aforementioned technical solution continuously improves decision-making quality during the financial risk control process through dynamic feedback and optimization mechanisms. By scoring implementation effectiveness, detecting decision gaps, formulating optimization and compensation strategies, and dynamically optimizing the decision flow, the accuracy and effectiveness of financial risk control decisions can be improved, thereby reducing risk and enhancing the security and sustainability of credit products.

[0135] The process of matching the optimal optimization strategy to the traversed decision gaps includes:

[0136] Analyze the optimization and compensation needs of the identified decision gaps to determine the set of optimization and compensation needs;

[0137] The optimization and compensation requirement set is characterized to obtain a feature description vector;

[0138] Obtain multiple sets of one-to-one corresponding standard feature description vectors and candidate optimization and compensation strategies;

[0139] The feature description vector is matched with any standard feature description vector to obtain the matching degree;

[0140] The candidate optimization strategy corresponding to the standard feature description vector with the highest matching degree with the feature description vector is taken as the optimal optimization strategy, and it is matched with the decision gaps encountered.

[0141] The working principle and beneficial effects of the above technical solution are as follows:

[0142] When traversing each decision gap, the first step is to conduct an optimization and remedy analysis. This step assesses the gap, understands its nature, and how to mitigate the risks it poses. This analysis helps the system understand which risk control rules, algorithms, or measures need adjustment or improvement. For example, a decision gap might be due to an inadequate risk assessment model for a specific user group; optimization and remedy would involve enhancing the risk identification capability for that group.

[0143] After conducting the optimization requirements analysis, the system generates an optimization compensation requirement set. This set includes all requirements that need adjustment, such as modifying risk control model parameters, adding monitoring measures, or adjusting decision-making processes. These requirement sets will be used to guide subsequent optimization decisions and ensure the relevance and effectiveness of the optimization solutions.

[0144] Feature description is a characteristic description of the set of optimization and compensation needs, including the optimization objective, requirements, and influencing factors. For example, in a decision gap, features such as customer credit scores and historical loan application information may need to be considered. Feature description vectors convert these descriptions into digital vector form for easy matching. Standard feature description vectors are pre-constructed and correspond to standard descriptions of past optimization and compensation needs. The feature description vector is matched with any standard feature description vector to obtain the matching degree. The candidate optimization and compensation strategy corresponding to the standard feature description vector with the highest matching degree is taken as the optimal optimization and compensation strategy. This strategy is most suitable for optimizing and compensating for the decision gap encountered in the current optimization and compensation needs set, and is then matched with the encountered decision gap. Optimization and compensation strategies may include adjusting risk control rules, increasing monitoring of certain groups, and designing new risk control models or algorithms to better predict risks.

[0145] The aforementioned technical solution continuously improves the accuracy of the financial risk control decision-making system by analyzing the needs of decision gaps, matching feature vectors, and selecting the optimal optimization strategy. Through real-time dynamic feedback optimization, the system can maintain efficient and accurate risk control decision-making capabilities despite external factors such as changes in the market environment and user behavior.

[0146] The financial risk intelligent risk control platform method based on multi-dimensional data integration, applied in a financial risk intelligent risk control platform system based on multi-dimensional data integration, includes the following steps:

[0147] S1: Acquire and integrate credit data, transaction data, and behavioral data of individual credit users, and build a unified data asset center;

[0148] S2: New features are derived from numerical data using statistical methods based on time interval characteristics to obtain fused personal credit user characteristic data variables;

[0149] S3: Construct a scoring model, a pricing model, and a fraud detection model based on the fused personal credit user characteristic data variables, and then train, validate, and optimize them;

[0150] S4: Using the personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model, output the personal credit user scoring, personal credit user pricing, and personal credit user fraud detection results.

[0151] S5: Utilize business segments to build dedicated decision flows for different risk control scenarios for individual loan users;

[0152] S6: Determine the risk control decision flow for individual credit users based on their personal credit user scores, pricing, and fraud identification results.

[0153] S7: Select risk control decision flows from the resource pool that match the characteristics of the product and the individual credit user based on the risk control decision flow of the individual credit user. If there is no risk control decision flow that matches the characteristics of the individual credit user in the resource pool, then formulate a risk control decision flow product that matches the characteristics of the individual credit user based on the risk control decision flow that matches the characteristics of the individual credit user.

[0154] In summary, this invention, through atomic risk control components and visualized strategy orchestration, allows business personnel to independently maintain strategies, significantly shortening the strategy iteration cycle and greatly improving real-time computing capabilities. This, in turn, meets the needs of high-concurrency business scenarios. Through multi-dimensional data fusion and intelligent algorithm optimization, it improves the accuracy of fraud detection and credit assessment, reducing false positive rates. By continuously adjusting and optimizing the model based on actual operational results and implementing a rapid iteration and update mechanism, it ensures the continuous evolution of risk control capabilities to adapt to market changes. Unified variable management reduces redundant development, improving data asset reuse rates. Furthermore, strategy transparency reduces management complexity, enhancing business self-management capabilities. The system prioritizes risk control decision flows. Based on the risk control decision flows of individual credit users, it selects risk control decision flows from the resource pool that match the characteristics of the product and individual credit users. If no risk control decision flows matching the characteristics of individual credit users are found in the resource pool, a risk control decision flow product matching the characteristics of individual credit users is developed based on the risk control decision flows that match the characteristics of individual credit users. This enables flexible output of risk control capabilities, supports continuous monitoring and dynamic adjustment, and ensures that risk control strategies evolve in sync with business needs. Through flexible output of risk control capabilities, it supports personalized product design (such as differentiated pricing, dynamic credit limit adjustments, etc.), helps business growth, and achieves the best balance between risk and return through deep synergy between risk control and business.

[0155] 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 process, method, article, or apparatus.

[0156] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A financial risk intelligent risk control platform system based on multi-dimensional data integration, characterized in that: include: The data processing module is used to acquire and integrate the credit data, transaction data, and behavioral data of individual credit users to obtain integrated individual credit user characteristic data variables. The model building module is used to construct scoring models, pricing models, and fraud detection models based on the fused personal credit user characteristic data variables, and to train, validate, and optimize them. The decision building module is used to construct dedicated decision flows for different risk control scenarios for personal credit users using the business module, and to manage them using the management module. The solution formulation module is used to select risk control decision flows that conform to the characteristics of the product and personal credit users from the resource pool. The financial risk control decision module is used to make corresponding financial risk control decisions based on the risk control decision flows that conform to the characteristics of the product and personal credit users, and to dynamically optimize the risk control decision flows based on the implementation effect of the financial risk control decision results. Specifically, this includes: scoring the implementation effect of the financial risk control decision results, detecting decision gaps, formulating optimization and compensation strategies, and dynamically optimizing the decision flow, and re-making corresponding financial risk control decisions based on the optimized and compensated risk control decision flow. The dynamic feedback optimization of the risk control decision flow based on the implementation effect of the financial risk control decision results specifically performs the following operations: The implementation effect of the financial risk control decision results is initially scored using the following formula: ;in, The scoring results represent the effectiveness of the implementation of financial risk control decisions. To assess the effectiveness of financial risk control decision-making outcomes in the implementation phase Scores under each scoring indicator For the first The first weight corresponding to each scoring indicator The total number of scoring indicators; when the score of the implementation effect of financial risk control decision results is lower than the threshold, based on the pre-trained decision gap detection model, multiple decision gaps in the risk control decision flow are detected according to the implementation effect of financial risk control decision results; the optimization compensation priority value of each decision gap is calculated by the following formula: ;in, For the first Priority values ​​for optimizing and filling decision gaps ， The total number of decision gaps. For the first The severity of the decision-making gap The second weight corresponds to the severity. For the first The degree of significance of optimizing and filling the decision-making gap To optimize the third weight corresponding to the degree of compensation significance, For the first The reproducibility of each decision-making gap The fourth weight corresponds to the reproducibility level; each decision gap is traversed sequentially according to the priority value of optimization and compensation from largest to smallest; during each traversal, the optimal optimization and compensation strategy is matched for the traversed decision gap; based on the optimal optimization and compensation strategy, the risk control decision flow is optimized and compensated for the traversed decision gaps, and based on the optimized and compensated risk control decision flow, the corresponding financial risk control decision is re-made; matching the optimal optimization and compensation strategy for the traversed decision gaps includes: performing optimization and compensation demand analysis on the traversed decision gaps to determine the optimization and compensation demand set; performing feature description on the optimization and compensation demand set to obtain feature description vectors; obtaining multiple sets of one-to-one corresponding standard feature description vectors and candidate optimization and compensation strategies; matching the feature description vector with any standard feature description vector to obtain the matching degree; taking the candidate optimization and compensation strategy corresponding to the standard feature description vector with the largest matching degree with the feature description vector as the optimal optimization and compensation strategy, and matching it with the traversed decision gap.

2. The financial risk intelligent risk control platform system based on multi-dimensional data integration as described in claim 1, characterized in that, The data processing module includes: a data acquisition unit for acquiring credit data, transaction data, and behavioral data of individual credit users; a data integration unit for integrating the acquired credit data, transaction data, and behavioral data of individual credit users to build a unified data asset center; and a variable expansion unit for expanding the risk control feature dimensions of individual credit users by utilizing feature engineering and variable derivation.

3. The financial risk intelligent risk control platform system based on multi-dimensional data integration as described in claim 2, characterized in that, The variable expansion unit specifically includes: using the IQR rule to correct outliers in personal credit user credit data, transaction data, and behavioral data; classifying personal credit user credit data, transaction data, and behavioral data into categorical variables, date variables, and continuous variables according to data type, and parsing date variables into time interval features; using Label Encoding to convert categorical variables into numerical data and standardizing the numerical data; and deriving new features from the numerical data according to the time interval features using statistical methods, which are the fused personal credit user feature data variables.

4. The intelligent financial risk control platform system based on multi-dimensional data integration as described in claim 1, characterized in that, The model building module includes: a building unit, used to build a personal credit user scoring model, a personal credit user pricing model, and a personal credit user fraud detection model using personal credit user characteristic data variables; and a training and optimization unit, used to train, verify, and optimize the personal credit user scoring model, the personal credit user pricing model, and the personal credit user fraud detection model.

5. The intelligent financial risk control platform system based on multi-dimensional data integration as described in claim 4, characterized in that, The model building module specifically includes: selecting feature data variables that are highly correlated with the target variable for personal credit user characteristic data variables, and removing irrelevant and redundant feature data variables; dividing the feature data variables into different model building data, specifically including: data related to scoring model building, data related to pricing model building, and data related to fraud detection model building; dividing the different model building data into training sets and test sets for relevant models, randomly selecting 70% of the samples as the training set and the remaining 30% as the test set; using a convolutional neural network as the basic framework, constructing a personal credit user scoring model using the data related to scoring model building; using a recurrent neural network as the basic framework, constructing a personal credit user pricing model using the data related to pricing model building; and using a recurrent neural network as the basic framework, constructing a personal credit user pricing model using the data related to fraud detection model building. A personal loan user fraud detection model is proposed. Recurrent neural networks can capture patterns in sequential data, making them suitable for fraud detection. The model is trained using corresponding training sets for a personal loan user scoring model, a personal loan user pricing model, and a personal loan user fraud detection model. After training, the performance of each model is validated using corresponding test sets. Based on the validation results, the performance of each model is optimized. Personal loan user feature data variables are input into the personal loan user scoring model, personal loan user pricing model, and personal loan user fraud detection model, respectively, and the outputs are the personal loan user score, personal loan user pricing, and personal loan user fraud detection results.

6. The intelligent financial risk control platform system based on multi-dimensional data integration as described in claim 1, characterized in that, The decision-making construction module includes: a business segment comprising a process engine, a decision engine, and a data base; determining sample time windows based on different risk control scenarios; utilizing the decision engine to process Boolean logic judgments, perform real-time feature calculations, and dynamically parse formulas to obtain a rule package; using the rule package to structurally store variables and configuring cyclic rule processing set objects; accessing personal credit user credit data, transaction data, and behavioral data, and integrating scoring models, pricing models, and fraud detection models to construct dedicated decision flows for different risk control scenarios for personal credit users.

7. The intelligent financial risk control platform system based on multi-dimensional data integration as described in claim 1, characterized in that, The scheme formulation module includes: a decision determination unit, used to determine the risk control decision flow for individual credit users based on individual credit user scores, individual credit user pricing, and individual credit user fraud identification results; and a decision selection unit, used to select risk control decision flows from the resource pool that match the characteristics of the product and the individual credit user based on the individual credit user's risk control decision flow.

8. A financial risk intelligent risk control platform method based on multi-dimensional data integration, applied in the financial risk intelligent risk control platform system based on multi-dimensional data integration as described in claim 7, characterized in that, Includes the following steps: S1: Acquire and integrate the credit data, transaction data, and behavioral data of individual credit users to obtain the integrated individual credit user characteristic data variables; S2: New features are derived from numerical data using statistical methods based on time interval characteristics to obtain fused personal credit user characteristic data variables; S3: Construct a scoring model, pricing model, and fraud detection model based on the fused personal credit user characteristic data variables, and train, validate, and optimize them; S4: Utilize the personal credit user scoring model, personal credit user pricing model, and personal credit user fraud detection model to output personal credit user scores, personal credit user pricing, and personal credit user fraud detection results; S5: Utilize business segments to construct dedicated decision flows for different risk control scenarios for personal credit users; S6: Determine the risk control decision flow for personal credit users based on personal credit user scores, personal credit user pricing, and personal credit user fraud detection results; S7: Based on the personal credit user's risk control decision flow, select risk control decision flows from the resource pool that match the product and personal credit user characteristics. If no risk control decision flow matches the personal credit user's characteristics in the resource pool, then develop a risk control decision flow product that matches the personal credit user's characteristics based on the risk control decision flow that matches the personal credit user's characteristics.

Images