Consumption finance intelligent risk control capability building method and system

Abstract

The invention relates to a method and a system for constructing intelligent risk control capability of consumer finance. The method comprises the following steps: step 1, constructing a full-process digital risk control system; 2, constructing a risk control model variable data link panorama; 3, constructing a big data intelligent risk control core capability; 4, realizing comprehensive and accurate risk description of the user; and step 5, multi-model credit risk decision-making, including multiple links of marketing, fraud prevention, pre-loan, during loan and after loan.

Description

Technical Field

[0001] The present invention belongs to the technical field of risk control, and particularly relates to a method and system for building intelligent risk control capabilities for consumer finance. Background Art

[0002] In the field of consumer finance, risk management and control are the core links in enterprise operation, directly related to the stability and profitability of the enterprise. Traditional risk control technologies mainly rely on manual review and simple data analysis methods, and there are many problems and drawbacks.

[0003] Firstly, traditional risk control technologies have limitations in data processing. They can often only process limited data types and data volumes, and it is difficult to comprehensively integrate and analyze multi-dimensional data sources, such as self-owned customer data, third-party credit investigation data, social media data, etc., resulting in an insufficient information basis for risk control decisions. Secondly, traditional risk control technologies are relatively backward in model construction and application, lacking the support of advanced intelligent algorithms such as machine learning and deep learning, and unable to accurately evaluate the credit status and risk level of customers, prone to misjudgment and missed judgment. In addition, traditional risk control technologies are not flexible and efficient enough in process management. The risk control process is often rigid and difficult to quickly adjust and optimize according to business needs and market changes, affecting the operation efficiency and competitiveness of the enterprise. Summary of the Invention

[0004] (I) Object of the Invention

[0005] In order to overcome the above deficiencies, the object of the present invention is to provide a method and system for building intelligent risk control capabilities for consumer finance to solve the above technical problems.

[0006] (II) Technical Solution

[0007] To achieve the above object, the technical solution provided by the present application is as follows:

[0008] A method for building intelligent risk control capabilities for consumer finance includes the following steps:

[0009] Step 1: Build a full-process digital risk control system, integrate multi-dimensional data sources for in-depth analysis, and design the architecture of the risk control technology system;

[0010] Step 2: Build a panoramic view of the data link of risk control model variables to realize the whole-process management of data from collection, processing to application;

[0011] Step 3: Build the core capabilities of big data intelligent risk control, including process orchestration, rule strategies, model decision-making, feature variables, data access, shunt experiments, and monitoring and verification;

[0012] Step 4: Achieve comprehensive and accurate risk profiling of users, integrate nautical and comprehensive multi-dimensional data sources through the "Wanwei Smart Space" platform, and use attribution quantitative methods to achieve differentiated customer services;

[0013] Step 5: Multi-model credit risk decision-making, covering marketing, anti-fraud, pre-loan, mid-loan, post-loan and other links, through customized models to achieve comprehensive assessment and decision-making of customer credit risk.

[0014] Preferably, the risk control technology system architecture includes anti-fraud and access management, credit assessment and offer management, mid-loan risk management, and post-loan risk management.

[0015] The anti-fraud and access management module is used to build an anti-fraud model. This model is based on machine learning or deep learning technology to achieve real-time monitoring and early warning of fraudulent behavior, and evaluate customer access by analyzing customer profiles and risk assessment results;

[0016] Credit assessment and offer management module, which is used to assess customer credit status based on multi-dimensional data sources and prediction models, formulate personalized loan policies and preferential strategies, and monitor customer credit changes to adjust loan policies and preferential strategies;

[0017] The loan risk management module is used to monitor customer loan usage and repayment behavior, identify potential risks, and use early warning models to manage risks by grading them. It also implements risk control measures, including adjusting loan amounts and early collection, and reveals customer repayment patterns and characteristics through data analysis.

[0018] The post-loan risk management module is used to establish an intelligent collection system, apply natural language processing and machine learning technologies to match and optimize collection strategies, identify the characteristics and patterns of overdue customers through data analysis, assist in collection work, and use smart contract technology to automate loan management and collection.

[0019] Preferably, in the multi-dimensional data source integration and in-depth analysis in step 1,

[0020] Data sources include, but are not limited to, proprietary customer data, third-party credit data, social media data, and public information data. Proprietary customer information includes basic information, transaction records, and credit history. These data sources are massive, multi-dimensional, and real-time, providing a rich information basis for risk control decisions.

[0021] The in-depth analysis includes forming customer portraits through data cleaning, integration, and mining to reveal the customer's credit status, consumption habits, and risk preferences. At the same time, it uses machine learning and deep learning techniques to build predictive models to predict and evaluate customer behavior, providing more accurate support for risk control decisions.

[0022] Preferably, the panoramic view of the risk control model variable data link includes:

[0023] 1.1 Data acquisition module, which is used to obtain the variable information required for the risk control model from the front end, business systems, credit investigation data, internal services, and third-party data, including but not limited to the basic information, transaction records, credit history, etc. of customers;

[0024] 1.2 Data cleaning module, which is used to preprocess the collected data, including duplicate removal, missing value processing, outlier detection and correction, etc., to ensure the accuracy and consistency of the data;

[0025] 1.3 Data storage module, which is used to store the cleaned data in a suitable data warehouse for subsequent data analysis and model training;

[0026] 1.4 Data access module, which is used to provide an efficient data access interface to support the data requirements in scenarios such as model training and real-time decision-making;

[0027] 1.5 Data warehouse module, as the core of data storage, stores the variable data required for the risk control model and provides data backup, recovery, security and other guarantee measures;

[0028] 1.6 Offline model training sub-module, which is used to train the risk control model based on historical data using machine learning algorithms to mine the rules and patterns in the data;

[0029] 1.7 Real-time variable calculation sub-module, which calculates the variable values required for the risk control model in real time when a transaction occurs to support real-time decision-making;

[0030] 1.8 Data service sub-module, which provides services such as data query, statistics, and analysis to support the formulation and optimization of risk control strategies;

[0031] 1.9 Model platform sub-module, which provides functions such as model management, model release, and model evaluation to support the rapid iteration and optimization of the risk control model;

[0032] 1.10 Decision engine sub-module, which makes real-time approval and decision on the loan applications of customers according to the risk control strategies and model results;

[0033] 1.11 Service interface module, which is used to provide interfaces for interacting with business systems and external data sources to support data import, export, and sharing.

[0034] Preferably, the overall framework of the panoramic view of the risk control model variable data link is divided into four main parts: "design", "implementation", "data processing", and "user interface". These four parts together constitute the macro view of the risk control model variable data link and provide a basic framework for subsequent detailed analysis.

[0035] Preferably, the core capabilities of big data intelligent risk control mainly include:

[0036] Flexible process orchestration management, which realizes the flexible configuration and management of risk control processes through visual process orchestration tools, supports multiple products to share process nodes, and reduces the risk of process changes;

[0037] Visual configuration and release, providing a visual interface to support users in quickly configuring and releasing risk control processes according to business requirements;

[0038] Flexible configuration of rules and strategies, supporting users to flexibly configure multiple rules according to business requirements to form a rule set, and at the same time providing various rule strategy construction methods such as decision trees and decision tables;

[0039] Monitoring of rules and strategies, real-time monitoring of the execution status and effects of rules and strategies, and supporting users to adjust and optimize strategies according to monitoring results;

[0040] Lifecycle management of rules and strategies, supporting the creation, release, testing, verification, online, monitoring, and offline of rules and strategies;

[0041] Fully self-service online release of models, providing a model self-service online release function to support users in quickly deploying models according to business requirements;

[0042] Fully self-service configuration of feature variables, supporting users to flexibly configure feature variables according to business requirements, and providing various feature variable construction templates and complex feature variable SQL construction tools;

[0043] Model monitoring, real-time monitoring of the performance and effects of models, and supporting users to adjust and optimize models according to monitoring results;

[0044] Model lifecycle management, supporting the creation, training, testing, verification, online, monitoring, update, and offline of models;

[0045] Model multi-version management, supporting the management of multiple versions of models, facilitating users to switch and compare between different versions;

[0046] Monitoring of feature variables, real-time monitoring of the changes and abnormal conditions of feature variables, and supporting users to adjust and optimize feature variables according to monitoring results;

[0047] Lifecycle management of feature variables, supporting the creation, configuration, verification, online, monitoring, and offline of feature variables;

[0048] Group management of feature variables, supporting the group management of feature variables, facilitating users to perform batch operations and management;

[0049] Data access, supporting the flexible configuration and access of internal and external data sources, and providing data source permission control functions;

[0050] The third-party data sources can be accessed and invoked on demand, supporting the on-demand access and invocation of third-party data sources according to business requirements, and providing a flexible billing rule definition function for data sources;

[0051] Split testing, including champion challenge experiments, pre-release online verification, and gray-scale production release verification. By comparing the effects of different risk control strategies or models, the optimal strategy or model is selected for online deployment. Pre-release online verification ensures stability and accuracy, and gray-scale production release verification gradually applies the risk control strategy or model to the production environment to reduce risks;

[0052] Monitoring and verification, including data source validity monitoring, abnormal business volume monitoring and alarm, and process historical data effect verification. It monitors the validity and accuracy of data sources in real time to ensure the reliability of the data foundation for risk control strategies or models. It monitors abnormal business volume situations in real time and issues alarm messages in a timely manner to support users for quick response and handling. By comparing the effects of historical data and current data, it verifies the improvement and optimization effects of risk control strategies or models.

[0053] Preferably, the core capabilities of the big data intelligent risk control are achieved through the following technologies:

[0054] Real-time data processing technology, adopting technical means such as real-time data collection, real-time analysis and decision-making, and real-time alarm notification, to achieve real-time adjustment and optimization of risk control strategies;

[0055] Automated deployment and monitoring technology, through automated deployment and monitoring tools, to achieve rapid deployment and real-time monitoring of risk control strategies or models, improving operation and maintenance efficiency and accuracy;

[0056] Data security and privacy protection technology, adopting technical means such as data encryption, access control, and privacy protection to ensure the security and privacy of risk control data;

[0057] Intelligent optimization and iteration technology, using machine learning, deep learning, or other intelligent optimization algorithms to continuously optimize and iterate risk control strategies or models, improving risk control efficiency and accuracy.

[0058] Preferably, the "Wanwei Intelligent Space" is a data analysis platform integrating big data, artificial intelligence, or other advanced technologies. This platform collects and analyzes multi-dimensional data of users to create different customer models to accurately depict the risk characteristics of users;

[0059] To achieve accurate risk characterization, multiple data sources including marine data are integrated. These data sources cover users' identity information, transaction records, credit history, and social behaviors, providing a rich data foundation for subsequent attribution quantification;

[0060] Attribution quantification is a key step in converting the collected data into useful information. By cleaning, processing, and mining the data, key indicators that can reflect the user's willingness and ability to repay debts are extracted and quantified. These quantified indicators provide strong data support for the subsequent risk assessment model.

[0061] Differentiated customer group service, based on the results of attribution quantification, divides users into different customer groups. Each customer group has unique characteristics and risk levels. For different customer groups, differentiated service strategies are formulated to meet their personalized needs.

[0062] Differentiated customer group service, based on the results of attribution quantification, divides users into different customer groups. Each customer group has unique characteristics and risk levels. For different customer groups, differentiated service strategies are formulated to meet their personalized needs.

[0063] Preferably, the multi-model credit risk decision includes:

[0064] The marketing model module is used to optimize the advertising placement strategy through media cooperation OCPA / RTA to improve marketing efficiency and reduce customer acquisition costs, and formulate differentiated marketing strategies for different channels to improve the conversion rate.

[0065] The anti-fraud model module includes a general anti-fraud model for constructing anti-fraud strategies based on historical data to identify potential fraud behaviors, an anti-fraud regional model for formulating targeted anti-fraud strategies in combination with regional characteristics, and a mid-loan fraud behavior model - loop for real-time monitoring of fraud behaviors in the scenario of revolving loans.

[0066] The pre-loan approval model module includes a general approval model for comprehensively evaluating based on information such as the customer's credit record and income status to determine the loan amount and interest rate, intelligent verification for quickly verifying the customer's identity information and contact information through automated means to improve the approval efficiency, and a customer risk portfolio score for comprehensively considering multiple risk factors of the customer and scoring to more accurately assess the risk.

[0067] The mid-loan management model module includes a mid-loan sub-model for dynamically adjusting the loan amount and interest rate according to the customer's repayment behavior and consumption habits after the loan is issued, a quota management model for flexibly adjusting the quota according to the customer's credit status and repayment ability to reduce risks, and a revolving quota risk model for real-time monitoring of risks in the scenario of revolving loans to ensure loan safety.

[0068] Post-loan management model module, including the collection rate prediction model used to predict the customer's future overdue probability and take measures in advance to reduce losses, the collection model used to formulate personalized collection strategies based on the customer's overdue situation to improve the collection rate, the bill of lading model used for post-loan risk monitoring and early warning, and the multi-head model used to evaluate the customer's borrowing and lending situation in multiple financial institutions to prevent risks;

[0069] Marketing response model module, used to predict customer response to marketing activities in order to optimize marketing strategies;

[0070] The consumption fusion model module is used to evaluate the credit status and repayment ability of customers based on their consumption data;

[0071] The disconnection prediction model module is used to predict whether a customer is likely to lose contact so that measures can be taken in advance;

[0072] The customer lifecycle management model module is used to formulate differentiated service strategies according to the customer lifecycle stage to improve customer satisfaction and loyalty.

[0073] A system for building intelligent risk control capabilities for consumer finance, including:

[0074] Data integration module, used to integrate multi-dimensional data sources;

[0075] The risk control system architecture module is used to design the risk control technology system architecture;

[0076] Data link panorama module, used to manage the entire process of data from collection, processing to application;

[0077] Intelligent risk control core capability module, used to build big data intelligent risk control core capabilities;

[0078] User risk profiling module, used to achieve comprehensive and accurate risk profiling of users;

[0079] Credit risk decision module, used to implement multi-model credit risk decision.

[0080] Beneficial effects:

[0081] 1. Improved user identification accuracy: Integrate multi-dimensional data sources and use intelligent algorithms to accurately form customer portraits. For example, discover potential credit risk transmission paths through social behavior data. The customer risk scoring algorithm combines multiple factors to generate accurate scores, effectively identify high-risk users, and reduce bad debt risks.

[0082] 2. Comprehensive enhancement of risk control capabilities: The full-process digital risk control system fine-tunes the management and control of each stage, such as intelligent pre-loan verification to prevent fraud, dynamic adjustment of loan amounts during the loan, post-loan prediction of overdue probability and personalized collection, and a multi-model credit risk decision-making system that flexibly responds to different credit needs and reduces the overall risk level.

[0083] 3. Improved risk control efficiency and flexibility: The intelligent decision-making platform automates decision-making, enables high-concurrency processing to quickly respond to business needs; continuous model verification and optimization shortens the iteration cycle, and flexible process orchestration configuration can quickly adapt to business innovation and market changes.

[0084] 4. Facilitate business innovation and optimize customer experience: Precise risk control supports the design of personalized financial products to meet the needs of different customers; reduces manual review processes, improves the speed and convenience of business handling, and enhances customer satisfaction.

[0085] In summary, through the application of a series of innovative technologies, this patent not only solves the problems existing in traditional consumer finance risk control technologies, but also brings many outstanding effects and advantages to the industry, promoting the development and progress of consumer finance risk control technologies. Brief Description of the Drawings

[0086] Figure 1 is the architecture diagram of the risk control technology system of the present invention;

[0087] Figure 2 is the panoramic diagram of the data link of risk control model variables according to an embodiment of the present invention;

[0088] Figure 3 is the architecture diagram of the core capabilities of big data intelligent risk control according to an embodiment of the present invention;

[0089] Figure 4 is the architecture diagram of the ten-thousand-dimensional intelligent space according to an embodiment of the present invention. Detailed Embodiment

[0090] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the specific embodiments and with reference to the attached Figures 1-4 , drawings. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In addition, in the following descriptions, the descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0091] A method for building an intelligent risk control capability for consumer finance includes the following steps:

[0092] Step 1: Build a full-process digital risk control system, integrate multi-dimensional data sources for in-depth analysis, and design the architecture of the risk control technology system;

[0093] Step 2: Build a panoramic diagram of the data link of risk control model variables to achieve the whole-process management of data from collection, processing to application;

[0094] Step 3: Build the core capabilities of big data intelligent risk control, including process orchestration, rule strategy, model decision, feature variables, data access, diversion experiments and monitoring verification;

[0095] Step 4: Achieve comprehensive and accurate risk profiling of users, integrate nautical and comprehensive multi-dimensional data sources through the "Wanwei Smart Space" platform, and use attribution quantitative methods to achieve differentiated customer services;

[0096] Step 5: Multi-model credit risk decision-making, covering marketing, anti-fraud, pre-loan, mid-loan, post-loan and other links, through customized models to achieve comprehensive assessment and decision-making of customer credit risk.

[0097] The primary task of building a digital risk control system is to integrate multi-dimensional data sources. These data sources include but are not limited to proprietary customer data, third-party credit data, social media data, public information data, etc. The customer data includes basic information, transaction records, and credit history. These data sources are massive, multi-dimensional, and real-time, providing a rich information basis for risk control decisions.

[0098] In-depth analysis is required based on the integration of data sources. Through data cleaning, integration, mining and other means, customer portraits are formed to reveal key information such as customer credit status, consumption habits, risk preferences, etc. At the same time, using machine learning, deep learning and other technologies, predictive models are constructed to predict and evaluate customer behavior, providing more accurate support for risk control decisions.

[0099] The key technologies and implementation strategy analysis include:

[0100] Intelligent decision-making platform: Building an autonomous and controllable decision-making platform is the key to achieving full-process digitalization. The platform should have the ability to flexibly adjust and continuously iterate, and be able to achieve high-concurrency processing based on cloud and microservice architecture. By introducing machine learning algorithms and automated decision-making engines, we can achieve intelligent and automated decision-making. At the same time, through visual interfaces and data analysis tools, we can monitor decision-making effects in real time and provide guidance for subsequent optimization and adjustment.

[0101] Model technology and experimental mechanism: In the risk control system, model technology plays a vital role. Continuously optimize and iterate the risk control model to improve its accuracy and stability. To this end, an experimental mechanism has been established to continuously verify and optimize the model through A / B testing and multi-armed bandit methods. At the same time, it is necessary to pay attention to the interpretability and robustness of the model to ensure that the model can maintain stable performance in a complex and changing risk environment.

[0102] Data Ecosystem Construction: Data is the foundation of the risk control system. To improve the quality and availability of data, it is necessary to strengthen the effective integration of various internal and external information sources. At the same time, by leveraging advanced technologies such as biometric recognition, machine vision, and graph database technology, the capabilities of data collection, processing, and analysis can be enhanced. In addition, a data governance system needs to be established to ensure data compliance, security, and privacy protection.

[0103] Scenario-Differentiated Risk Control Process: The consumer finance field involves various scenarios, such as car purchases, house purchases, education, etc. The risk control requirements and risk characteristics vary in different scenarios. Therefore, we need to formulate differentiated risk control processes for different scenarios. This includes segmenting and profiling customers in different scenarios, formulating differentiated risk assessment criteria and access conditions, designing differentiated loan policies and preferential strategies, etc. Through the scenario-differentiated risk control process, we can ensure the effectiveness and pertinence of risk control measures.

[0104] Compliance and Risk Management: In the process of building a digital risk control system, compliance and risk management are equally important. We need to ensure that all risk control measures comply with relevant laws, regulations, and regulatory requirements. At the same time, by establishing a risk management system and emergency response plan, we can effectively respond to potential risk events and crisis situations.

[0105] Preferably, the risk control technology system architecture includes anti-fraud and access management, credit assessment and offer management, in-loan risk management, and post-loan risk management.

[0106] The anti-fraud and access management module is used to build an anti-fraud model, which is based on machine learning or deep learning technology, realizes real-time monitoring and early warning of fraud behaviors, and evaluates customer access through analyzing customer portraits and risk assessment results.

[0107] The credit assessment and offer management module is used to evaluate the customer's credit status based on multi-dimensional data sources and prediction models, formulate personalized loan policies and preferential strategies, and monitor changes in customer credit to adjust loan policies and preferential strategies.

[0108] The in-loan risk management module is used to monitor the customer's loan usage and repayment behaviors, identify potential risks, and use an early warning model for risk grading management, implement risk control measures, including adjusting the loan amount and early collection, and reveal the customer's repayment patterns and characteristics through data analysis.

[0109] The post-loan risk management module is used to establish an intelligent collection system, apply natural language processing and machine learning technologies to match and optimize collection strategies, identify the characteristics and patterns of overdue customers through data analysis to assist the collection work, and use intelligent contract technology to automate loan management and collection.

[0110] Preferably, in the multi-dimensional data source integration and in-depth analysis in step one,

[0111] the data sources include but are not limited to self-owned customer data, third-party credit investigation data, social media data, and public information data. The self-owned customer information includes basic information, transaction records, and credit history. These data sources are characterized by being massive, multi-dimensional, and real-time, providing a rich information basis for risk control decisions;

[0112] The in-depth analysis includes forming a customer portrait by means of data cleaning, integration, and mining to reveal the customer's credit status, consumption habits, and risk preferences. At the same time, using machine learning and deep learning technologies to build a prediction model to predict and evaluate the customer's behavior, providing more accurate support for risk control decisions.

[0113] Preferably, the panoramic diagram of the variable data link of the risk control model includes:

[0114] 1.1 Data acquisition module, used to obtain the variable information required for the risk control model from the front end, business systems, credit investigation data, internal services, and third-party data, including but not limited to the customer's basic information, transaction records, credit history, etc.;

[0115] 1.2 Data cleaning module, used to preprocess the collected data, including duplicate removal, missing value processing, outlier detection and correction, etc., to ensure the accuracy and consistency of the data;

[0116] 1.3 Data storage module, used to store the cleaned data in a suitable data warehouse for subsequent data analysis and model training;

[0117] 1.4 Data access module, used to provide an efficient data access interface to support the data requirements in scenarios such as model training and real-time decision-making;

[0118] 1.5 Data warehouse module, as the core of data storage, stores the variable data required for the risk control model and provides data backup, recovery, security, and other guarantee measures;

[0119] 1.6 Offline model training sub-module, used to train the risk control model based on historical data using machine learning algorithms to mine the rules and patterns in the data;

[0120] 1.7 Real-time variable calculation sub-module, when a transaction occurs, it calculates the variable values required for the risk control model in real time to support real-time decision-making;

[0121] 1.8 Data service sub-module, providing data query, statistics, analysis, and other services to support the formulation and optimization of risk control strategies;

[0122] 1.9 Model platform sub-module, which provides functions such as model management, model release, and model evaluation, and supports the rapid iteration and optimization of risk control models;

[0123] 1.10 Decision engine sub-module, which conducts real-time approval and decision-making on customers' loan applications according to risk control strategies and model results;

[0124] 1.11 Service interface module, which is used to provide interfaces for interacting with business systems and external data sources, and supports data import, export, and sharing.

[0125] Key technologies and tools include:

[0126] Database: such as Oracle, which is used to store and manage risk control model variable data.

[0127] Operating system: such as Linux, which provides a stable operating environment for data processing and analysis.

[0128] Programming language: such as Java, which is used to develop data processing and analysis programs.

[0129] Software products: such as Eagle (a big data processing and analysis tool), which is used to process and analyze massive data; BLAZE (a high-performance computing framework), which is used to accelerate data processing and model training.

[0130] Details of the data link include:

[0131] Data input, where the data comes from front-end collected data, business system data, credit investigation data, internal services, and third-party data, etc. After being cleaned and preprocessed, the data is stored in the data warehouse.

[0132] Data processing, in the data warehouse, the data is further processed and formed into variables required by the risk control model. Offline model training uses historical data to train the risk control model and updates the model parameters in real time.

[0133] Decision output, the decision engine conducts approval and decision-making on customers' loan applications according to risk control strategies and model results. The decision results are returned to the business system through the service interface to guide subsequent loan issuance and management.

[0134] Feedback and optimization, the business system continuously optimizes and improves the risk control strategies and models according to the decision results and customer feedback. The optimized strategies and models are applied to the approval of new loan applications again, forming a closed-loop risk control system.

[0135] The applications of the risk control model variables include:

[0136] In the pre-loan risk control stage, the risk control model variables are used to evaluate customers' credit status and repayment ability to support the approval and credit granting decisions of loan applications.

[0137] In the post-loan management stage, risk control model variables are used to monitor customers' repayment behaviors and overdue risks to support the formulation and execution of collection strategies.

[0138] In consumption scenarios and cross-selling, risk control model variables are also used to evaluate customers' consumption capabilities and potential demands to support the formulation and optimization of marketing strategies.

[0139] Preferably, the overall framework of the panoramic diagram of the risk control model variable data link is divided into four main parts: "design", "implementation", "data processing", and "user interface". These four parts together constitute a macroscopic view of the risk control model variable data link, providing a basic framework for subsequent detailed analysis.

[0140] Preferably, the core capabilities of big data intelligent risk control mainly include:

[0141] Flexible process orchestration management, realizing flexible configuration and management of risk control processes through a visual process orchestration tool, supporting multiple products to share process nodes, and reducing the risk of process changes;

[0142] Visual configuration and release, providing a visual interface to support users in quickly configuring and releasing risk control processes according to business requirements;

[0143] Flexible configuration of rule strategies, supporting users to flexibly configure multiple rules according to business requirements to form a rule set, and at the same time providing various rule strategy construction methods such as decision trees and decision tables;

[0144] Monitoring of rule strategies, real-time monitoring of the execution status and effects of rule strategies, and supporting users to adjust and optimize strategies according to monitoring results;

[0145] Lifecycle management of rule strategies, supporting the creation, release, testing, verification, online, monitoring, and offline of rule strategies;

[0146] Fully self-service online release of models, providing a model self-service online release function to support users in quickly deploying models according to business requirements;

[0147] Fully self-service configuration of feature variables, supporting users to flexibly configure feature variables according to business requirements, and providing multiple feature variable construction templates and complex feature variable SQL construction tools;

[0148] Model monitoring, real-time monitoring of the performance and effects of models, and supporting users to adjust and optimize models according to monitoring results;

[0149] Model lifecycle management, supporting the creation, training, testing, verification, online, monitoring, update, and offline of models;

[0150] Model multi-version management, which supports the management of multiple versions of models, facilitating users to switch between and compare different versions;

[0151] Feature variable monitoring, which monitors the changes and anomalies of feature variables in real time and supports users to adjust and optimize feature variables according to the monitoring results;

[0152] Feature variable lifecycle management, which supports the creation, configuration, verification, online deployment, monitoring, and offline of feature variables;

[0153] Feature variable grouping management, which supports the grouping management of feature variables, facilitating users to perform batch operations and management;

[0154] Data access, which supports the flexible configuration and access of internal and external data sources and provides data source permission control functions;

[0155] On-demand access and invocation of third-party data sources, which supports the on-demand access and invocation of third-party data sources according to business needs and provides a flexible billing rule definition function for data sources;

[0156] Traffic splitting experiments, including champion challenge experiments, pre-release online verification, and gray-scale production release verification. By comparing the effects of different risk control strategies or models, the optimal strategy or model is selected for online deployment. Pre-release online verification ensures stability and accuracy, and gray-scale production release verification gradually applies the risk control strategy or model to the production environment to reduce risks;

[0157] Monitoring and verification, including data source validity monitoring, abnormal business volume monitoring and alarming, and process historical data effect verification. It monitors the validity and accuracy of data sources in real time to ensure a reliable data foundation for risk control strategies or models, monitors abnormal business volume situations in real time, issues alarm messages in a timely manner, supports users to respond and process quickly, and verifies the improvement and optimization effects of risk control strategies or models by comparing the effects of historical data and current data.

[0158] Preferably, the core capabilities of big data intelligent risk control are realized through the following technologies:

[0159] Real-time data processing technology, which adopts technical means such as real-time data collection, real-time analysis and decision-making, and real-time alarm notification to realize the real-time adjustment and optimization of risk control strategies;

[0160] Automated deployment and monitoring technology, which realizes the rapid deployment and real-time monitoring of risk control strategies or models through automated deployment and monitoring tools, improving operation and maintenance efficiency and accuracy;

[0161] Data security and privacy protection technology, which adopts technical means such as data encryption, access control, and privacy protection to ensure the security and privacy of risk control data;

[0162] Intelligent optimization and iterative techniques, using machine learning, deep learning, or other intelligent optimization algorithms, continuously optimize and iterate the risk control strategies or models to improve the effectiveness and accuracy of risk control.

[0163] Preferably, the "Wanwei Smart Space" is a data analysis platform integrating big data, artificial intelligence, or other advanced technologies. This platform collects and analyzes multi-dimensional data of users to create different customer models for accurately depicting the risk characteristics of users.

[0164] To achieve accurate risk characterization, multiple data sources including marine data are integrated. These data sources cover users' identity information, transaction records, credit history, and social behavior, providing a rich data basis for subsequent attribution quantification.

[0165] Attribution quantification is a key step in converting the collected data into useful information. Through data cleaning, processing, and mining, key indicators reflecting users' willingness and ability to repay debts are extracted and quantified. These quantified indicators provide strong data support for subsequent risk assessment models.

[0166] Differentiated customer group services. Based on the results of attribution quantification, users are divided into different customer groups, each with unique characteristics and risk levels. For different customer groups, differentiated service strategies are formulated to meet their personalized needs.

[0167] Differentiated customer group services. Based on the results of attribution quantification, users are divided into different customer groups, each with unique characteristics and risk levels. For different customer groups, differentiated service strategies are formulated to meet their personalized needs.

[0168] To quantify the risk level of users, a risk scoring model is designed. This model comprehensively considers multiple factors such as users' identity information, transaction behavior, and credit records, and generates a risk score for each user. This score helps quickly identify high-risk users, enabling corresponding risk control measures to be taken. Key variables include credit record variables such as days past due, number of past due times, and default rate, which are used to evaluate customers' willingness and ability to repay debts; financial information variables such as income stability, debt level, and savings rate, which help understand customers' financial status and predict their future debt repayment ability; and consumption behavior variables such as consumption frequency, consumption amount, and consumption type, which can reflect customers' consumption habits and preferences and evaluate their potential risks.

[0169] The scoring formula is based on a complex mathematical model, usually including multiple linear or non-linear equations. The basic score is set according to variables such as the customer's credit record, financial information, and consumption behavior, and then adjusted according to the customer's specific situation to form the final score. The specific formula is: Final score = Basic score + Σ(Adjustment factor i * Weight i), where the adjustment factor i represents the adjustment value for the i-th specific situation, and the weight i represents the relative importance of this adjustment factor in the final score.

[0170] In terms of model application, more than 20,000 basic decision variables including Model A score, Model B score, Model C score, etc. are extracted. These variables cover multiple aspects such as the user's identity information, economic status, and credit history. Using these basic decision variables, multiple risk assessment models such as Model A, Model B, and Model C are constructed, which can accurately assess the user's risk level according to different business scenarios and requirements. In addition, the user's consumption behavior is monitored and predicted through risk control models such as Model F to further improve the risk control efficiency.

[0171] According to the analysis results of the risk assessment model, a number of risk control strategies are carefully designed, covering key links such as credit approval, quota pricing, and collection. Each strategy has been rigorously verified by historical data and carefully optimized to ensure that it can achieve the expected risk control effect in actual application, effectively reduce credit risks, and ensure the stable operation of financial services. In the process of implementing risk control strategies, the principles of diversification and refinement are adhered to. A variety of strategy means such as adjusting the quota and restricting transactions are comprehensively used to accurately control the user's risk and achieve differential management of users with different risk levels. At the same time, a real-time monitoring and feedback mechanism is established to continuously track the implementation effect of the strategy, timely collect feedback information, and dynamically adjust the strategy parameters according to market changes and user behavior to continuously improve the accuracy and effectiveness of the risk control strategy and ensure that the risk control system can flexibly respond to various risk challenges.

[0172] To gain in-depth insights into user characteristics and needs, a comprehensive customer group characteristics management system has been built. This system has powerful real-time monitoring and analysis capabilities, can comprehensively capture users' behavior trajectories and preference changes, and accurately depict the customer group portraits. Through in-depth mining of user data, it provides strong data support for strategy design and optimization, ensuring that various strategies can accurately fit the characteristics of different customer groups and improving the pertinence and effectiveness of strategy implementation. By deeply mining the potential value information in user data, a customer operation value management system has been created. The system comprehensively evaluates key indicators such as user growth and stability, and accurately identifies high-value customer groups. Based on the user value evaluation results, it tailors personalized product and service recommendation plans for them to meet the diverse needs of different users, and improves customer satisfaction and loyalty. At the same time, a continuous tracking mechanism is established to closely monitor changes in user value, timely adjust business strategies, fully tap the potential value of customers, maximize customer value, and promote the sustainable development of the business.

[0173] Preferably, the multi-model credit risk decision-making includes:

[0174] A marketing model module for optimizing advertising placement strategies through media cooperation OCPA / RTA to improve marketing efficiency and reduce customer acquisition costs, and formulating differentiated marketing strategies for different channels to increase conversion rates;

[0175] An anti-fraud model module, including a general anti-fraud model for building anti-fraud strategies based on historical data to identify potential fraud behaviors, an anti-fraud regional model for formulating targeted anti-fraud strategies in combination with regional characteristics, and a mid-loan fraud behavior model - loop for real-time monitoring of fraud behaviors in the context of revolving loans;

[0176] A pre-loan approval model module, including a general approval model for comprehensively evaluating based on information such as customer credit records and income status to determine loan amounts and interest rates, intelligent verification for quickly verifying customer identity information and contact details through automated means to improve approval efficiency, and a customer risk portfolio score for comprehensively considering multiple risk factors of customers and scoring to more accurately assess risks;

[0177] A mid-loan management model module, including a mid-loan sub-model for dynamically adjusting loan amounts and interest rates according to customers' repayment behaviors and consumption habits after loan disbursement, a quota management model for flexibly adjusting quotas according to customers' credit status and repayment ability to reduce risks, and a revolving quota risk model for real-time monitoring of risks in the context of revolving loans to ensure loan safety;

[0178] Post-loan management model module, including the collection rate prediction model used to predict the customer's future overdue probability and take measures in advance to reduce losses, the collection model used to formulate personalized collection strategies based on the customer's overdue situation to improve the collection rate, the bill of lading model used for post-loan risk monitoring and early warning, and the multi-head model used to evaluate the customer's borrowing and lending situation in multiple financial institutions to prevent risks;

[0179] Marketing response model module, used to predict customer response to marketing activities in order to optimize marketing strategies;

[0180] The consumption fusion model module is used to evaluate the credit status and repayment ability of customers based on their consumption data;

[0181] The disconnection prediction model module is used to predict whether a customer is likely to lose contact so that measures can be taken in advance;

[0182] The customer lifecycle management model module is used to formulate differentiated service strategies according to the customer lifecycle stage to improve customer satisfaction and loyalty.

[0183] In the process of building a multi-model credit risk decision-making system, model development and optimization are crucial. We dispatch dedicated data modeling experts to conduct on-site modeling, and develop customized models with the help of China Science and Technology Financial's strong technical capabilities in the field of financial technology. At the same time, we also regularly evaluate and optimize the model to ensure that it can always accurately reflect the credit risk status of customers.

[0184] A system for building intelligent risk control capabilities for consumer finance, including:

[0185] Data integration module, used to integrate multi-dimensional data sources;

[0186] The risk control system architecture module is used to design the risk control technology system architecture;

[0187] Data link panorama module, used to manage the entire process of data from collection, processing to application;

[0188] Intelligent risk control core capability module, used to build big data intelligent risk control core capabilities;

[0189] User risk profiling module, used to achieve comprehensive and accurate risk profiling of users;

[0190] Credit risk decision module, used to implement multi-model credit risk decision.

[0191] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0192] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for building intelligent risk control capabilities for consumer finance, characterized in that: The following steps are involved: Step 1: Build a full-process digital risk control system, integrate multi-dimensional data sources for in-depth analysis, and design the risk control technology system architecture; Step 2: Build a panoramic view of the risk control model variable data link to achieve full-process management of data from collection, processing to application; Step 3: Build the core capabilities of big data intelligent risk control, including process orchestration, rule strategy, model decision, feature variables, data access, diversion experiments and monitoring verification; Step 4: Achieve comprehensive and accurate risk profiling of users, integrate nautical and comprehensive multi-dimensional data sources through the "Wanwei Smart Space" platform, and use attribution quantitative methods to achieve differentiated customer services; Step 5: Multi-model credit risk decision-making, covering marketing, anti-fraud, pre-loan, loan and post-loan links, and achieving comprehensive assessment and decision-making on customer credit risk through customized models.

2. A method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: The risk control technology system architecture includes anti-fraud and access management, credit assessment and offer management, mid-loan risk management, and post-loan risk management. The anti-fraud and access management module is used to build an anti-fraud model. This model is based on machine learning or deep learning technology to achieve real-time monitoring and early warning of fraudulent behavior, and evaluate customer access by analyzing customer profiles and risk assessment results; Credit assessment and offer management module, which is used to assess customer credit status based on multi-dimensional data sources and prediction models, formulate personalized loan policies and preferential strategies, and monitor customer credit changes to adjust loan policies and preferential strategies; The loan risk management module is used to monitor customer loan usage and repayment behavior, identify potential risks, and use early warning models to manage risks by grading them. It also implements risk control measures, including adjusting loan amounts and early collection, and reveals customer repayment patterns and characteristics through data analysis. The post-loan risk management module is used to establish an intelligent collection system, apply natural language processing and machine learning technologies to match and optimize collection strategies, identify the characteristics and patterns of overdue customers through data analysis, assist in collection work, and use smart contract technology to automate loan management and collection.

3. The method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: In the multi-dimensional data source integration and in-depth analysis in step 1, Data sources include, but are not limited to, proprietary customer data, third-party credit data, social media data, and public information data. Proprietary customer information includes basic information, transaction records, and credit history. These data sources are massive, multi-dimensional, and real-time, providing a rich information basis for risk control decisions. The in-depth analysis includes forming customer portraits through data cleaning, integration, and mining to reveal the customer's credit status, consumption habits, and risk preferences. At the same time, it uses machine learning and deep learning techniques to build predictive models to predict and evaluate customer behavior, providing more accurate support for risk control decisions.

4. A method for building intelligent risk control capabilities for consumer finance according to claim 1 or 2, characterized in that: Panorama of the variable data link of the risk control model include: 1.1 Data collection module, used to obtain variable information required by the risk control model from the front-end, business system, credit data, internal services and third-party data, including but not limited to basic customer information, transaction records, and credit history; 1.2 Data cleaning module, used to pre-process the collected data, including deduplication, missing value processing, outlier detection and correction, to ensure the accuracy and consistency of the data; 1.3 Data storage module, used to store the cleaned data in a suitable data warehouse for subsequent data analysis and model training; 1.4 Data access module, which is used to provide an efficient data access interface to support the data requirements of model training and real-time decision-making scenarios; 1.5 Data warehouse module, as the core of data storage, stores variable data required by the risk control model and provides data backup, recovery and security measures; 1.6 Offline model training submodule, which is used to train risk control models based on historical data using machine learning algorithms to mine regularities and patterns in the data; 1.7 Real-time variable calculation submodule: when a transaction occurs, it calculates the variable values ​​required by the risk control model in real time to support real-time decision-making; 1.8 Data service submodule, which provides data query, statistics, and analysis services, and supports the formulation and optimization of risk control strategies; 1.9 Model platform submodule, which provides functions of model management, model publishing, and model evaluation, and supports rapid iteration and optimization of risk control models; 1.10 Decision engine submodule: to review and make decisions on customers’ loan applications in real time based on risk control strategies and model results; 1.11 Service interface module, used to provide interfaces for interacting with business systems and external data sources, and support data import, export and sharing.

5. A method for building intelligent risk control capabilities for consumer finance according to claim 1 or 2, characterized in that: The overall framework of the risk control model variable data link panorama is divided into four main parts: "design", "implementation", "data processing" and "user interface". These four parts together constitute the macro view of the risk control model variable data link and provide a basic framework for subsequent detailed analysis.

6. A method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: The core capabilities of big data intelligent risk control mainly include: Flexible process orchestration management: flexible configuration and management of risk control processes can be achieved through visual process orchestration tools, which supports multiple products sharing process nodes and reduces the risk of process changes; Visual configuration release, providing a visual interface to support users to quickly configure and release risk control processes according to business needs; Flexible configuration of rule strategies supports users to flexibly configure multiple rules according to business needs to form a rule set. It also provides decision trees, decision tables and multiple rule strategy construction methods. Rule strategy monitoring: real-time monitoring of the execution and effects of rule strategies, and support for users to adjust and optimize strategies based on monitoring results; Rule policy lifecycle management, supporting the creation, release, testing, verification, launch, monitoring and de-launch of rule policies; The model can be released online completely by itself, providing the function of self-service model release online, supporting users to quickly deploy models according to business needs; The feature variables are fully self-configured, supporting users to flexibly configure feature variables according to business needs, and providing a variety of feature variable construction templates and complex feature variable SQL construction tools; Model monitoring: real-time monitoring of model performance and effects, supporting users to adjust and optimize models based on monitoring results; Model lifecycle management, supporting model creation, training, testing, verification, launch, monitoring, update, and decommissioning; Model multi-version management: supports multiple version management of models, making it convenient for users to switch and compare between different versions; Feature variable monitoring: real-time monitoring of changes and abnormalities in feature variables, and support for users to adjust and optimize feature variables based on monitoring results; Feature variable lifecycle management, supporting the creation, configuration, verification, launch, monitoring, and de-launch of feature variables; Feature variable group management: support group management of feature variables, convenient for users to perform batch operations and management; Data access, supports flexible configuration and access of internal and external data sources, and provides data source permission management and control functions; On-demand access and invocation of third-party data sources, supporting on-demand access and invocation of third-party data sources according to business needs, and providing flexible data source billing rule definition function; Diversion experiments, including champion challenge experiments, pre-release online verification, and grayscale production release verification, compare the effects of different risk control strategies or models, and select the best strategy or model for online release. Pre-release online verification ensures stability and accuracy, and grayscale production release verification gradually applies risk control strategies or models to the production environment to reduce risks. Monitoring and verification, including data source validity monitoring, business volume anomaly monitoring alarms, and process historical data effect verification, real-time monitoring of the validity and accuracy of data sources, ensuring the data basis of risk control strategies or models is reliable, real-time monitoring of business volume anomalies, timely issuance of alarm information, supporting users to respond and process quickly, and verifying the improvement and optimization effects of risk control strategies or models by comparing the effects of historical data and current data.

7. A method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: The core capabilities of big data intelligent risk control are realized through the following technologies: Real-time data processing technology uses real-time data collection, real-time analysis and decision-making, and real-time alarm notification to achieve real-time adjustment and optimization of risk control strategies; Automated deployment and monitoring technology: Through automated deployment and monitoring tools, rapid deployment and real-time monitoring of risk control strategies or models can be achieved, improving operation and maintenance efficiency and accuracy; Data security and privacy protection technology uses data encryption, access control, and privacy protection techniques to ensure the security and privacy of risk control data; Intelligent optimization and iteration technology uses machine learning, deep learning or other intelligent optimization algorithms to continuously optimize and iterate risk control strategies or models to improve risk control efficiency and accuracy.

8. The method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: The "Wanwei Smart Space" is a data analysis platform that integrates big data, artificial intelligence or other advanced technologies. The platform creates different customer models by collecting and analyzing multi-dimensional data of users to accurately depict the risk characteristics of users; In order to achieve accurate risk characterization, multiple data sources including nautical mile data are integrated. These data sources cover users’ identity information, transaction records, credit history, and social behaviors, providing a rich data foundation for subsequent attribution quantification; Attribution quantification is a key step in converting collected data into useful information. By cleaning, processing and mining data, key indicators that can reflect users' willingness and ability to repay debts are extracted and quantified. These quantitative indicators provide strong data support for subsequent risk assessment models. Differentiated customer service: Based on the results of attribution quantification, users are divided into different customer groups. Each customer group has unique characteristics and risk levels. Differentiated service strategies are formulated for different customer groups to meet their personalized needs. Differentiated customer services divide users into different customer groups based on the results of attribution quantification. Each customer group has unique characteristics and risk levels. Differentiated service strategies are formulated for different customer groups to meet their personalized needs.

9. The method for building intelligent risk control capabilities for consumer finance according to claim 1, characterized in that: The multi-model credit risk decision-making includes: Marketing model module, which is used to optimize advertising delivery strategies through media cooperation OCPA / RTA to improve marketing efficiency and reduce customer acquisition costs, and formulate differentiated marketing strategies for different channels to improve conversion rates; Anti-fraud model module, including the general anti-fraud model for building anti-fraud strategies based on historical data to identify potential fraudulent behaviors, the anti-fraud regional model for formulating targeted anti-fraud strategies based on regional characteristics, and the loan fraud behavior model - cycle for real-time monitoring of fraudulent behaviors in revolving loan scenarios; Pre-loan approval model module, including the general approval model for comprehensive evaluation based on customer credit history and income information to determine the loan amount and interest rate, intelligent verification for rapid verification of customer identity information and contact information through automated means to improve approval efficiency, and customer risk portfolio scoring for comprehensive consideration of multiple risk factors of customers and scoring for more accurate risk assessment; The loan management model module includes a loan sub-model for dynamically adjusting the loan amount and interest rate based on the customer's repayment behavior and consumption habits after the loan is issued, a loan amount management model for flexibly adjusting the loan amount based on the customer's credit status and repayment ability to reduce risks, and a revolving loan risk model for real-time risk monitoring of revolving loans to ensure loan safety; Post-loan management model module, including the collection rate prediction model used to predict the customer's future overdue probability and take measures in advance to reduce losses, the collection model used to formulate personalized collection strategies based on the customer's overdue situation to improve the collection rate, the bill of lading model used for post-loan risk monitoring and early warning, and the multi-head model used to evaluate the customer's borrowing and lending situation in multiple financial institutions to prevent risks; Marketing response model module, used to predict customer response to marketing activities in order to optimize marketing strategies; The consumption fusion model module is used to evaluate the credit status and repayment ability of customers based on their consumption data; The disconnection prediction model module is used to predict whether a customer is likely to lose contact so that measures can be taken in advance; The customer lifecycle management model module is used to formulate differentiated service strategies according to the customer lifecycle stage to improve customer satisfaction and loyalty.

10. A system for building intelligent risk control capabilities for consumer finance, characterized in that: include: Data integration module, used to integrate multi-dimensional data sources; The risk control system architecture module is used to design the risk control technology system architecture; Data link panorama module, used to manage the entire process of data from collection, processing to application; Intelligent risk control core capability module, used to build big data intelligent risk control core capabilities; User risk profiling module, used to achieve comprehensive and accurate risk profiling of users; Credit risk decision module, used to implement multi-model credit risk decision.

Images