Automatic pre-loan risk control model construction method and system based on large model technology

Abstract

The invention provides a pre-loan automatic scoring model construction method and system based on a large model technology, and the method comprises the steps: comprehensively collecting a multi-dimensional basic data set, and providing rich data support for a pre-loan scoring model. According to the method, a pre-loan risk assessment index system of key indexes is customized according to business requirements, and comprehensiveness and accuracy of model assessment are ensured. Through deep feature processing of initial feature derivative data, effective features are extracted, feature data and a subset thereof are constructed, and the prediction capability of the model is enhanced. By applying a large model agent technology, model parameters are automatically iterated and optimized, and optimal model configuration is searched. And finally, through strict model verification steps, the accuracy and reliability of the model are ensured, so that the construction of the pre-loan automatic scoring model is completed. According to the method, the automation level of model construction is improved, the risk assessment capability of the model is enhanced, and a powerful risk management tool is provided for financial institutions.

Description

Technical Field

[0001] The present invention relates to the field of financial risk control technology, and in particular to a method and system for constructing an automatic pre-loan risk control model based on large model technology. Background Art

[0002] With the rapid development of financial services, traditional risk control models have been unable to keep up with the new requirements of risk management. At present, the construction of risk control models mostly relies on the experience of experts and manual feature engineering, which is not only inefficient but also difficult to adapt to the rapid changes in the market environment. At the same time, although some automated machine learning technologies have made progress in improving the efficiency of model construction, they still have limitations in terms of the automatic update speed of the model, the level of intelligence, the diversity of risk assessment indicators, and the prevention of overfitting. Therefore, there is an urgent need to develop a new type of automated and intelligent risk control model construction method that can improve the efficiency and accuracy of the risk control process and have the ability to quickly adapt to market changes.

[0003] In this context, financial institutions are increasingly demanding intelligent risk control and model algorithms, and financial risk control is undergoing earth-shaking changes. Financial institutions need a new approach that can fully utilize the capabilities of large models to improve the accuracy and efficiency of risk control models through automatic learning and pattern recognition capabilities. At the same time, this approach needs to have strong adaptability and generalization capabilities, be able to automatically learn based on new data, and adjust the parameters and structure of the model in a timely manner to maintain the accuracy and stability of the model. Summary of the invention

[0004] The present invention aims at the technical problems existing in the prior art and provides a method and system for constructing an automatic pre-loan risk control model based on large model technology. By constructing a detailed hyperparameter space and using large model technology to drive hyperparameter optimization, it ensures that the model can be effectively optimized for the preset risk assessment target. The invention can automatically construct an intelligent and efficient pre-loan risk assessment model to adapt to the rapid changes in the financial market and provide strong risk management support for financial institutions.

[0005] According to a first aspect of the present invention, a method for constructing an automatic pre-loan risk control model based on a large model technology is provided, comprising the following steps:

[0006] Step 1. Obtain initial feature data through borrower personal information and enterprise information, and form an original feature data set with all user original feature data used for modeling;

[0007] Step 2. Perform feature derivation on the original feature data set to generate a derived sample data set;

[0008] Step 3. Divide the derived sample data set into a training sample data set, a test data set, and an out-of-time sample test set according to requirements;

[0009] Step 4. Establish a pre-loan risk control model learning flow;

[0010] Step 5. Set up a risk assessment target system, build a model optimization agent based on the big model, and perform intelligent model iteration;

[0011] Step 6. Validate the model using the test dataset and the out-of-temporal sample dataset.

[0012] Based on the above technical solution, the present invention can also make the following improvements.

[0013] Optionally, the obtaining of initial characteristic data through borrower personal information and enterprise information includes:

[0014] Obtain the borrower's basic personal information or the company's credit report, collect the company's financial statements, and classify and process the data to form a comprehensive business indicator system that covers multiple aspects of personal information, corporate credit, and financial status. At the same time, based on the actual default situation of the borrower, form an original feature data set, where 0 indicates no default and 1 indicates default.

[0015] Optionally, the basic personal information includes name, ID number, and corresponding personal credit report and credit record; the enterprise's credit report includes overdue status and litigation information; the enterprise's financial statements include income statement, balance sheet and tax invoice data.

[0016] Optionally, the deriving features from the original feature data set includes:

[0017] Feature derivation is performed by deriving features based on business logic, deriving features with the help of data analysis technology, and building an automatic feature derivation agent.

[0018] Optionally, the features derived based on business logic include: industry professional features, operating characteristics, market share and competitiveness characteristics, key customer relationship characteristics; the features derived with the help of data analysis technology include: polynomial features, cross features, sliding window statistical features, mean, standard deviation and clustering features; building an automatic feature derivation agent includes: automatically deriving features using large model technology.

[0019] Optionally, the pre-loan risk control model includes data preprocessing, feature engineering, machine learning model and model evaluation; the data preprocessing link includes missing value processing, outlier processing and noise data processing; the feature engineering link includes data conversion, feature extraction, feature selection and feature screening; the machine learning link includes linear regression model, ensemble tree model, deep learning model; model evaluation includes regression model evaluation method and classification model integration method, and all methods and parameters in the entire model together constitute the parameter space.

[0020] Optionally, constructing a model optimization agent based on a large model to perform model intelligent iteration includes:

[0021] Set the initial values ​​of decision variables;

[0022] For a given decision variable, calculate the value of the objective function under the current decision variable;

[0023] Agent-based model updating optimizes model parameters by simulating the functional modules of observation, thinking, action and memory of biological agents.

[0024] Optionally, the optimization of model parameters by simulating the functional modules of observation, thinking, action and memory of biological intelligent bodies includes:

[0025] The agent first analyzes the relationship between the current parameters and the objective function value through the observation module and stores this key information in the memory module;

[0026] The agent enters the thinking module to conduct in-depth analysis of the collected data, and uses logical reasoning and decision-making capabilities to predict the potential impact of different parameter selections on model performance, and find parameter combinations that can maximize or optimize the objective function value;

[0027] The action module implements the decisions made by the agent in the thinking module, adjusts the model parameters, and tests the effects of these new parameters in the real environment;

[0028] The memory module stores historical data and experience, and continuously updates and optimizes the agent's knowledge base.

[0029] According to a second aspect of the present invention, there is provided an automatic pre-loan risk control model construction system based on a large model technology, comprising:

[0030] A data acquisition module is used to acquire initial feature data through borrower personal information and enterprise information, and form an original feature data set from all user original feature data used for modeling;

[0031] The data processing module is used to derive features from the original feature data set to generate a derived sample data set; the derived sample data set is divided into a training sample data set, a test data set, and an out-of-time sample test set according to requirements;

[0032] The risk control model building module is used to establish the pre-loan risk control model learning flow, build the hyperparameter space and risk assessment target system, and build a model optimization agent based on the large model to perform model intelligent iteration;

[0033] The model validation module is used to validate the model using the test dataset and the out-of-time sample dataset.

[0034] According to a third aspect of the present invention, there is provided an electronic device comprising a memory and a processor, wherein the processor is configured to implement the steps of a method for constructing an automatic pre-loan risk control model based on large model technology when executing a computer program stored in the memory.

[0035] Technical effects and advantages of the present invention:

[0036] The present invention proposes a method and system for constructing an automatic pre-loan risk control model based on large model technology. By constructing a detailed hyperparameter space and using large model technology to drive hyperparameter optimization, it ensures that the model can be effectively optimized for the preset risk assessment goals. This method accurately controls the model optimization process through a carefully formulated risk assessment target system and a large model intelligent agent, thereby effectively avoiding overfitting while improving model performance and enhancing the generalization ability of the model. Ultimately, this method can automatically construct a pre-loan risk assessment model that is both intelligent and efficient to adapt to the rapid changes in the financial market and provide strong risk management support for financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flow chart of a method for constructing an automatic pre-loan risk control model based on a large model technology provided by an embodiment of the present invention;

[0038] Figure 2 A flowchart of automatic iteration of a large model agent provided by an embodiment of the present invention;

[0039] Figure 3 A schematic diagram of the hardware structure of a possible electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] It is understandable that based on the defects in the background technology, the embodiment of the present invention proposes a method for constructing an automatic pre-loan risk control model based on a large model technology, specifically as follows: Figure 1 As shown, the following steps are included:

[0042] Step 1: Obtain initial feature data through borrower personal information and enterprise information. All user original feature data used for modeling form an original feature data set;

[0043] In the above technical solution, the acquisition of initial feature data through borrower personal information and enterprise information includes the following steps:

[0044] First, obtain the borrower’s basic personal information, such as name, ID number, and corresponding personal credit report and credit record.

[0045] For enterprises, we obtain the credit reports of credit bureaus and analyze the credit records of enterprises in detail, including overdue situations and litigation information. At the same time, we collect the financial statements of enterprises, including profit and loss statements, balance sheets, etc., as well as tax invoice data, such as value-added tax invoices and corporate income tax reports.

[0046] Then, the above data are classified and processed to form a comprehensive business indicator system, covering personal information, corporate credit, financial status and other aspects. At the same time, based on the actual default situation of the borrower, the original feature data set is formed, where 0 indicates no default and 1 indicates default.

[0047] Step 2: derive features from the original feature data set to generate a derived sample data set;

[0048] In the above technical solution, the original feature data set is derived from three aspects: specifically, on the one hand, features are derived based on business logic, such as industry professional features, business features, market share and competitiveness features, key customer relationship features, etc.; on the other hand, features are derived with the help of data analysis technology, such as polynomial features, cross features, sliding window statistical features, mean, standard deviation, clustering features, etc.; on the third hand, an automatic feature derivation agent is constructed to automatically derive features using large model technology. The features obtained by the above three methods form a derived sample data set.

[0049] Step 3: Divide the derived sample data set into a training sample data set, a test data set, and an out-of-time sample test set according to requirements;

[0050] Step 4: Establish a pre-loan risk control model learning flow;

[0051] It should be noted that the pre-loan risk control model is a pipeline flow in which the optimization target and iteration direction are controlled by a large model. It can combine the data preprocessing steps, feature engineering and machine learning models that may be used in the entire modeling process to build the entire machine learning process. The entire machine learning flow is a model. The entire model contains four core nodes: data preprocessing->feature engineering->machine learning model->model evaluation. The data preprocessing link includes missing value processing, outlier processing, and noise data processing; the feature engineering link includes data conversion, feature extraction, feature selection, and feature screening; the machine learning link includes linear regression models, ensemble tree models, deep learning models and other algorithm models; model evaluation includes regression model evaluation methods and classification model integration methods. All methods and parameters in the entire model together constitute the parameter space.

[0052] Step 5: Construct a hyperparameter space and risk assessment target system, and build a model optimization agent based on the large model to perform model intelligent iteration;

[0053] In the above technical solution, building a hyperparameter space includes "creating a detailed hyperparameter space, which will include the feature engineering methods and machine learning models and their corresponding parameters that may be involved in each link of the machine learning process. In addition, the knowledge base will contain a corresponding hyperparameter space explanation document, which details the function and purpose of each parameter, so that the large model can accurately understand and apply it when making optimization decisions to improve the efficiency and effect of model tuning.

[0054] Building a risk assessment target system includes: When building a pre-loan risk assessment model for small and micro enterprises, first clearly define multiple assessment targets to comprehensively measure the performance of the model. These targets include but are not limited to maximizing AUC and maximizing KS. In addition to these deterministic indicators, a fuzzy target assessment system can be formulated using the large model's prompt word engineering.

[0055] Building a model optimization agent based on a large model to perform model intelligent iteration specifically includes:

[0056] Initialization: In this optimization method, you first need to set the initial values ​​of the decision variables;

[0057] Objective function calculation: For a given decision variable, calculate the value of the objective function under the current decision variable.

[0058] Agent-based model updating mechanism: It optimizes model parameters by simulating the four basic functional modules of biological agents: observation, thinking, action and memory.

[0059] Specific as Figure 2 As shown in the figure, by simulating the four basic functional modules of biological intelligence, namely observation, thinking, action and memory, the model parameters are optimized, including:

[0060] The agent first analyzes the relationship between the current parameters and the objective function value through the observation module, and stores this key information in the memory module for subsequent reference and learning. This observation is not just a simple record of the current state, but a deep understanding of the model performance and parameter impact.

[0061] The agent then enters the thinking module, which is the core of the agent. It needs to conduct in-depth analysis of the collected data and use logical reasoning and decision-making capabilities to predict the potential impact of different parameter selections on model performance. Here, the agent will use various tools and algorithms, such as Bayesian optimization, genetic algorithms, or reinforcement learning, to explore the hyperparameter space and find parameter combinations that can maximize or optimize the objective function value.

[0062] The action module is responsible for implementing the decisions made by the agent in the thinking module, adjusting the model parameters, and testing the effects of these new parameters in the actual environment. This step is the key to verifying the results of the agent's thinking and is also an indispensable practical link in the model update process.

[0063] Finally, the memory module not only stores historical data and experience, but also continuously updates and optimizes the agent’s knowledge base. These memories are crucial for the agent’s future decisions, as they provide valuable information about which parameter choices are effective and which are not.

[0064] Through the close collaboration and iterative cycles of the above four modules, the intelligent agent can gradually approach the optimal model parameters.

[0065] Step 6: Validate the model using the test dataset and the out-of-time sample dataset.

[0066] The embodiment of the present invention adopts the above technical solution and has the following beneficial effects:

[0067] Automation and efficiency improvement: Automated parameter adjustment and model optimization reduce the need for manual intervention, save a lot of time and resources, and improve the efficiency of model development and deployment.

[0068] Precise Optimization: The agent is able to accurately find the optimal solution in the hyperparameter space, which is often more accurate than manually adjusting parameters and can discover optimal parameter combinations that humans may overlook.

[0069] Adaptability and flexibility: The agent can quickly adjust model parameters based on new data and feedback, allowing the model to adapt to changes in data distribution and changes in the environment.

[0070] Reduce overfitting: By continuously iterating and testing different parameters, the agent helps find the best parameters to balance model complexity and generalization ability, reducing the risk of overfitting. At the same time, the continuous learning and optimization of the agent helps improve the performance of the model on unseen data and enhance the generalization ability of the model.

[0071] Improve decision quality: The thinking module of the intelligent agent can integrate a variety of decision support tools and algorithms to improve the quality and reliability of decision making.

[0072] Continuous learning and improvement: The agent’s memory module enables the model to learn from historical experience and continuously improve, achieving continuous self-optimization. This approach provides new avenues for research and innovation in machine learning, especially in the fields of automated machine learning and intelligent decision-making systems.

[0073] According to a second aspect of the present invention, there is provided an automatic pre-loan risk control model construction system based on a large model technology, comprising:

[0074] A data acquisition module is used to acquire initial feature data through borrower personal information and enterprise information, and form an original feature data set from all user original feature data used for modeling;

[0075] The data processing module is used to derive features from the original feature data set to generate a derived sample data set; the derived sample data set is divided into a training sample data set, a test data set, and an out-of-time sample test set according to requirements;

[0076] The risk control model building module is used to establish the pre-loan risk control model learning flow, build the hyperparameter space and risk assessment target system, and build a model optimization agent based on the large model to perform model intelligent iteration;

[0077] The model validation module is used to validate the model using the test dataset and the out-of-time sample dataset.

[0078] It can be understood that the automatic pre-loan risk control model construction system based on big model technology provided by the present invention corresponds to the automatic pre-loan risk control model construction method based on big model technology provided in the aforementioned embodiments. The relevant technical features of the automatic pre-loan risk control model construction system based on big model technology can refer to the relevant technical features of the automatic pre-loan risk control model construction method based on big model technology, which will not be repeated here.

[0079] According to a third aspect of the present invention, an embodiment of the present invention further provides an electronic device, specifically as follows Figure 3As shown, the electronic device includes: a processor, a communication interface, a memory, and a communication bus, wherein the processor 810, the communication interface, and the memory 830 communicate with each other through the communication bus. The processor 810 can call the logic instructions in the memory 830 to execute the implementation steps of a method for building an automatic pre-loan risk control model based on a large model technology.

[0080] In addition, the logic instructions in the above-mentioned memory 810 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0081] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0082] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0083] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0084] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0085] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a pre-loan automatic risk control model based on large model technology, characterized in that: The following steps are involved: Step 1. Obtain initial feature data through borrower personal information and enterprise information, and form an original feature data set with all user original feature data used for modeling; Step 2. Perform feature derivation on the original feature data set to generate a derived sample data set; Step 3. Divide the derived sample data set into a training sample data set, a test data set, and an out-of-time sample test set according to requirements; Step 4. Establish a pre-loan risk control model learning flow; Step 5. Construct a hyperparameter space and risk assessment target system, and build a model optimization agent based on the large model to perform model intelligent iteration; Step 6. Validate the model using the test dataset and the out-of-temporal sample dataset.

2. According to claim 1, a method for constructing a pre-loan automatic risk control model based on large model technology is characterized in that: The initial feature data obtained through the borrower's personal information and corporate information includes: Obtain the borrower's basic personal information or the company's credit report, collect the company's financial statements, and classify and process the data to form a comprehensive business indicator system that covers multiple aspects of personal information, corporate credit, and financial status. At the same time, based on the actual default situation of the borrower, form an original feature data set, where 0 indicates no default and 1 indicates default.

3. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 2 is characterized in that: The basic personal information includes name, ID number, and corresponding personal credit report and credit record; the company's credit report includes overdue status and litigation information; the company's financial statements include income statement, balance sheet and tax invoice data.

4. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 1 is characterized in that: The feature derivation of the original feature data set includes: Feature derivation is performed by deriving features based on business logic, deriving features with the help of data analysis technology, and building an automatic feature derivation agent.

5. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 4 is characterized in that: The derived features based on business logic include: industry professional features, operating characteristics, market share and competitiveness characteristics, key customer relationship characteristics; the derived features with the help of data analysis technology include: polynomial features, cross features, sliding window statistical features, mean, standard deviation and clustering features; building an automatic feature derivation agent includes: automatically deriving features using large model technology.

6. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 1 is characterized in that: The pre-loan risk control model includes data preprocessing, feature engineering, machine learning model and model evaluation; the data preprocessing link includes missing value processing, outlier processing and noise data processing; the feature engineering link includes data conversion, feature extraction, feature selection and feature screening; the machine learning link includes linear regression model, ensemble tree model and deep learning model; model evaluation includes regression model evaluation method and classification model integration method. All methods and parameters in the entire model together constitute the parameter space.

7. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 1 is characterized in that: The method of constructing a model optimization agent based on a large model to perform model intelligent iteration includes: Set the initial values ​​of decision variables; For a given decision variable, calculate the value of the objective function under the current decision variable; Agent-based model updating optimizes model parameters by simulating the functional modules of observation, thinking, action and memory of biological agents.

8. The method for constructing a pre-loan automatic risk control model based on large model technology according to claim 7 is characterized in that: The optimization of model parameters by simulating the functional modules of observation, thinking, action and memory of biological intelligent bodies includes: The agent analyzes the relationship between the current parameters and the objective function value through the observation module, and stores this relationship information in the memory module; The agent enters the thinking module to conduct in-depth analysis of the collected data, and uses logical reasoning and decision-making capabilities to predict the potential impact of different parameter selections on model performance, and find parameter combinations that can maximize or optimize the objective function value; The action module implements the decisions made by the agent in the thinking module, adjusts the model parameters, and tests the effects of the parameters in the actual environment; The memory module stores historical data and experience, and continuously updates and optimizes the agent's knowledge base.

9. A pre-loan automatic risk control model construction system based on large model technology, characterized in that: include: A data acquisition module is used to acquire initial feature data through borrower personal information and enterprise information, and form an original feature data set from all user original feature data used for modeling; The data processing module is used to derive features from the original feature data set to generate a derived sample data set; the derived sample data set is divided into a training sample data set, a test data set, and an out-of-time sample test set according to requirements; The risk control model building module is used to establish the pre-loan risk control model learning flow, build the hyperparameter space and risk assessment target system, and build a model optimization agent based on the large model to perform model intelligent iteration; The model validation module is used to validate the model using the test dataset and the out-of-time sample dataset.

10. An electronic device, characterized in that: It includes a memory and a processor, and the processor is used to implement a pre-loan automatic risk control model construction method based on large model technology as described in any one of claims 1 to 8 when executing a computer program stored in the memory.

Images