Credit assessment method and device, electronic equipment, storage medium and program product
By quantifying and labeling basic enterprise information, and combining backpropagation algorithm and adaptive learning rate to train a credit assessment model, the problem of low accuracy in credit assessment in traditional methods is solved, and high-precision and robust credit assessment is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA UNITED NETWORK COMM GRP CO LTD
- Filing Date
- 2025-12-31
- Publication Date
- 2026-04-28
AI Technical Summary
The accuracy of corporate credit assessment in existing technologies has decreased. Traditional expert scoring models are easily affected by subjective factors, and traditional machine learning models require manual feature design, resulting in poor feature extraction quality, making it difficult to meet the needs of large-scale, dynamic and high-precision credit scoring.
By acquiring basic enterprise information and performing numerical quantification, an enterprise information dataset is generated. The data is then labeled, and a credit assessment model is trained using the backpropagation algorithm and adaptive learning rate to improve assessment accuracy.
It improves the accuracy of credit assessment models, enhances the accuracy and precision of corporate credit assessment, and improves robustness to complex data distributions.
Smart Images
Figure CN121935531A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a credit assessment method, apparatus, electronic device, storage medium, and program product. Background Technology
[0002] Corporate credit assessment is a core component of financial risk control, supply chain management, and business partnership decisions. The accuracy of corporate credit scores directly impacts financial institutions' lending decisions and partners' credit authorization. With the surge in the number of enterprises and the increasing complexity of economic activities, traditional manual assessment methods are no longer sufficient to meet the demands for large-scale, dynamic, and highly accurate credit scoring.
[0003] In existing technologies, calculating corporate credit scores mainly relies on expert scoring models or traditional machine learning models.
[0004] However, existing expert scoring models rely on manual rules, and the scoring results are easily affected by subjective factors. Traditional machine learning models require manual feature design, and the quality of feature extraction directly affects model performance. Due to the dependence on input features, the existing credit scoring methods lead to a decrease in the accuracy of corporate credit assessment. Summary of the Invention
[0005] This application provides a credit assessment method, apparatus, electronic device, storage medium, and program product to address the problem of reduced accuracy in corporate credit assessments in the prior art.
[0006] In a first aspect, embodiments of this application provide a request scheduling method for a model inference engine, including:
[0007] Acquire basic enterprise information and perform numerical quantification on the basic enterprise information to generate an enterprise information dataset;
[0008] The enterprise information dataset is labeled to generate model input data;
[0009] The input data of the model is input into the credit assessment model to generate a credit assessment result. The credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
[0010] In one possible implementation, the step of acquiring basic enterprise information and performing numerical quantification processing on the basic enterprise information to generate an enterprise information dataset includes: acquiring basic enterprise information; performing quantification processing on the basic enterprise information to generate quantified information; calculating a two-dimensional matrix of enterprise information based on the quantified information; and generating an enterprise information dataset based on the two-dimensional matrix of enterprise information.
[0011] In one possible implementation, the step of labeling the enterprise information dataset to generate model input data includes: labeling the enterprise information dataset to generate an enterprise credit information label dataset; scoring the enterprise credit information label dataset according to a preset scoring rule to generate an enterprise credit information label scoring dataset; and performing data preprocessing on the enterprise credit information label scoring dataset and the enterprise information dataset to generate model input data.
[0012] In one possible implementation, after labeling the enterprise information dataset and generating model input data, the method further includes: initializing the connection weights of the credit assessment model; filtering neurons in the credit assessment model using a temporary fallback method to obtain a filtered credit assessment model; and training the filtered credit assessment model according to the connection weights and a preset number of iterations to generate a trained credit assessment model.
[0013] In one possible implementation, training the selected credit assessment model based on the connection weights and a preset number of iterations to generate the trained credit assessment model includes: calculating the output value of each neuron in the selected credit assessment model based on the connection weights; calculating a loss function based on the output value of each neuron and the credit assessment prediction value; and training the credit assessment model based on the loss function and a preset number of iterations to generate the trained credit assessment model.
[0014] In one possible implementation, after calculating the loss function based on the output value of each neuron and the credit assessment prediction value, the method further includes: calculating the gradient of each connection weight in the credit assessment model based on the loss function; determining the learning rate for updating each connection weight based on the number of iterations of the credit assessment model; updating each connection weight based on the gradient of each connection weight and the learning rate for updating each connection weight, thereby generating optimized connection weights.
[0015] Secondly, embodiments of this application provide a credit assessment device, comprising:
[0016] The acquisition module is used to acquire basic enterprise information and perform numerical quantification processing on the basic enterprise information to generate an enterprise information dataset.
[0017] The tagging module is used to tag the enterprise information dataset and generate model input data;
[0018] The generation module is used to input the model input data into the credit assessment model and generate credit assessment results, wherein the credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
[0019] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;
[0020] The memory stores computer-executed instructions;
[0021] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.
[0022] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.
[0023] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.
[0024] The credit assessment method, apparatus, electronic device, storage medium, and program product of the model inference engine provided in this application embodiment generate an enterprise information dataset by numerically quantifying the basic information of enterprises, marking the enterprise information, generating model input data, assessing enterprise credit using a credit assessment model, and training the credit assessment model through backpropagation algorithm and adaptive learning rate to improve the assessment accuracy of the credit assessment model, thereby improving the accuracy of enterprise credit assessment. Attached Figure Description
[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0026] Figure 1 This is a schematic diagram of the system structure of a computer device provided in an embodiment of this application;
[0027] Figure 2 Flowchart of the credit assessment method provided for this application Figure 1 ;
[0028] Figure 3 Flowchart of the credit assessment method provided for this application Figure 2 ;
[0029] Figure 4 A schematic diagram of the backpropagation algorithm provided in an embodiment of this application;
[0030] Figure 5 A schematic diagram illustrating the temporary withdrawal method for screening neurons provided in an embodiment of this application;
[0031] Figure 6 Flowchart of the credit assessment method provided for this application Figure 3 ;
[0032] Figure 7 A schematic diagram of the credit assessment device provided for this application;
[0033] Figure 8 A schematic diagram of the structure of the electronic device provided in this application.
[0034] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation
[0035] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0036] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0037] Corporate credit assessment is a core component in scenarios such as financial risk control, supply chain management, and business cooperation decisions. The accuracy of corporate credit scores directly impacts financial institutions' lending decisions and partners' credit authorization. With the surge in the number of enterprises and the increasing complexity of economic activities, traditional manual assessment methods are struggling to meet the demands for large-scale, dynamic, and high-precision credit scoring. Current technologies primarily rely on expert scoring models or traditional machine learning models to calculate corporate credit scores. However, expert scoring models depend on manually defined rules, making the results susceptible to subjective influences. Traditional machine learning models require manually designed features, and the quality of feature extraction directly affects model performance. This reliance on input features leads to a decrease in the accuracy of current credit scoring methods.
[0038] To address the aforementioned technical problems, this application proposes the following technical concept: Considering the acquisition of multi-dimensional enterprise basic information, the inventors quantify and process this information into a uniform, calculable numerical form to solve the data heterogeneity problem. The quantified data is then labeled to generate model input data. This model input data is then fed into a credit assessment model to evaluate the enterprise's credit. Furthermore, considering the use of backpropagation algorithm and adaptive learning rate to train the credit assessment model, the adaptive learning rate mechanism balances model training speed and accuracy, and the backpropagation algorithm enables multi-dimensional evaluation, thereby improving the accuracy of enterprise credit assessment.
[0039] Figure 1 This is a schematic diagram of the system architecture of the computer device provided in an embodiment of this application. Figure 1 As shown, the computer device includes: a receiving device 101, a processing device 102, and a display device 103.
[0040] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the credit assessment method. In other feasible embodiments of this application, the above architecture may include more or fewer components than illustrated, or combine some components, or split some components, or arrange different components, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown can be implemented in hardware, software, or a combination of both.
[0041] In the specific implementation process, the receiving device 101 can be an input / output interface or a communication interface, which can obtain basic enterprise information.
[0042] The processing device 102 can perform credit assessments based on basic enterprise information and generate credit assessment results.
[0043] The display device 103 can be used to display the aforementioned credit assessment results, etc.
[0044] The display device can also be a touch screen, used to receive user commands while displaying the above content, so as to realize the operation interaction with the user.
[0045] It should be understood that the above-mentioned processing device can be implemented by a processor reading instructions from memory and executing those instructions, or it can be implemented by a chip circuit.
[0046] Furthermore, the network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
[0047] Figure 2 Flowchart of the credit assessment method provided for this application Figure 1 ,like Figure 2 As shown, the method includes:
[0048] S201: Obtain basic enterprise information and perform numerical quantification on the basic enterprise information to generate an enterprise information dataset.
[0049] Specifically, the process involves acquiring basic enterprise information, quantifying it to generate quantitative information, calculating a two-dimensional matrix of enterprise information based on the quantitative information, and generating an enterprise information dataset based on the two-dimensional matrix of enterprise information.
[0050] S202: Label the enterprise information dataset to generate model input data.
[0051] Specifically, the enterprise information dataset is labeled, the enterprise loan information labeled dataset is scored according to preset scoring rules to generate an enterprise credit information labeled scoring dataset, and the enterprise credit information labeled scoring dataset and the enterprise information dataset are preprocessed to generate model input data.
[0052] S203: Input the model input data into the credit assessment model to generate the credit assessment result. The credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
[0053] In this embodiment, the credit assessment result is denoted as .
[0054] For example, based on the credit assessment results Determine whether to grant credit, if ,and ,and ,and ,and Then the loan amount is A1, the loan period is B1, and the enterprise credit information marker data is recorded. .
[0055] in, A1 and B1 are hyperparameters that are set manually.
[0056] Specifically, update the enterprise credit information tagging dataset. Data after enterprise evaluation Update to the Enterprise Credit Information Tag Dataset This provides data support for subsequent modeling and scoring.
[0057] As can be seen from the above embodiments, by quantifying the basic information of enterprises, an enterprise information dataset is generated, the enterprise information is labeled, model input data is generated, the credit assessment model is used to assess the credit of enterprises, and the credit assessment model is trained by backpropagation algorithm and adaptive learning rate to improve the assessment accuracy of the credit assessment model, thereby improving the accuracy of enterprise credit assessment.
[0058] In one embodiment of this application, step S201 includes:
[0059] S2011: Obtain basic enterprise information, quantify the basic enterprise information, and generate quantitative information.
[0060] In this embodiment, the methods for obtaining basic enterprise information include, but are not limited to, obtaining it in text form, obtaining it in voice form, and obtaining it by calling the enterprise website interface.
[0061] In this embodiment, the basic enterprise information includes, but is not limited to, asset information, profit information, and cash flow information.
[0062] In this embodiment, the basic enterprise information is denoted as... .
[0063] Where N represents the total number of dimensions describing the basic information of an enterprise, which can be asset information, profit information, and cash flow information. In this embodiment, the number of basic enterprise information is n. This represents the individual information corresponding to the enterprise-level information.
[0064] For example, profit information is recorded as Net operating profit margin is denoted as The net profit margin on total assets is denoted as .
[0065] Specifically, the basic information of enterprises is quantified numerically, based on the description of each piece of enterprise information in the input information. The basic information of an enterprise is quantified to obtain quantified values. .
[0066] For example, corporate credit records are recorded as Among them, the number of historical defaults is recorded as Historical loan amounts are recorded as Historical loan terms are recorded as Historical loan rating is recorded as The quantified value of the number of historical defaults If the quantified number of historical defaults is less than 3, then .
[0067] in, and These are hyperparameters, set by the developers, and and The value is greater than or equal to 0 and less than or equal to 255.
[0068] S2012: Calculate a two-dimensional matrix of enterprise information based on quantitative information.
[0069] Specifically, based on the enterprise's basic information dimension N and the number of information descriptions in each dimension. , to perform dimension calculation of the two-dimensional information matrix.
[0070] Specifically, finding the maximum number of elements in array M is expressed as:
[0071]
[0072] The dimensions of a two-dimensional matrix are N×M.
[0073] Specifically, construct a two-dimensional data matrix of level 0 enterprise information. Based on the calculated dimensions of the two-dimensional matrix, supplement data with fewer than M columns by adding (M-) columns. Construct a two-dimensional data matrix of enterprise information of size N×M using 0 elements. .
[0074] Specifically, a Level 1 enterprise information data matrix is generated based on the two-dimensional enterprise information data matrix. The calculation method is as follows:
[0075]
[0076] In the formula, This represents the Hadamard product operation; This indicates the calculation factors required to generate Level 1 information data, which can be set manually and then input by the computer.
[0077] For example, net operating profit margin The score can be adjusted based on industry or output value to obtain a higher score.
[0078] S2013: Generate an enterprise information dataset based on a two-dimensional matrix of enterprise information.
[0079] Specifically, each enterprise (numbered as follows) is counted repeatedly. 0-level enterprise information two-dimensional data matrix and Level 1 information data matrix ,in The calculation formula is:
[0080]
[0081] In the formula, This is a Level 1 information data matrix, representing the first level. Enterprise information data of individual companies ; Indicates the first A Level 0 information data matrix for each enterprise; Indicates the first The calculation factors required to generate Level 1 information data for each enterprise.
[0082] In this embodiment, the enterprise information dataset is denoted as , represented as:
[0083]
[0084] As can be seen from the above embodiments, by quantifying the basic information of enterprises into calculable values, calculating a two-dimensional matrix of enterprise information based on the quantified information, and generating an enterprise information dataset, the credit assessment model's ability to process complex data is improved.
[0085] In one embodiment of this application, step S202 includes:
[0086] S2021: Label the enterprise information dataset to generate an enterprise credit information labeled dataset.
[0087] Specifically, let Execute the first Each enterprise credit information tag is labeled, and the enterprise credit information tag dataset is denoted as [database name missing]. .
[0088] For example, Indicates the first Whether the corporate loan was successful. If a company successfully obtains a loan, the label is defined as 1; if the company fails to obtain a loan, the label is defined as 0.
[0089] Indicates the first Loan amount obtained by individual enterprises , .
[0090] Indicates the first Loan time for individual enterprises , .
[0091] Indicates the first Number of times a company successfully obtains credit and repays it on time , .
[0092] Indicates the first Other markers for individual companies , .
[0093] No. Individual enterprise tag vector Represented as:
[0094]
[0095] Repeatedly mark corporate credit information until Generate a labeled dataset of corporate credit information. , Represented as:
[0096]
[0097] S2022: Score the enterprise credit information tagging dataset according to the preset scoring rules and generate an enterprise credit information tagging scoring dataset.
[0098] Specifically, the credit information of enterprises is scored, as follows:
[0099]
[0100] In the formula, This indicates a score for the marking of corporate credit information; The values of the hyperparameters are set, where, and , and , and , and .
[0101] In this embodiment, the enterprise credit information is labeled and scored into a dataset. Represented as:
[0102]
[0103] S2023: Perform data preprocessing on the enterprise credit information labeling and scoring dataset and the enterprise information dataset to generate model input data.
[0104] Specifically, the corporate credit information labeling and scoring dataset and enterprise information datasets Data preprocessing, format conversion, and segmentation are performed to generate model input data.
[0105] As can be seen from the above embodiments, by labeling the enterprise information dataset, an enterprise credit information label dataset is generated. The enterprise credit information label dataset is scored according to the scoring rules. The model input data is generated based on the enterprise credit information label scoring dataset and the enterprise information dataset to evaluate the enterprise credit, thereby improving the accuracy of the credit evaluation model in evaluating enterprise credit.
[0106] Figure 3 Flowchart of the credit assessment method provided for this application Figure 2 ,like Figure 3 As shown, in one embodiment of this application, after step S202, the following step is further included:
[0107] S301: Initialize the connection weights of the credit assessment model.
[0108] In this embodiment, the connection weights of the initialized credit assessment model are denoted as: , , , .
[0109] S302: The neurons in the credit assessment model are screened by the temporary withdrawal method to obtain the screened credit assessment model.
[0110] Specifically, some neurons in the hidden layer of the credit assessment model are temporarily discarded, and the selected neurons are used for calculation.
[0111] Figure 4 This is a schematic diagram of the backpropagation algorithm provided in an embodiment of this application.
[0112] Figure 5 This is a schematic diagram of the temporary withdrawal method for screening neurons provided in an embodiment of this application.
[0113] like Figure 4 and Figure 5 As shown, dashed lines represent some neurons in the hidden layer that are discarded, to avoid too many neurons participating in the calculation and to prevent overfitting of the output data.
[0114] S303: Train the selected credit assessment model based on the connection weights and the preset number of iterations, and generate the trained credit assessment model.
[0115] Specifically, the output value of each neuron is calculated, the loss function is calculated based on the credit assessment prediction value, the credit assessment model is trained through a preset number of iterations, and the trained credit assessment model is generated.
[0116] As can be seen from the above embodiments, by initializing the connection weights of the evaluation model, using the temporary withdrawal method to screen the neurons in the credit evaluation model, the generalization performance of the credit evaluation model is improved, and the recognition accuracy of the credit evaluation model is improved by training the credit evaluation model according to the connection weights and the number of iterations.
[0117] In one embodiment of this application, step S303 includes:
[0118] S3031: Calculate the output value of each neuron in the filtered credit assessment model based on the connection weights.
[0119] Specifically, the input data to be input into the model is fed into the input layer of the credit assessment model. By connecting weights and activation functions, the output of each neuron is calculated layer by layer, and the calculated data is output through the output layer.
[0120] In this embodiment, the calculation formula from the input layer to the hidden layer is:
[0121]
[0122] In the formula, This represents the weighted input to the hidden layer; This represents the output of the hidden layer; and Indicates connection weight; Indicates the activation function; This represents the input data for the model.
[0123] In this embodiment, the calculation formula from the hidden layer to the output layer is:
[0124]
[0125] In the formula, This represents the weighted input of the output layer; This represents the output of the hidden layer; and Indicates connection weight; This represents the activation function.
[0126] In this embodiment, the data output by the output layer is:
[0127]
[0128] In the formula, , Activation function for corporate credit information prediction scoring dataset and The calculation formula is:
[0129]
[0130] In the formula, This is a hyperparameter.
[0131] S3032: Calculate the loss function based on the output value of each neuron and the credit assessment prediction value.
[0132] In this embodiment, the formula for calculating the loss function is:
[0133]
[0134]
[0135] In the formula, Represents the loss function; and For hyperparameters; For corporate credit information prediction and scoring datasets, This is a dataset for accurate credit scoring of enterprises. A loss function representing whether a company's credit application was successful; A loss function representing the loss on the loan amount obtained by the company; The loss function representing the duration of the loan obtained by the company; This represents the number of times a company successfully obtains credit and repays it on time; A loss function indicating whether other markers for the enterprise were successful.
[0136] S3033: Train the credit assessment model based on the loss function and the preset number of iterations, and generate the trained credit assessment model.
[0137] Specifically, the learning rate of the model is adjusted according to the gradient of the loss function with respect to the network parameters and the preset number of iterations, the model is trained iteratively, and a trained credit assessment model is generated.
[0138] As can be seen from the above embodiments, by obtaining the output value of each neuron, comparing the output value with the credit assessment prediction value, generating a loss function, and training the credit assessment model based on the loss function, the problem of insufficient scoring discrimination caused by the single objective of the traditional model is solved, and the accuracy of the assessment is improved.
[0139] Figure 6 Flowchart of the credit assessment method provided for this application Figure 3 ,like Figure 6 As shown, in one embodiment of this application, after step S3032, the method further includes:
[0140] S401: Calculate the gradient of each connection weight in the credit assessment model based on the loss function.
[0141] Specifically, the gradient of the loss function is calculated by using the backpropagation algorithm, starting from the output layer and working backward through each layer of the network until the input layer:
[0142]
[0143] In the formula, For the loss function with respect to the variable gradient, For the loss function with respect to the variable The gradient.
[0144] In this embodiment, the error term of the hidden unit, i.e., the formula for calculating the gradient or partial derivative of the loss function with respect to the hidden unit, is as follows:
[0145]
[0146] In the formula, For the loss function with respect to the variable gradient, For the loss function with respect to the variable The gradient.
[0147] S402: Determine the learning rate for updating each connection weight based on the number of iterations of the credit assessment model.
[0148] In this embodiment, the formula for calculating the learning rate is:
[0149]
[0150] In the formula, Indicates the learning rate; and Indicates hyperparameters; Indicates the number of iterations. .
[0151] S403: Update each connection weight according to the gradient of each connection weight and the learning rate of each connection weight, and generate optimized connection weights.
[0152] In this embodiment, the formula for updating the output unit parameters is:
[0153]
[0154]
[0155] In this embodiment, the formula for updating the hidden unit parameters is:
[0156]
[0157] As can be seen from the above embodiments, by calculating the gradient value of each connection weight, dynamically adjusting the learning rate of the connection weights according to the number of iterations, and updating the connection weights according to the learning rate and gradient, the robustness of the model under complex data distribution is improved.
[0158] Figure 7 A schematic diagram of the credit assessment device provided in this application is shown below. Figure 7 As shown, the credit assessment device 70 provided in this embodiment includes: an acquisition module 701, a marking module 702, and a generation module 703.
[0159] The acquisition module 701 is used to acquire basic enterprise information and perform numerical quantification processing on the basic enterprise information to generate an enterprise information dataset.
[0160] The labeling module 702 is used to label the enterprise information dataset and generate model input data.
[0161] The generation module 703 is used to input the model input data into the credit assessment model and generate the credit assessment result. The credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
[0162] In one embodiment of this application, the acquisition module 701 includes:
[0163] The acquisition unit is used to acquire basic enterprise information, quantify the basic enterprise information, and generate quantitative information.
[0164] The calculation unit is used to calculate a two-dimensional matrix of enterprise information based on quantitative information.
[0165] The generation unit is used to generate an enterprise information dataset based on a two-dimensional matrix of enterprise information.
[0166] In one embodiment of this application, the marking module 702 includes:
[0167] The tagging unit is used to tag enterprise information datasets to generate enterprise credit information tagging datasets.
[0168] The scoring unit is used to score the enterprise credit information tagging dataset according to preset scoring rules, and generate an enterprise credit information tagging scoring dataset.
[0169] The data preprocessing unit is used to preprocess the enterprise credit information tagging and scoring dataset and the enterprise information dataset to generate model input data.
[0170] In one embodiment of this application, the marking module 702 further includes:
[0171] The initialization unit is used to initialize the connection weights of the credit assessment model.
[0172] The screening unit is used to screen neurons in the credit assessment model using the temporary withdrawal method to obtain the screened credit assessment model.
[0173] The training unit is used to train the selected credit assessment model based on the connection weights and a preset number of iterations, and to generate the trained credit assessment model.
[0174] In one embodiment of this application, the training unit includes:
[0175] The first computational subunit is used to calculate the output value of each neuron in the filtered credit assessment model based on the connection weights.
[0176] The second computational subunit is used to calculate the loss function based on the output value of each neuron and the credit assessment prediction value.
[0177] The training subunit is used to train the credit assessment model based on the loss function and the preset number of iterations, and generate the trained credit assessment model.
[0178] In one embodiment of this application, the training unit further includes:
[0179] The third computational subunit is used to calculate the gradient of each connection weight in the credit assessment model based on the loss function.
[0180] Determine the sub-units to determine the learning rate for updating each connection weight based on the number of iterations of the credit assessment model.
[0181] The update sub-unit is used to update each connection weight based on the gradient of each connection weight and the learning rate updated by each connection weight, thereby generating optimized connection weights.
[0182] The credit assessment device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.
[0183] Figure 8 A schematic diagram of the structure of the electronic device provided in this application. Figure 8 As shown, the electronic device 80 provided in this embodiment includes at least one processor 801 and a memory 802. Optionally, the electronic device 80 further includes a communication component 803. The processor 801, memory 802, and communication component 803 are connected via a bus.
[0184] In the specific implementation process, at least one processor 801 executes computer execution instructions stored in memory 802, causing at least one processor 801 to execute the above-mentioned credit assessment method.
[0185] The specific implementation process of processor 801 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.
[0186] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.
[0187] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.
[0188] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.
[0189] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the credit assessment method described above.
[0190] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the aforementioned credit assessment method.
[0191] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.
[0192] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.
[0193] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.
[0194] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0195] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0196] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0197] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.
[0198] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.
Claims
1. A credit assessment method, characterized in that, include: Acquire basic enterprise information and perform numerical quantification on the basic enterprise information to generate an enterprise information dataset; The enterprise information dataset is labeled to generate model input data; The input data of the model is input into the credit assessment model to generate a credit assessment result. The credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
2. The method according to claim 1, characterized in that, The process of acquiring basic enterprise information and performing numerical quantification on the basic enterprise information to generate an enterprise information dataset includes: Obtain basic enterprise information, quantify the basic enterprise information, and generate quantitative information; Calculate a two-dimensional matrix of enterprise information based on the quantitative information; Generate an enterprise information dataset based on the two-dimensional matrix of enterprise information.
3. The method according to claim 1, characterized in that, The step of labeling the enterprise information dataset to generate model input data includes: The enterprise information dataset is labeled to generate an enterprise credit information labeled dataset; The enterprise credit information tagging dataset is scored according to preset scoring rules to generate an enterprise credit information tagging scoring dataset. Data preprocessing is performed on the enterprise credit information tagging and scoring dataset and the enterprise information dataset to generate model input data.
4. The method according to claim 1, characterized in that, After labeling the enterprise information dataset and generating model input data, the process further includes: Initialize the connection weights of the credit assessment model; The credit assessment model is obtained by filtering neurons in the credit assessment model using the temporary withdrawal method; The selected credit assessment model is trained based on the connection weights and the preset number of iterations, thus generating the trained credit assessment model.
5. The method according to claim 4, characterized in that, The step of training the selected credit assessment model based on the connection weights and a preset number of iterations to generate the trained credit assessment model includes: The output value of each neuron in the filtered credit assessment model is calculated based on the connection weights. The loss function is calculated based on the output value of each neuron and the credit assessment prediction value. The credit assessment model is trained based on the loss function and the preset number of iterations, and the trained credit assessment model is generated.
6. The method according to claim 4, characterized in that, After calculating the loss function based on the output values of each neuron and the credit assessment prediction value, the method further includes: Calculate the gradient of each connection weight in the credit assessment model based on the loss function; The learning rate for updating each connection weight is determined based on the number of iterations of the credit assessment model. The connection weights are updated based on their gradients and the learning rates, resulting in optimized connection weights.
7. A credit assessment device, characterized in that, include: The acquisition module is used to acquire basic enterprise information and perform numerical quantification processing on the basic enterprise information to generate an enterprise information dataset. The tagging module is used to tag the enterprise information dataset and generate model input data; The generation module is used to input the model input data into the credit assessment model and generate credit assessment results, wherein the credit assessment model is a model trained using the backpropagation algorithm and adaptive learning rate.
8. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the credit assessment method as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the credit assessment method as described in any one of claims 1 to 6.
10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the credit assessment method according to any one of claims 1 to 6.