Credit risk assessment method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a credit risk assessment method and device, equipment and a storage medium, and the method comprises the steps: obtaining the pre-loan exhaustion data of a target business person for a target enterprise, and the in-loan data corresponding to the target enterprise; determining first enterprise information corresponding to the target enterprise based on the pre-loan exhaustion data, and determining second enterprise information corresponding to the target enterprise based on the in-loan data; and performing cross verification based on the first enterprise information and the second enterprise information to obtain a target risk assessment result of the target enterprise. According to the technical scheme of the embodiment of the invention, the problem of single analysis data dimension during credit risk assessment of small and micro enterprises in the prior art is solved, the pre-loan exhaustion data and the in-loan data of the target enterprise can be fused and verified, and the accuracy of credit risk assessment is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of financial data analysis technology, and in particular to a credit risk assessment method, apparatus, device and storage medium. Background Technology

[0002] In existing micro and small enterprise (MSE) lending practices, financial institutions typically assign loan officers to conduct on-site visits to businesses before granting loans. This involves understanding the business premises, production equipment, and employee situation, and interviewing business owners to understand the company's business strategies and financial development. However, this due diligence approach has several problems. The data collected is often limited, typically only including basic structured data or relying on photographs. Furthermore, the collected data is not efficiently integrated into the loan application process, failing to fully leverage its value. Compared to large and medium-sized enterprises, MSEs, due to their smaller size, less standardized management, and unclear finances, may submit incomplete or falsified data, posing challenges to financial institutions in risk control. Summary of the Invention

[0003] This invention provides a credit risk assessment method, apparatus, device, and storage medium that can integrate and verify pre-loan due diligence data and mid-loan data of a target enterprise, thereby improving the accuracy of credit risk assessment.

[0004] In a first aspect, embodiments of the present invention provide a credit risk assessment method, the method comprising:

[0005] Obtain pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and determine the second enterprise information corresponding to the target enterprise based on the loan data; perform cross-validation based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

[0006] Secondly, embodiments of the present invention provide a credit risk assessment device, the device comprising:

[0007] The data acquisition module is used to acquire pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; the enterprise information extraction module is used to determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and to determine the second enterprise information corresponding to the target enterprise based on the loan data; the enterprise information verification module is used to perform cross-verification based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

[0008] Thirdly, embodiments of the present invention provide a computer device, the computer device comprising:

[0009] One or more processors;

[0010] Memory, used to store one or more programs;

[0011] When the one or more programs are executed by the one or more processors, the one or more processors implement the credit risk assessment method described in any embodiment.

[0012] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the credit risk assessment method described in any embodiment.

[0013] The technical solution provided by this invention involves acquiring pre-loan due diligence data from target business personnel for a target enterprise, as well as corresponding loan data for the target enterprise; determining first enterprise information based on the pre-loan due diligence data, and determining second enterprise information based on the loan data; and cross-validating the first and second enterprise information to obtain the target enterprise's target risk assessment result. This invention solves the problem of limited data dimensions in existing technologies for credit risk assessment of micro and small enterprises, by integrating and validating pre-loan due diligence data and loan data for the target enterprise, thereby improving the accuracy of credit risk assessment. Attached Figure Description

[0014] Figure 1 This is a flowchart of a credit risk assessment method provided by an embodiment of the present invention;

[0015] Figure 2 This is a flowchart of another credit risk assessment method provided by an embodiment of the present invention;

[0016] Figure 3 This is a flowchart illustrating a credit risk assessment process provided by an embodiment of the present invention;

[0017] Figure 4 This is a schematic diagram of the structure of a credit risk assessment device provided in an embodiment of the present invention;

[0018] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The acquisition, storage, use, and processing of data in the technical solutions of the embodiments of the present invention all comply with the relevant provisions of national laws and regulations.

[0020] Figure 1 This is a flowchart of a credit risk assessment method provided by an embodiment of the present invention. The embodiment of the present invention can be applied to scenarios of risk assessment for micro and small enterprises. The method can be executed by a credit risk assessment device, which can be implemented by software and / or hardware.

[0021] like Figure 1 As shown, the credit risk assessment method includes the following steps:

[0022] S110. Obtain the pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise.

[0023] The target enterprise can be a micro or small enterprise requiring credit risk assessment. The technical solution of this invention can analyze multiple dimensions of enterprise information to determine whether credit risk exists. The target business personnel can be staff members conducting credit due diligence on the target enterprise. Before issuing a loan, financial institutions conduct a comprehensive and in-depth investigation and assessment of the borrower's credit status, repayment ability, financial condition, operating conditions, and collateral value to fully understand the borrower's risk profile, provide a basis for loan decisions, reduce loan default risk, and ensure fund security. Simultaneously, borrowers should actively cooperate with the due diligence process and provide truthful and complete information to improve the success rate of loan approval.

[0024] Pre-loan due diligence data can be enterprise data obtained by target business personnel after conducting due diligence on target companies. Specifically, target business personnel can use survey pads to collect relevant data, which is divided into four categories: photos, videos, GPS tracking, and audio. For example: In photo data collection, target business personnel will take photos of the business owner and company documents for archiving. Photos of the business owner's face are used for consistency verification by loan due diligence personnel later. Document photos include tax invoices, property certificates, and business licenses obtained during the loan process, used for cross-verification with the business owner's business registration data. Second, in video data collection, target business personnel will photograph the company's real estate and employee work environments. When entering companies with factories, construction sites, or land, video recordings of the real estate will be recorded. When investigating operating companies with multiple employees, such as technology companies, videos of personnel and their work environment will be taken. Third, in GPS data collection, the physical location of target business personnel will be continuously collected and stored while they are using the PAD for pre-loan marketing. Fourth, during the voice collection process, the target business personnel will ask the business owner to say a specified sentence during the pre-loan investigation to obtain the business owner's voice information.

[0025] Furthermore, loan data can be enterprise data provided by the target company during the loan application process. Specifically, loan data includes at least one of the following: credit data, business registration data, tax data, internal bank data, and invoice data. The system can acquire customer data from different dimensions after customer authorization for customer identity verification, access determination, and credit assessment; it can also capture data involved in the loan application process, including credit data collection, business registration data collection, internal bank data collection, invoice data collection, and tax data collection. For example, credit data collection includes personal / corporate credit reports, credit scores, loan records, and overdue records, used to assess a customer's creditworthiness; business registration data collection includes company registration information, shareholder information, business scope, and operating status, used to verify a company's qualifications and operations; internal bank data collection includes data on a customer's deposits, loans, and wealth management transactions within the bank, used to analyze a customer's financial situation and risk appetite; invoice data collection includes information on value-added tax invoices issued and received by the company, used to verify the company's operating income and the authenticity of transactions; and tax data collection includes tax returns and tax payment certificates, used to verify the company's tax payment status and operating income.

[0026] S120. Based on the pre-loan due diligence data, determine the first enterprise information corresponding to the target enterprise, and based on the loan data, determine the second enterprise information corresponding to the target enterprise.

[0027] The first type of enterprise information can be obtained based on due diligence data. Specifically, the data types in the pre-loan due diligence data can be identified, and then corresponding data feature extraction steps can be performed for different due diligence data types to finally obtain the corresponding enterprise information. The second type of enterprise information can be extracted from the loan data dimension. For example, the second type of enterprise information may include: the business owner's voice recording, the business owner's facial recognition, the business's credit information, registered location, the business's monthly operating status, the business owner's credit information, and the number of taxpayers, etc. Specifically, key data features can be extracted from the loan data to obtain the second type of enterprise information.

[0028] S130. Based on the first enterprise information and the second enterprise information, cross-validation is performed to obtain the target risk assessment result of the target enterprise.

[0029] The target risk assessment result can be the final result of a credit risk assessment of the target company. Specifically, consistency and similarity analysis can be performed on the information of the first company and the information of the second company, and the credit risk can be determined based on the analysis results, thereby determining the target risk assessment result.

[0030] Optionally, after obtaining the pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise, the pre-loan due diligence data and the loan data can be normalized and standardized, and enterprise information can be extracted based on the pre-loan due diligence data and the loan data after data processing.

[0031] The technical solution provided by this invention obtains pre-loan due diligence data from target business personnel for a target enterprise, as well as corresponding loan data for the target enterprise; determines first enterprise information based on the pre-loan due diligence data, and determines second enterprise information based on the loan data; and cross-validates the first and second enterprise information to obtain the target enterprise's target risk assessment result. This invention solves the problem of limited data dimensions in existing technologies for credit risk assessment of micro and small enterprises, allowing for the fusion and verification of pre-loan due diligence data and loan data for the target enterprise, thereby improving the accuracy of credit risk assessment.

[0032] Figure 2This is a flowchart of another credit risk assessment method provided by an embodiment of the present invention. The embodiments of the present invention can be applied to scenarios of risk assessment for micro and small enterprises. Based on the above embodiments, this embodiment further explains how to determine the first enterprise information corresponding to the target enterprise based on pre-loan due diligence data, and how to determine the second enterprise information corresponding to the target enterprise based on mid-loan data; and how to perform cross-validation based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise. The device can be implemented by software and / or hardware and integrated into a computer device with application development capabilities.

[0033] like Figure 2 As shown, the credit risk assessment method includes the following steps:

[0034] S210. Obtain the pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise.

[0035] S220. Identify the various due diligence data types in the pre-loan due diligence data, and extract data features according to the data processing methods corresponding to the due diligence data types to obtain the first enterprise information.

[0036] The due diligence data types include at least one of the following: due diligence image data, due diligence video data, due diligence trajectory data, and due diligence voice data. The first enterprise information can be enterprise information obtained based on the due diligence data. Specifically, the data types in the pre-loan due diligence data can be identified, and then corresponding data feature extraction steps can be performed for different due diligence data types to finally obtain the corresponding enterprise information.

[0037] Specifically, feature extraction can be performed on due diligence image data to obtain enterprise text features and enterprise owner image features; feature extraction can be performed on due diligence video data to obtain enterprise real estate features and enterprise employee features; due diligence trajectory data can obtain due diligence location features corresponding to target business personnel; and due diligence voice data can obtain enterprise owner voice features.

[0038] First, when processing due diligence image data, the image processing module captures and processes the business owner's photo. The backend system automatically captures the business owner's photo, performs image quality assessment, and preprocesses the image, including face detection, alignment, and illumination normalization. After image preprocessing, the backend uses the deep learning model FaceNet to extract the business owner's facial features. The extracted features are then dimensionality-reduced and standardized to generate fixed-dimensional feature vectors. These feature vectors are then normalized to ensure the stability of the vector distribution. After feature vector generation, the extracted facial features are mapped to an embedded vector space. Metric learning techniques are used to optimize the vector space, improve feature discriminative power, and generate standardized embedded vectors to support subsequent comparison and analysis. Finally, the generated embedded vectors are stored in an encrypted database.

[0039] In image capture, the target business personnel input basic information such as business licenses and invoices through the front-end system. The back-end system automatically captures relevant images, including scanned copies of business licenses and invoices. After image acquisition, preprocessing is performed, including noise reduction, rotation correction, and resolution optimization. In OCR text extraction, the deep learning OCR model PaddleOCR is used to extract text information from the images. The extracted text undergoes structured processing to generate standardized text data. Finally, the OCR results are verified and corrected to ensure data accuracy. In NLP semantic recognition, NLP technology is used to perform semantic analysis on the extracted text, including entity recognition, keyword extraction, and semantic classification. Key information such as company name, registration number, invoice amount, and tax ID is identified. The extracted information is matched and integrated with the customer's input data to generate a complete customer data record. Finally, the extracted text data, semantic analysis results, and customer metrics are stored in the database.

[0040] Secondly, when processing due diligence video data, the system focuses on processing video information of business operations recorded by target personnel during loan investigations. This includes detecting people in the video and removing duplicates. For person detection and extraction, an improved YOLOv5 algorithm is used for real-time person detection, extracting bounding boxes and feature vectors. For person deduplication, the DeepSORT algorithm is used for cross-frame person matching, calculating similarity based on appearance features and motion information, and setting dynamic thresholds to achieve accurate deduplication. After person extraction and deduplication are completed, the system stores the corresponding number of people for subsequent loan verification.

[0041] Third, when processing the due diligence voice data, the recorded voice of the business owner undergoes noise reduction processing, employing an adaptive filter to eliminate environmental noise. After noise reduction, a Gammatone filter bank is used to extract the time-frequency features of the voice signal. After feature extraction, the extracted features are normalized to eliminate dimensional differences between features, and Principal Component Analysis (PCA) is used for feature dimensionality reduction to remove redundant features and improve computational efficiency. Finally, the optimized feature vectors are stored in an encrypted database to ensure data security.

[0042] Fourth, when processing due diligence trajectory data, the GPS information of the target business personnel during the pre-loan investigation is processed and analyzed. First, location data is captured; the backend system collects the location information of the target business personnel in real time through the GPS module of the PAD device. The collected location information is preprocessed, including noise reduction, interpolation, and coordinate transformation. Then, the collected location information is used to generate the investigation trajectory of the target business personnel. The trajectory data is then optimized using the Douglas-Peucker trajectory compression algorithm. Finally, the trajectory data is segmented to identify the start point, end point, and key nodes of the investigation task. The start point, end point, and key nodes of the investigation task are then stored in the database.

[0043] This solution proposes that loan officers use dedicated devices such as tablets to input company information on-site, collect images, videos, and audio, and extract information such as GPS data. Secondly, the collected multimodal data is categorized and processed, and algorithms such as OCR, YOLO, and Gammatone are used to extract key company information. When customers apply for loans online, GPS tracking is used to determine whether the target business personnel have actually visited the company, accurately confirming whether they have indeed conducted on-site due diligence and improving the accuracy of credit risk assessment.

[0044] S230. Extract enterprise features from the loan data according to the preset extraction dimensions, and use the extracted feature information as the second enterprise information.

[0045] The second type of enterprise information can be enterprise information extracted from the loan data dimension. For example, the second type of enterprise information may include: the voice of the enterprise owner, the face of the enterprise owner, the enterprise's credit information, the registered location, the enterprise's monthly operating status, the credit information of the enterprise owner, and the number of taxpayers, etc.

[0046] S240. Perform a consistency check on the first enterprise information and the second enterprise information to obtain the first risk assessment result.

[0047] The first risk assessment result can be the result of a credit risk assessment from the perspective of consistency of enterprise information. Specifically, according to preset verification rules, the data in the first enterprise information and the corresponding data in the second enterprise information can be checked for consistency. A consistency score is determined based on the number of data items that meet the requirements, and the determined consistency score is used as the first risk assessment result.

[0048] Optionally, consistency analysis can be performed from four verification dimensions: entity information, operating entity, financial relationship, biometrics, and device behavior. For example, entity information verification can be performed based on first and second enterprise information to obtain entity verification results; operating entity authenticity verification can be performed based on first and second enterprise information to obtain operating entity verification results; financial data relationship verification can be performed based on first and second enterprise information to obtain financial data verification results; device and trajectory detection can be performed based on first and second enterprise information to obtain device and trajectory verification results; and the first risk assessment result is determined based on the entity verification results, operating entity verification results, financial data verification results, and device and trajectory verification results. Specific rules are shown in the table below:

[0049] Enterprise Entity Information Verification Form

[0050]

[0051]

[0052] Verification of the authenticity of the business entity

[0053]

[0054] Financial data correlation verification

[0055]

[0056] Equipment and Behavior Anomaly Detection

[0057]

[0058]

[0059] S250. Input the first enterprise information and the second enterprise information into the dual-tower encoder to obtain the second risk assessment result.

[0060] The second risk assessment result can be the result of credit risk assessment from the perspective of enterprise information consistency. Specifically, the dual-tower encoder includes: a first encoder, a second encoder, and a dual-tower analysis module. The first enterprise information can be input into the first encoder to obtain the first enterprise representation vector; the second enterprise information can be input into the second encoder to obtain the second enterprise representation vector; and the first and second enterprise representation vectors can be input into the dual-tower analysis module to output the second risk assessment result.

[0061] First Encoder (Pre-Loan Data Encoder): Image Feature Encoding: An improved ResNet-50 is used as the backbone network. Inputs include photos of the company premises (224×224RGB) and workstation camera footage. A Spatial Attention Module is introduced to enhance the focusing ability on key areas such as business licenses and inventory labels. Voiceprint Feature Encoding: A pre-trained Wav2Vec 2.0 model is used to extract 128-dimensional MFCC features from the company's main voiceprint. After capturing temporal dynamic features via a bidirectional LSTM, a 256-dimensional embedding vector is output. Spatiotemporal Trajectory Encoding: The GPS trajectory data of the target business personnel is transformed into a spatiotemporal grid sequence (1km×1km grid, 5-minute time window). A Transformer architecture is used to capture the spatiotemporal correlation of the movement path. Multimodal Fusion Layer: The weights of image, voiceprint, and trajectory features are dynamically allocated through a gated cross-attention mechanism, outputting a 512-dimensional comprehensive representation vector of the first encoder.

[0062] The second encoder (loan data encoder): Structured data encoding: For tax invoice amounts (time series), business registration information (categorical), enterprise credit information (numerical), and enterprise tax payments (time series), a field-specific encoding strategy is adopted: Numerical fields (such as average monthly invoice amount) are standardized and then input into a fully connected layer; categorical fields (such as industry codes) are converted into 32-dimensional vectors through an embedding layer; time series fields (such as tax records for the past 12 months) have periodicity and trend features extracted by a TCN (Temporal Convolutional Network). All encoding results are concatenated and then input into an attention layer to capture the dependencies between fields, outputting a 512-dimensional representation vector of the second encoder.

[0063] The dual-tower analysis module calculates the similarity between the 512-dimensional representation vectors output by the first and second encoders, thereby indicating the probability that pre-loan data and mid-loan data come from the same customer. This scheme uses a cosine similarity function to calculate the similarity, as follows:

[0064] The vector similarity of the outputs of the two towers is calculated using cosine similarity:

[0065]

[0066] Where VA represents the representation vector of the first encoder and VB represents the representation vector of the second encoder.

[0067] Comparison loss function:

[0068] An improved Triplet Margin Loss is employed to optimize the feature space through dynamic hard sample mining:

[0069]

[0070] Where Spos represents the similarity of matching sample pairs, Sneg represents the similarity of negative sample pairs, and the marginal value α is dynamically adjusted according to the difficulty of samples within the batch.

[0071] S260. Determine the target risk assessment result based on the first risk assessment result and the second risk assessment result.

[0072] The target risk assessment result can be the final result of the credit risk assessment of the target enterprise. Specifically, the target risk assessment result can be determined by combining the first risk assessment result and the second risk assessment result. For example, if the enterprise entity information verification and equipment and behavior anomaly detection verification fail due to rule failures, it is marked as high risk and the loan is directly rejected; if the operating entity authenticity verification and operating entity authenticity verification fail due to rule verification failures, it is marked as medium risk and the judgment is made based on the consistency score predicted by the model. If the score is less than a certain threshold, it is marked as high risk and the loan is directly rejected; if it is higher than a certain threshold, it is marked as medium risk and the judgment is made manually. If all cross-validation rules pass, the judgment is made based on the consistency score predicted by the model. If the score is higher than a certain threshold, it is marked as low risk; if it is lower than a certain threshold, it is marked as medium risk and the judgment is made manually.

[0073] For example, to better understand the technical solution provided by this invention, a specific embodiment is described below: The evaluation and verification stage of this patent aims to construct a cross-validation system for pre-loan due diligence data and mid-loan data through multimodal data fusion and intelligent analysis technology, solving the problems of low efficiency, strong subjectivity, and information silos in traditional manual verification. This stage is driven by a dual-engine approach of "rule engine + deep learning model," using technologies such as automated comparison, semantic understanding, and spatiotemporal correlation analysis to achieve multi-dimensional verification of the authenticity, consistency, and rationality of enterprise information. Figure 3As shown, the main steps include: 1. Data input: pre-loan multimodal data (images, videos, audio, GPS) and mid-loan data (credit, tax, business registration) are input in parallel; 2. Preprocessing: data cleaning, feature extraction, and spatiotemporal alignment; 3. Cross-validation: the rule engine performs strong rule comparison, and the knowledge graph mines the associated risks; 4. Model prediction: the dual-tower model calculates the data similarity; Result generation: outputs the risk level (low risk / medium risk / high risk).

[0074] The technical solution provided by this invention involves acquiring pre-loan due diligence data from target business personnel for a target enterprise, as well as corresponding loan data for the target enterprise; identifying multiple due diligence data types in the pre-loan due diligence data, and extracting data features according to the data processing methods corresponding to each due diligence data type to obtain first enterprise information; extracting enterprise features from the loan data according to preset extraction dimensions, and using the extracted feature information as second enterprise information; verifying the consistency between the first and second enterprise information to obtain a first risk assessment result; inputting the first and second enterprise information into a dual-tower encoder to obtain a second risk assessment result; and determining the target risk assessment result based on the first and second risk assessment results. This technical solution solves the problem of a single data dimension in the prior art when conducting credit risk assessment for micro and small enterprises, and can integrate and verify the pre-loan due diligence data and loan data of the target enterprise, thereby improving the accuracy of credit risk assessment.

[0075] Figure 4 This is a schematic diagram of the structure of a credit risk assessment device provided in an embodiment of the present invention. The embodiment of the present invention can be applied to scenarios for risk assessment of micro and small enterprises. The device can be implemented by software and / or hardware and integrated into a computer device with application development capabilities.

[0076] like Figure 4 As shown, the credit risk assessment device includes: a data acquisition module 310, an enterprise information extraction module 320, and an enterprise information verification module 330.

[0077] The data acquisition module 310 is used to acquire pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; the enterprise information extraction module 320 is used to determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and to determine the second enterprise information corresponding to the target enterprise based on the loan data; the enterprise information verification module 330 is used to perform cross-verification based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

[0078] The technical solution provided by this invention involves acquiring pre-loan due diligence data from target business personnel for a target enterprise, as well as corresponding loan data for the target enterprise; determining first enterprise information based on the pre-loan due diligence data, and determining second enterprise information based on the loan data; and cross-validating the first and second enterprise information to obtain the target enterprise's target risk assessment result. This invention solves the problem of limited data dimensions in existing technologies for credit risk assessment of micro and small enterprises, by integrating and validating pre-loan due diligence data and loan data for the target enterprise, thereby improving the accuracy of credit risk assessment.

[0079] In one optional implementation, the enterprise information extraction module 320 is specifically used to: identify multiple due diligence data types in the pre-loan due diligence data, extract data features according to the data processing methods corresponding to the due diligence data types, and obtain the first enterprise information; wherein, the due diligence data types include at least one of due diligence image data, due diligence video data, due diligence trajectory data, and due diligence voice data; extract enterprise features from the loan data according to a preset extraction dimension, and use the extracted feature information as the second enterprise information; wherein, the loan data includes at least one of credit data, business registration data, tax data, bank data, and invoice data.

[0080] In one optional implementation, the enterprise information extraction module 320 includes: a due diligence data feature extraction unit, used to: extract features from the due diligence image data to obtain enterprise text features and enterprise owner image features; extract features from the due diligence video data to obtain enterprise real estate features and enterprise employee features; extract due diligence location features corresponding to the target business personnel from the due diligence trajectory data; and extract enterprise owner voice features from the due diligence voice data.

[0081] In one optional implementation, the enterprise information verification module 330 is specifically used to: perform consistency verification on the first enterprise information and the second enterprise information to obtain a first risk assessment result; input the first enterprise information and the second enterprise information into a dual-tower encoder to obtain a second risk assessment result; and determine the target risk assessment result based on the first risk assessment result and the second risk assessment result.

[0082] In one optional implementation, the enterprise information verification module 330 includes: a data consistency verification unit, configured to: verify enterprise entity information based on the first enterprise information and the second enterprise information to obtain an enterprise entity verification result; verify the authenticity of the operating entity based on the first enterprise information and the second enterprise information to obtain an operating entity verification result; verify the correlation of financial data based on the first enterprise information and the second enterprise information to obtain a financial data verification result; detect equipment and trajectory based on the first enterprise information and the second enterprise information to obtain an equipment and trajectory verification result; and determine the first risk assessment result based on the enterprise entity verification result, the operating entity verification result, the financial data verification result, and the equipment and trajectory verification result.

[0083] In one optional implementation, the enterprise information verification module 330 includes: a data similarity verification unit, configured to: input the first enterprise information into the first encoder to obtain a first enterprise representation vector; input the second enterprise information into the second encoder to obtain a second enterprise representation vector; and input the first enterprise representation vector and the second enterprise representation vector into the dual-tower analysis module to obtain the second risk assessment result.

[0084] In one optional implementation, the credit risk assessment device further includes a data preprocessing module, used to: normalize and standardize the pre-loan due diligence data and the mid-loan data, and extract enterprise information based on the pre-loan due diligence data and the mid-loan data after data processing.

[0085] The credit risk assessment device provided in the embodiments of the present invention can execute the credit risk assessment method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0086] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Figure 5 A block diagram of an exemplary computer device 12 suitable for implementing embodiments of the present invention is shown. Figure 5 The computer device 12 shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of the present invention. The computer device 12 can be any terminal device with computing capabilities and can be configured within a credit risk assessment device.

[0087] like Figure 5 As shown, the computer device 12 is represented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and a bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0088] Bus 18 can be one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0089] Computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0090] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (…). Figure 5 Not shown; usually referred to as a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0091] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0092] Computer device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with the computer device 12, and / or with any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through input / output (I / O) interface 22. Furthermore, computer device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) through network adapter 20. Figure 5 As shown, network adapter 20 communicates with other modules of computer device 12 via bus 18. It should be understood that, although... Figure 5 As not shown, it can be used in conjunction with computer device 12 with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0093] Processing unit 16 executes various functional applications and data processing by running programs stored in system memory 28, such as implementing the credit risk assessment method provided in this embodiment of the invention, which includes:

[0094] Obtain pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and determine the second enterprise information corresponding to the target enterprise based on the loan data; perform cross-validation based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

[0095] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the credit risk assessment method as provided in any embodiment of the present invention, including:

[0096] Obtain pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and determine the second enterprise information corresponding to the target enterprise based on the loan data; perform cross-validation based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

[0097] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0098] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0099] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0100] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as C, Java, Smalltalk, C++, C#, and Python, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0101] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computing device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0102] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A credit risk assessment method, characterized in that, include: Obtain the pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; Based on the pre-loan due diligence data, the first enterprise information corresponding to the target enterprise is determined, and based on the loan data, the second enterprise information corresponding to the target enterprise is determined; Based on cross-validation of the first enterprise information and the second enterprise information, the target risk assessment result of the target enterprise is obtained.

2. The method according to claim 1, characterized in that, The step of determining the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and determining the second enterprise information corresponding to the target enterprise based on the loan data, includes: Identify multiple due diligence data types in the pre-loan due diligence data, and extract data features according to the data processing methods corresponding to the due diligence data types to obtain the first enterprise information; wherein, the due diligence data types include at least one of due diligence image data, due diligence video data, due diligence trajectory data, and due diligence voice data; The loan data is processed according to a preset extraction dimension to extract enterprise features, and the extracted feature information is used as the second enterprise information; wherein, the loan data includes at least one of the following: credit data, business registration data, tax data, bank data, and invoice data.

3. The method according to claim 2, characterized in that, The step of extracting data features according to the data processing method corresponding to the due diligence data type includes: Feature extraction is performed on the due diligence image data to obtain enterprise text features and enterprise main image features; Feature extraction is performed on the due diligence video data to obtain the characteristics of the company's real estate and the characteristics of the company's employees; From the due diligence trajectory data, the due diligence location characteristics corresponding to the target business personnel are obtained; The voice characteristics of the business owner are obtained from the due diligence voice data.

4. The method according to claim 1, characterized in that, The step of cross-validating the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise includes: A consistency check is performed on the first enterprise information and the second enterprise information to obtain the first risk assessment result; The first enterprise information and the second enterprise information are input into the dual-tower encoder to obtain the second risk assessment result; The target risk assessment result is determined based on the first risk assessment result and the second risk assessment result.

5. The method according to claim 4, characterized in that, The step of performing consistency verification on the first enterprise information and the second enterprise information to obtain the first risk assessment result includes: Based on the first enterprise information and the second enterprise information, the enterprise entity information is verified to obtain the enterprise entity verification result; The authenticity of the business entity is verified based on the information of the first enterprise and the information of the second enterprise, and the verification result of the business entity is obtained. The correlation between financial data is verified based on the information of the first company and the information of the second company, and the verification results of the financial data are obtained. Based on the first enterprise information and the second enterprise information, device and trajectory detection is performed to obtain device and trajectory verification results; The first risk assessment result is determined based on the enterprise entity verification results, operating entity verification results, financial data verification results, and equipment and trajectory verification results.

6. The method according to claim 4, characterized in that, The dual-tower encoder includes: a first encoder, a second encoder, and a dual-tower analysis module. The step of inputting the first enterprise information and the second enterprise information into the dual-tower encoder to obtain a second risk assessment result includes: The first enterprise information is input into the first encoder to obtain the first enterprise representation vector; The second enterprise information is input into the second encoder to obtain the second enterprise representation vector; The first enterprise representation vector and the second enterprise representation vector are input into the dual-tower analysis module to obtain the second risk assessment result.

7. The method according to claim 1, characterized in that, After obtaining the pre-loan due diligence data of the target business personnel for the target enterprise, and the corresponding loan data of the target enterprise, the process also includes: The pre-loan due diligence data and the loan data are normalized and standardized, and enterprise information is extracted based on the processed pre-loan due diligence data and loan data.

8. A credit risk assessment device, characterized in that, The device includes: The data acquisition module is used to acquire the pre-loan due diligence data of the target business personnel for the target enterprise, as well as the corresponding loan data of the target enterprise; The enterprise information extraction module is used to determine the first enterprise information corresponding to the target enterprise based on the pre-loan due diligence data, and to determine the second enterprise information corresponding to the target enterprise based on the loan data. The enterprise information verification module is used to perform cross-verification based on the first enterprise information and the second enterprise information to obtain the target risk assessment result of the target enterprise.

9. A computer device, characterized in that, The computer device includes: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the credit risk assessment method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the credit risk assessment method as described in any one of claims 1-7.

Images