Credit anti-fraud blacklist area identification system and method based on grid division

Abstract

The application discloses a credit anti-fraud blacklist area identification system and method based on grid division, which comprises a data acquisition and preprocessing module for acquiring and preprocessing GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, and acquiring business data; a gridding division and coding module for performing spatial discretization and grid coding according to the preprocessed GPS coordinate data to obtain GPS grids, and according to the business data, counting a plurality of indexes in each GPS grid; a feature engineering construction module for determining a plurality of feature vectors based on the plurality of indexes and the preprocessed image features; and a black grid identification model training and execution module for inputting the plurality of feature vectors into a trained black grid identification model to obtain a black grid identification result. The application can improve the accuracy of identifying credit anti-fraud blacklist areas.

Description

Technical Field

[0001] This application relates to the field of credit anti-fraud technology, and in particular to a credit anti-fraud blacklist region identification system and method based on grid partitioning. Background Technology

[0002] In the process of digitalizing online lending, organized fraud exhibits a clear regional clustering characteristic. Loan applications processed centrally by intermediaries frequently occur within the same geographical area and in a short period of time, resulting in financial institutions incurring significantly higher losses in localized areas compared to randomly distributed scenarios. Traditional anti-fraud models mainly rely on static customer attributes and credit history data, lacking the ability to deeply analyze the spatiotemporal trajectory of application GPS locations, making it difficult to effectively identify such regionally clustered fraud patterns.

[0003] Although existing technologies have attempted to incorporate geographic location information into risk assessment systems, the following key deficiencies still exist: (1) Existing systems do not have modules for gridded density analysis and cluster modeling of GPS coordinates, making it impossible to quantify the intensity and boundaries of regional risks; (2) There is a lack of dynamic update and decay mechanisms for blacklisted areas, resulting in a delayed response to the migration behavior of fraud gangs; (3) It fails to achieve joint modeling of geographic area risks and individual credit risks of applicants, leading to the wrong rejection of high-quality customers or the failure to prevent high-risk applications.

[0004] In summary, existing technologies have relatively low accuracy in identifying blacklisted areas for credit fraud prevention. Summary of the Invention

[0005] This application aims to propose a grid-based system and method for identifying blacklisted areas in credit fraud prevention, which can improve the accuracy of identifying such areas.

[0006] In a first aspect, embodiments of this application provide a credit anti-fraud blacklist region identification system based on grid partitioning, the system comprising: The data acquisition and preprocessing module is used to acquire and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as acquired business data. The grid division and encoding module is communicatively connected to the data acquisition and preprocessing module. It is used to receive the preprocessed GPS coordinate data and the business data, perform spatial discretization and grid encoding on the preprocessed GPS coordinate data to obtain a GPS grid, and statistically analyze multiple indicators within each GPS grid based on the business data. The feature engineering construction module is communicatively connected to the data acquisition and preprocessing module and the gridding and encoding module, and is used to receive the multiple indicators and the preprocessed image features, and determine multiple feature vectors based on the multiple indicators and the preprocessed image features; The black grid recognition model training and execution module is communicatively connected to the feature engineering construction module. It is used to receive the multiple feature vectors, input the multiple feature vectors into the trained black grid recognition model, and obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

[0007] In some implementations, the multiple metrics include total number of credit application customers, number of credit approved customers, total credit limit, number of credit approved customers, number of overdue customers, number of credit applications per day, and the average number of credit applications per day within a GPS grid of the same size.

[0008] In some implementations, the feature engineering construction module includes: The credit density feature submodule is used to calculate the daily credit application density within each grid based on the daily credit application volume and the average daily credit application volume; to obtain the credit application volume of each adjacent time window and to calculate the credit volume growth rate of adjacent time windows based on the credit application volume of each adjacent time window; and to obtain abnormal patterns. The face background consistency feature analysis submodule is used to calculate the similarity between the background features corresponding to different applicant face recognition images within the same GPS grid based on the preprocessed image features; calculate the proportion of high similarity image pairs based on the similarity, wherein the high similarity image pairs are image pairs with a similarity greater than a first preset threshold; and also to obtain target background features from the preprocessed image features. The risk indicator feature statistics submodule is used to calculate multiple risk indicator features based on the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, and the number of overdue customers; it is also used to analyze the correlation between customers within the grid and other known fraudulent customers. The time series feature extraction submodule is used to identify abnormally concentrated applications during non-working hours through time series analysis models, and to calculate the credit volume change trend and periodic pattern between adjacent time periods.

[0009] In some implementations, the multiple risk indicators include the grid's historical first delinquency rate, credit approval rate, average credit limit, and delinquency rate.

[0010] In some implementations, the calculation of multiple risk indicator characteristics based on the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, and the number of overdue customers includes: The historical first delinquency rate of a grid is calculated based on the number of first-time delinquent customers and the total number of credit application customers within the grid. The credit approval rate is calculated based on the number of customers who have been granted credit and the total number of customers who have applied for credit. Calculate the average credit limit based on the total credit limit and the number of customers who have granted credit. The delinquency rate is calculated based on the number of overdue customers and the number of customers whose credit has been granted.

[0011] In some implementations, the trained black grid recognition model is trained using a training sample set with normal labels or black grid labels, including: The historical first-overage data within the acquired grid is compared with a second preset threshold. If the historical first-overage data reaches the second preset threshold, the grid is marked as a black grid. If the historical first-overage data is less than the second preset threshold but greater than a third preset threshold, the grid is marked as a normal grid or a black grid through manual verification. If the historical first-overage data is less than the third preset threshold, the grid is marked as a normal grid. Construct a training sample set using grids labeled with normal or black grids; The black grid recognition model is trained using the training sample set to obtain the trained black grid recognition model.

[0012] In some implementations, the system further includes a real-time risk warning and response module for making risk decisions and issuing warnings based on the black grid identification results.

[0013] Secondly, embodiments of this application also provide a method for identifying blacklisted regions for credit fraud prevention based on grid partitioning, applied to the aforementioned system for identifying blacklisted regions for credit fraud prevention based on grid partitioning, the method comprising: Collect and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as collected business data; Spatial discretization and grid encoding are performed on the preprocessed GPS coordinate data to obtain a GPS grid. Based on the business data, multiple indicators within each GPS grid are statistically analyzed. Based on the aforementioned multiple indicators and the preprocessed image features, multiple feature vectors are determined; The multiple feature vectors are input into the trained black grid recognition model to obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

[0014] Thirdly, embodiments of this application also provide an electronic device, including at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor, which, when executed by the at least one control processor, enable the at least one control processor to perform a grid-based credit anti-fraud blacklist region identification method as described above.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a grid-based credit anti-fraud blacklist region identification method as described above.

[0016] Compared with the prior art, this application has the following beneficial effects: The data acquisition and preprocessing module of this application is used to acquire and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as acquired business data. The gridding and encoding module is connected to the data acquisition and preprocessing module and is used to receive the preprocessed GPS coordinate data and business data. Based on the preprocessed GPS coordinate data, it performs spatial discretization and grid encoding to obtain GPS grids. Based on the business data, it statistically analyzes multiple indicators within each GPS grid. The feature engineering construction module is connected to the data acquisition and preprocessing module and the gridding and encoding module and is used to receive multiple indicators and preprocessed image features. Based on the multiple indicators and preprocessed image features, it determines multiple feature vectors. The black grid recognition model training and execution module is connected to the feature engineering construction module and is used to receive multiple feature vectors. It inputs the multiple feature vectors into the trained black grid recognition model to obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels. Thus, by dividing the system into GPS grids and then comprehensively considering GPS coordinate data and business data to statistically analyze multiple indicators within each GPS grid, and then determining multiple feature vectors based on these indicators and facial recognition images, and finally performing black grid identification based on these feature vectors, the accuracy of identifying credit fraud blacklist areas can be improved, achieving precise location and identification of high-fraud areas. The analysis results based on gridded features have clear business implications, and the output conclusions are clear and interpretable, facilitating understanding, verification, and decision support for risk control personnel, and promoting the rapid deployment and promotion of the system in business environments. Through automated and intelligent model recognition and interception mechanisms, the reliance on manual verification is significantly reduced, lowering risk control operating costs. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a schematic diagram of the structure of an embodiment of the credit anti-fraud blacklist region identification system based on grid partitioning provided in this application; Figure 2 This is a schematic diagram of a real-world application scenario in the best embodiment of the credit anti-fraud blacklist region identification system based on grid partitioning provided in this application; Figure 3 This is a schematic diagram showing the client results in the best embodiment of the grid-based credit anti-fraud blacklist region identification system provided in this application; Figure 4 This is a flowchart illustrating an embodiment of the credit anti-fraud blacklist region identification method based on grid partitioning provided in this application; Figure 5This is a schematic diagram of the structure of an embodiment of the electronic device provided in this application. Detailed Implementation

[0018] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0019] In the description of this application, the use of terms such as "first," "second," etc., is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0020] In the description of this application, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0021] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0022] First, let's analyze some of the terms used in this application: GPS gridding module: This refers to a hardware or logic module that discretizes a continuous geographic coordinate space into a series of regular or irregular grid cells according to a predetermined level of precision (such as several decimal places for latitude and longitude). This module converts spatial location data into structured data that can be used for statistical analysis and pattern recognition.

[0023] Black grid identification module: Specifically refers to the component used in this embodiment to identify GPS grids with extremely high risk levels. Its judgment logic typically includes, but is not limited to: customers within the grid have an abnormally high first-time delinquency rate, and exhibit high-risk characteristics such as batch credit granting and facial background consistency, which are highly correlated with intermediary fraud activities.

[0024] Historical First Delinquency Rate: Within a specified observation period, the proportion of customers in a customer group (such as customers within a specific grid) who experience a "first payment followed by delinquency" event out of the total number of customers in that group. This indicator is a key performance indicator for measuring initial credit risk.

[0025] Anomaly pattern detection: Within a GPS grid, monitor whether the number or density of credit applications within a unit of time (e.g., hourly, daily) is significantly higher than the historical normal level or the overall average level of the area.

[0026] The Face Background Consistency Feature Analysis Submodule is a visual analysis module used to detect whether the background environment (such as wall decorations, furniture, and lighting angles) of different applicants' face recognition verification images shows high similarity or regularity. This result can serve as an auxiliary basis for judging the existence of concentrated operation locations.

[0027] Feature Engineering Construction Module: In this embodiment, this specifically refers to the logical unit that constructs derived features from data such as original GPS coordinates, application time, and customer information to train and execute the risk grid identification model. For example, it generates "number of applications per unit area," "average first-time delinquency rate," and "background consistency index" for each grid.

[0028] To address the issue of low accuracy in identifying blacklisted regions for credit fraud prevention in related technologies, this application proposes a grid-based system and method for identifying blacklisted regions for credit fraud prevention.

[0029] Reference Figure 1 This application provides a schematic flowchart of a credit anti-fraud blacklist region identification system based on grid partitioning. The system may include: The data acquisition and preprocessing module 100 is used to acquire and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as acquired business data. The grid division and coding module 200 is connected to the data acquisition and preprocessing module. It is used to receive preprocessed GPS coordinate data and business data, perform spatial discretization and grid coding on the preprocessed GPS coordinate data to obtain GPS grids, and statistically analyze multiple indicators within each GPS grid based on the business data. The feature engineering construction module 300 communicates with the data acquisition and preprocessing module and the gridding and encoding module. It is used to receive multiple indicators and preprocessed image features, and determine multiple feature vectors based on the multiple indicators and preprocessed image features. The black grid recognition model training and execution module 400 is connected to the feature engineering construction module. It is used to receive multiple feature vectors, input the multiple feature vectors into the trained black grid recognition model, and obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

[0030] In this embodiment, the data acquisition and preprocessing module is used to acquire and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as acquired business data. The gridding and encoding module is communicatively connected to the data acquisition and preprocessing module and is used to receive the preprocessed GPS coordinate data and business data, perform spatial discretization and grid encoding based on the preprocessed GPS coordinate data to obtain GPS grids, and statistically analyze multiple indicators within each GPS grid based on the business data. The feature engineering construction module is communicatively connected to the data acquisition and preprocessing module and the gridding and encoding module and is used to receive multiple indicators and preprocessed image features, and determine multiple feature vectors based on the multiple indicators and preprocessed image features. The black grid recognition model training and execution module is communicatively connected to the feature engineering construction module and is used to receive multiple feature vectors, input the multiple feature vectors into the trained black grid recognition model, and obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels. Thus, by dividing the system into GPS grids and then comprehensively considering GPS coordinate data and business data to statistically analyze multiple indicators within each GPS grid, and then determining multiple feature vectors based on these indicators and facial recognition images, and finally performing black grid identification based on these feature vectors, the accuracy of identifying credit fraud blacklist areas can be improved, achieving precise location and identification of high-fraud areas. The analysis results based on gridded features have clear business implications, and the output conclusions are clear and interpretable, facilitating understanding, verification, and decision support for risk control personnel, and promoting the rapid deployment and promotion of the system in business environments. Through automated and intelligent model recognition and interception mechanisms, the reliance on manual verification is significantly reduced, lowering risk control operating costs.

[0031] The aforementioned process of collecting and preprocessing GPS coordinate data and facial recognition images to obtain preprocessed GPS coordinate data and preprocessed image features can be integrated with a client mobile application or web page. When a customer submits a credit application, the system obtains the user's authorized real-time location information (i.e., GPS coordinate data) and facial recognition image. Then, the data processing subunit cleans, deduplicates, and standardizes the collected GPS coordinate data to obtain preprocessed GPS coordinate data. The image processing subunit processes the facial recognition image to obtain preprocessed image features.

[0032] The aforementioned data collection can be achieved by connecting the credit business system with the post-loan management database to collect relevant business data in real time.

[0033] The aforementioned black grid recognition model can be constructed using algorithms such as XGBoost or LightGBM, or it can be constructed using other neural network models known to those skilled in the art. This embodiment does not provide a specific description or limitation of this model.

[0034] In some implementations, multiple metrics include total number of credit application customers, number of credit approved customers, total credit limit, number of credit approved customers, number of overdue customers, number of credit applications per day, and the average number of credit applications per day within a GPS grid of the same size.

[0035] In this embodiment, by comprehensively considering the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, the number of overdue customers, the number of credit applications per day, and the average number of credit applications per day within a GPS grid of the same size, a good data foundation can be laid for the subsequent feature engineering construction module to build multiple feature vectors.

[0036] The total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, the number of overdue customers, the number of credit applications per day, and the average number of credit applications per day within a GPS grid of the same size can be obtained through data statistics, which will not be described in detail in this embodiment.

[0037] In some implementations, the feature engineering construction module includes: The credit density feature submodule is used to calculate the daily credit application density within each grid based on the daily credit application volume and the average daily credit application volume; to obtain the credit application volume of adjacent time windows and to calculate the credit volume growth rate of adjacent time windows based on the credit application volume of adjacent time windows; and to obtain abnormal patterns. The face background consistency feature analysis submodule is used to calculate the similarity between the background features corresponding to different applicant face recognition images within the same GPS grid based on the preprocessed image features; based on the similarity, it calculates the proportion of high similarity image pairs, which are image pairs with a similarity greater than a first preset threshold; it is also used to obtain target background features from the preprocessed image features. The risk indicator feature statistics submodule is used to calculate multiple risk indicator features based on the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, and the number of overdue customers; it is also used to analyze the correlation between customers within the grid and other known fraudulent customers. The time series feature extraction submodule is used to identify abnormally concentrated applications during non-working hours through time series analysis models, and to calculate the credit volume change trend and periodic pattern between adjacent time periods.

[0038] In this embodiment, various feature vectors are extracted through the credit density feature submodule, the face background consistency feature analysis submodule, the risk indicator feature statistics submodule, and the time series feature extraction submodule, which can lay a good data foundation for the subsequent identification of black grids and improve the accuracy of black grid identification.

[0039] The aforementioned acquisition of anomaly patterns can include, but are not limited to, measuring anomalies through fluctuation levels. For example, two standard deviations can be considered a general anomaly pattern, while three standard deviations can be considered a comparative anomaly pattern.

[0040] The above-mentioned extraction of target background features from preprocessed image features can be achieved by directly extracting key object features (i.e., target background features) in the image background from the preprocessed image features, such as features of identical walls, furniture, decorations, etc.

[0041] The above analysis of the relationship between customers within the grid and other known fraudulent customers can be achieved by using a network of connections (such as graph technology) to quantify the network of connections between customers within the grid and other known fraudulent customers. The results show that customers within the grid have first-degree relationships, second-degree relationships, etc. with other known fraudulent customers.

[0042] The aforementioned method of identifying abnormally concentrated applications during non-working hours and calculating the credit volume change trend and periodic pattern between adjacent time periods using a time series analysis model can be achieved by directly identifying abnormally concentrated applications during non-working hours and calculating the credit volume change trend and periodic pattern between adjacent time periods using an ARIMA or ARMA model. It should be noted that ARIMA or ARMA models are technologies well-known to those skilled in the art, and this embodiment will not describe them in detail.

[0043] The aforementioned first preset threshold can be a pre-set threshold, which can be changed according to the actual situation. This embodiment does not impose any specific limitations on it.

[0044] In some implementations, multiple risk metrics include the grid's historical first delinquency rate, credit approval rate, average credit limit, and delinquency rate.

[0045] In this embodiment, by comprehensively considering multiple risk indicators such as historical first-time delinquency rate, credit approval rate, average credit limit, and delinquency rate, a good data foundation can be laid for the subsequent identification of black grids, thereby improving the accuracy of black grid identification.

[0046] In some implementations, multiple risk indicator characteristics are calculated based on the total number of credit application customers, the number of customers whose credit has been approved, the total credit limit, the number of customers whose credit has been approved, and the number of overdue customers, including: The historical first delinquency rate of a grid is calculated based on the number of first-time delinquent customers and the total number of credit application customers within the grid. The credit approval rate is calculated based on the number of customers who have been granted credit and the total number of customers who have applied for credit. Calculate the average credit limit based on the total credit limit and the number of customers who have been granted credit. The delinquency rate is calculated based on the number of overdue customers and the number of customers whose credit has been approved.

[0047] In this embodiment, by comprehensively considering multiple risk indicators such as historical first-time delinquency rate, credit approval rate, average credit limit, and delinquency rate, a good data foundation can be laid for the subsequent identification of black grids, thereby improving the accuracy of black grid identification.

[0048] In some implementations, the trained black grid recognition model is obtained by training a set of training samples with normal labels or black grid labels, including: The historical first-overage data within the acquired grid is compared with the second preset threshold. If the historical first-overage data reaches the second preset threshold, the grid is marked as a black grid. If the historical first-overage data is less than the second preset threshold but greater than the third preset threshold, the grid is marked as a normal grid or a black grid through manual verification. If the historical first-overage data is less than the third preset threshold, the grid is marked as a normal grid. Construct a training sample set using grids labeled with normal or black grids; The black grid recognition model is trained by training the training sample set to obtain the trained black grid recognition model.

[0049] In this embodiment, training the black grid recognition model with accurately classified grids bearing normal or black grid labels can improve the recognition accuracy of the trained black grid recognition model.

[0050] The aforementioned second and third preset thresholds can be pre-set thresholds and can be changed according to actual conditions. This embodiment does not impose specific limitations on them.

[0051] In some implementations, the system also includes a real-time risk warning and response module for making risk decisions and issuing warnings based on the black grid identification results.

[0052] In this embodiment, risk decision-making and early warning are carried out through the real-time risk warning and response module, which can complete millisecond-level risk scanning and decision-making when applying for credit, realizing a fundamental shift from post-event tracing to in-event proactive interception, and greatly improving the timeliness of risk prevention and control.

[0053] To facilitate understanding by those skilled in the art, a set of preferred embodiments is provided below: This embodiment provides a credit fraud anti-fraud black grid intelligent identification system based on GPS gridding and multi-source feature fusion. By integrating functional modules such as spatial clustering pattern analysis of application locations, visual consistency detection of facial backgrounds, and historical credit performance statistics, it automatically and accurately identifies high-risk geographical areas (i.e., black grids) centrally operated by intermediaries, thereby solving the problem of insufficient ability of existing systems to identify regional gang fraud and improving the automation level and decision-making accuracy of credit risk control. The specific technical solution of this embodiment includes the following: 1. Data Acquisition and Preprocessing Module. This module consists of three sub-units, responsible for collecting and standardizing input data from different sources, specifically: (1) GPS coordinate data acquisition unit: It interfaces with the client mobile application or web page to obtain the real-time location information authorized by the user when the customer submits a credit application. This unit records the precise GPS coordinates (latitude and longitude, retaining at least 6 decimal places) of each application, and at the same time collects auxiliary data such as application timestamp, device fingerprint and IP address, and outputs them to the data cleaning sub-unit. The data processing sub-unit is used to clean, deduplicate and standardize the collected GPS coordinate data.

[0054] (2) Face Image Acquisition and Background Analysis Unit: Integrated into the identity verification process, this unit is used to acquire customer face recognition images. It incorporates an image processing subunit capable of extracting color distribution, texture features, and key object information from the image background area. It also converts background features into numerical feature vectors (i.e., pre-processed image features) using vectorization techniques. The image processing subunit incorporates techniques such as color moments, improved LBP and Haar cascade detection, feature stitching, Min-Max normalization, PCA dimensionality reduction, and low-dimensional numerical feature vector transformation for feature extraction and transformation.

[0055] It should be noted that the customer facial recognition images collected in this embodiment have all been authorized by the customer.

[0056] (3) Business Data Collection Unit: Connects the credit business system and the post-loan management database to collect data such as customer credit application results (approval / rejection) and credit limits in real time, and continuously tracks repayment performance, recording whether there is a first default and the number of overdue customers. This unit also includes a tag management submodule, which supports manual or automatic tagging of confirmed fraud cases.

[0057] 2. GPS Grid Generation and Encoding Module. This module receives preprocessed GPS coordinate data, performs spatial discretization and grid encoding to obtain a GPS grid, and then encodes the GPS grid. Specifically: (1) Grid precision configuration submodule: Supports configuring grid precision parameters according to business needs (e.g., 3-digit precision is about 100 meters × 100 meters, 4-digit precision is about 10 meters × 10 meters). Allows setting differentiated precision strategies by city or region.

[0058] (2) Grid Coding Generation Submodule: This module truncates the GPS latitude and longitude coordinates in the GPS grid to a specified number of decimal places, generating a unique grid code. For example, coordinates (116.397428, 39.909843) → grid code (116.397, 39.909). It supports maintaining a mapping table between grid codes and geographic regions. Specifically, grid codes can be mapped to corresponding coordinates in the coordinate system, and then these coordinates can be mapped to specific geographic regions.

[0059] (3) Grid Statistical Indicators Submodule: Based on grid coding and business data, it statistically analyzes indicators such as the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, the number of overdue customers, the average number of credit applications per day, and the number of credit applications per day within each GPS grid; it can further analyze the distribution of different account managers and channel sources.

[0060] It should be noted that statistical indicators can be calculated using simple statistical methods such as cumulative summation and averaging. This embodiment will not describe this in detail.

[0061] 3. Feature Engineering Construction Module. This module extracts multi-dimensional risk features from the gridded data indicators of the GPS grid division and coding module for use by the subsequent black grid identification model. Specifically: (1) Credit Density Feature Submodule: Calculates the daily or hourly credit application density within each grid. The daily credit application density within a grid = daily credit application volume of a certain grid / average daily credit application volume of grids of the same size nationwide. The calculation of the hourly credit application density is similar to that of the daily credit application density, and will not be described in detail in this embodiment. Calculates the credit volume growth rate of adjacent time windows. The credit volume growth rate of adjacent time windows (assuming time windows A and B are adjacent) = (credit application volume of time window A - credit application volume of time window B) / credit application volume of time window B. Identifies abnormal patterns of sudden increases in credit volume within a short period of time.

[0062] (2) Face Background Consistency Feature Analysis Submodule: This module calls the numerical feature vector output by the face image acquisition and background analysis unit and uses image similarity algorithms (such as SSIM and perceptual hashing) to calculate the similarity of the face background features of different applicants within the same grid. Based on the similarity, the proportion of high-similarity image pairs is calculated as a consistency index; images with similarity exceeding a first preset threshold are considered high-similarity image pairs. The ratio between the number of high-similarity image pairs and the total number of image pairs within the same grid is calculated to obtain the proportion of high-similarity image pairs; image pairs are face recognition image pairs. Key object features (i.e., target background features, such as identical walls, furniture, decorations, etc.) in the image background are directly extracted from the numerical feature vector.

[0063] (3) Risk Indicator Characteristics Statistical Submodule: Calculates the historical first-time delinquency rate of the grid = number of first-time delinquent customers within the grid / total number of credit application customers within the grid; calculates derived indicators such as credit approval rate, average credit limit, and delinquency rate of the grid; analyzes the relationship network between customers within the grid and other known fraudulent customers. Through the relationship network (constructed using graph technology), the relationship network between customers within the grid and other known fraudulent customers is quantified, and the results show the first-degree relationship, second-degree relationship, etc., between customers within the grid and other known fraudulent customers. Among them: ; ; ; (4) Time series feature extraction submodule: The ARIMA or ARMA model is used to analyze the temporal distribution of grid credit volume, identify abnormal concentrated applications during non-working hours, and calculate the credit volume change trend and periodic pattern of adjacent time periods.

[0064] 4. Black Mesh Recognition Model Training and Execution Module. This module includes two sub-modules: offline training and online service, enabling model construction and deployment. Specifically: (1) The offline training submodule includes multiple units, specifically: Sample labeling unit: Based on historical first-time violation data and manual verification results, grids are labeled as "normal" or "black grids," constructing a labeled training sample set. Specifically, this is done through online, telephone, or on-site surveys. If the first-time violation data reaches the second preset threshold, it is defined as a "black grid." If the first-time violation data does not reach the second preset threshold but is relatively close (i.e., greater than the third preset threshold), a supplementary manual survey is conducted, and the results of the manual survey prevail. Each grid is considered a sample, with labels divided into "normal" and "black grid."

[0065] Feature Engineering Unit: Calls multiple features from the feature engineering construction module (including daily credit application density, credit growth rate of adjacent time windows, abnormal patterns, consistency indicators, and key object features, etc.) to generate grid-level feature vectors, and performs preprocessing such as missing value imputation and normalization.

[0066] Model training unit: The black grid recognition model is constructed using algorithms such as XGBoost or LightGBM. The preprocessed grid-level feature vector is used as input and the grid label is used as output. Hyperparameter tuning is completed through grid search and cross-validation to train the optimal classification model (i.e., the trained black grid recognition model is obtained).

[0067] Evaluation and Optimization Unit: Uses metrics such as AUC-ROC, precision, and recall to evaluate the model, analyze feature importance, and support regular iteration and retraining of the black grid recognition model.

[0068] (2) Online service submodule: Deploy the trained black grid recognition model as a service interface that can be called in real time, receive the feature vector of the newly applied grid, and output the risk score and black grid probability of the grid.

[0069] 5. Real-time Risk Warning and Response Module. This module implements a closed-loop process from risk identification to handling and feedback. Specifically: (1) Real-time feature calculation unit: For newly submitted credit applications, determine the GPS grid to which it belongs in real time, and process it through some steps in steps 1 to 3 above to quickly extract the recent behavioral features of the grid (such as recent application density, background consistency, etc.), that is, extract the same feature vector as when training the black grid recognition model.

[0070] (2) Risk Decision Engine: The feature vector of the newly submitted credit application is input into the trained black grid identification model to obtain the risk score and black grid probability of the grid (i.e., the black grid identification result). For applications with a risk score lower than the fourth preset threshold, "approved" is automatically returned and credit is granted. For applications with a score in the middle gray zone, an early warning is triggered and the application is pushed to the risk control specialist for manual review. For applications with a score higher than the threshold or that hit a black grid, the application is automatically rejected and the subsequent process is blocked. For grids that are not yet confirmed but have abnormal behavior, they are included in the observation list for continuous monitoring. The middle gray zone is the model scoring range, which is an intuitive way to express the predicted probability of the black grid identification model as a score through statistical techniques, making it easy for risk management personnel to use. For example, the risk score given by the black grid identification model training and execution module is 50% for grid A, which is 600 points after conversion. Since the 50% probability is highly uncertain, whether the grid is a "black grid" needs to be pushed to the risk control specialist for manual review.

[0071] It should be noted that the fourth preset threshold in this embodiment can be a pre-set threshold, which can be changed according to the actual situation. This embodiment does not make specific limitations on this.

[0072] (3) Early warning execution and feedback unit: executes automatic rejection or manual review instructions, generates risk warning reports in real time and pushes them to the risk control system; records the handling results and forms a feedback data stream for iterative optimization of the black grid identification model.

[0073] (4) Monitoring and display unit: The high-risk grid heat map, TOP risk grid list and handling progress are displayed in real time through the risk control screen; it supports alarm push through multiple channels such as SMS and email; it regularly generates BI analysis reports, monitors model performance and feature drift, and triggers automatic retraining and hot update of black grid recognition model when anomalies occur.

[0074] For practical applications of this embodiment, which uses multiple modules from steps 1 to 5 to construct the overall system, please refer to [the relevant documentation]. Figures 2 to 3 As shown.

[0075] Compared with the prior art, the technical solution of this embodiment has the following advantages: 1. High recognition accuracy: By integrating multiple functional modules such as GPS grid behavior analysis, face background consistency detection and historical overdue data statistics, it achieves accurate positioning and identification of areas with high incidence of fraud by intermediary gangs, significantly reducing the false alarm rate and false negative rate of traditional methods.

[0076] 2. Strong real-time response capability: The system has an embedded real-time risk warning and response module, which can complete millisecond-level risk scanning and decision-making when applying for credit, realizing a fundamental shift from post-event tracing to in-event proactive interception, greatly improving the timeliness of risk prevention and control.

[0077] 3. Adaptive and continuously evolving capabilities: The black grid recognition model training and execution module in the system supports online learning and regular updates, can automatically track the spatial migration and behavioral evolution of fraud patterns, realize the dynamic adjustment and optimization of risk strategies, and has a good level of intelligence.

[0078] 4. The results are highly interpretable and easy to deploy: The analysis results based on gridded features have clear business implications, and the output conclusions are clear and interpretable, which is convenient for risk control personnel to understand, verify and support decision-making, and facilitates the rapid implementation and promotion of the system in the business environment.

[0079] 5. Significantly reduce operating costs and risk losses: Through automated and intelligent identification and interception mechanisms, the reliance on manual verification is greatly reduced, thereby lowering risk control operating costs; at the same time, by blocking high-risk fraudulent applications in advance, potential bad debt losses are directly reduced, resulting in significant economic benefits.

[0080] 6. Broad coverage of risk dimensions and strong forward-looking prevention and control: Starting from the analysis of group behavior patterns and spatial correlations, the system can effectively identify new, cross-regional and concealed group fraud that is difficult to detect by traditional rule models or single-point analysis, thus expanding the coverage and forward-looking nature of risk prevention and control.

[0081] 7. Modular design, high integration, easy maintenance and expansion: The system adopts a modular architecture design, with clear responsibilities and interfaces for each functional unit (such as grid partitioning, feature engineering and model execution), which facilitates system maintenance, function upgrades and integration with existing risk control platforms.

[0082] Reference Figure 4 This application also provides a grid-based method for identifying blacklisted regions in credit fraud prevention, applied to the aforementioned grid-based system for identifying blacklisted regions in credit fraud prevention. The method includes the following steps: Step S100: Collect and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as collected business data; Step S200: Spatial discretization and grid encoding are performed on the preprocessed GPS coordinate data to obtain a GPS grid. Based on the business data, multiple indicators within each GPS grid are statistically analyzed. Step S300: Based on multiple indicators and preprocessed image features, determine multiple feature vectors; Step S400: Input multiple feature vectors into the trained black grid recognition model to obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

[0083] It should be noted that since the grid-based credit fraud blacklist region identification method in this embodiment is based on the same inventive concept as the grid-based credit fraud blacklist region identification system described above, the corresponding content in the system embodiment is also applicable to this method embodiment, and will not be described in detail here.

[0084] Reference Figure 5 This application also provides an electronic device, which includes: At least one memory; At least one processor; At least one program; The program is stored in memory, and the processor executes at least one program to implement the above-described grid-based credit anti-fraud blacklist region identification method.

[0085] This electronic device can be any smart terminal, including mobile phones, tablets, personal digital assistants (PDAs), and in-vehicle computers.

[0086] The electronic devices according to embodiments of this application will now be described in detail.

[0087] The processor 1600 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure. The memory 1700 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 1700 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1700 and is called and executed by the processor 1600 to execute the grid-based credit anti-fraud blacklist region identification method of this disclosure.

[0088] The input / output interface 1800 is used to implement information input and output. The communication interface 1900 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 2000 transmits information between various components of the device (e.g., processor 1600, memory 1700, input / output interface 1800, and communication interface 1900); The processor 1600, memory 1700, input / output interface 1800 and communication interface 1900 are connected to each other within the device via bus 2000.

[0089] This disclosure also provides a storage medium, which is a computer-readable storage medium storing computer-executable instructions for causing a computer to execute the above-described grid-based credit anti-fraud blacklist region identification method.

[0090] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0091] The embodiments described in this disclosure are for the purpose of more clearly illustrating the technical solutions of this disclosure and do not constitute a limitation on the technical solutions provided by this disclosure. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by this disclosure are also applicable to similar technical problems.

[0092] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this disclosure, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0093] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0094] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0095] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0096] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and 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 through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0098] 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.

[0099] Furthermore, the functional units in the various embodiments of this application 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. The integrated unit can be implemented in hardware or as a software functional unit.

[0100] If the integrated unit 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 application, in essence, or the part that contributes to the prior art, or all or 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 multiple instructions to cause an electronic 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 application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. The embodiments of this application have been described in detail above with reference to the accompanying drawings, but this application is not limited to the above embodiments. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of this application.

[0101] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application.

Claims

1. A credit fraud blacklist region identification system based on grid partitioning, characterized in that, The system includes: The data acquisition and preprocessing module is used to acquire and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as acquired business data. The grid division and encoding module is communicatively connected to the data acquisition and preprocessing module. It is used to receive the preprocessed GPS coordinate data and the business data, perform spatial discretization and grid encoding on the preprocessed GPS coordinate data to obtain a GPS grid, and statistically analyze multiple indicators within each GPS grid based on the business data. The feature engineering construction module is communicatively connected to the data acquisition and preprocessing module and the gridding and encoding module, and is used to receive the multiple indicators and the preprocessed image features, and determine multiple feature vectors based on the multiple indicators and the preprocessed image features; The black grid recognition model training and execution module is communicatively connected to the feature engineering construction module. It is used to receive the multiple feature vectors, input the multiple feature vectors into the trained black grid recognition model, and obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

2. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 1, characterized in that, The metrics include total number of credit application customers, number of credit approved customers, total credit limit, number of credit approved customers, number of overdue customers, daily credit application volume, and the average daily credit application volume within a GPS grid of the same size.

3. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 2, characterized in that, The feature engineering construction module includes: The credit density feature submodule is used to calculate the daily credit application density within each grid based on the daily credit application volume and the average daily credit application volume; to obtain the credit application volume of each adjacent time window and to calculate the credit volume growth rate of adjacent time windows based on the credit application volume of each adjacent time window; and to obtain abnormal patterns. The face background consistency feature analysis submodule is used to calculate the similarity between the background features corresponding to different applicant face recognition images within the same GPS grid based on the preprocessed image features; calculate the proportion of high similarity image pairs based on the similarity, wherein the high similarity image pairs are image pairs with a similarity greater than a first preset threshold; and also to obtain target background features from the preprocessed image features. The risk indicator feature statistics submodule is used to calculate multiple risk indicator features based on the total number of credit application customers, the number of credit approved customers, the total credit limit, the number of credit approved customers, and the number of overdue customers; it is also used to analyze the correlation between customers within the grid and other known fraudulent customers. The time series feature extraction submodule is used to identify abnormally concentrated applications during non-working hours through time series analysis models, and to calculate the credit volume change trend and periodic pattern between adjacent time periods.

4. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 3, characterized in that, The aforementioned risk indicators include the grid's historical first-time delinquency rate, credit approval rate, average credit limit, and delinquency rate.

5. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 4, characterized in that, The method calculates multiple risk indicator characteristics based on the total number of credit application customers, the number of customers whose credit has been approved, the total credit limit, the number of customers whose credit has been approved, and the number of overdue customers, including: The historical first delinquency rate of a grid is calculated based on the number of first-time delinquent customers and the total number of credit application customers within the grid. The credit approval rate is calculated based on the number of customers who have been granted credit and the total number of customers who have applied for credit. Calculate the average credit limit based on the total credit limit and the number of customers who have granted credit. The delinquency rate is calculated based on the number of overdue customers and the number of customers whose credit has been granted.

6. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 1, characterized in that, The trained black grid recognition model is obtained by training a set of training samples with normal labels or black grid labels, including: The historical first-overage data within the acquired grid is compared with a second preset threshold. If the historical first-overage data reaches the second preset threshold, the grid is marked as a black grid. If the historical first-overage data is less than the second preset threshold but greater than a third preset threshold, the grid is marked as a normal grid or a black grid through manual verification. If the historical first-overage data is less than the third preset threshold, the grid is marked as a normal grid. Construct a training sample set using grids labeled with normal or black grids; The black grid recognition model is trained using the training sample set to obtain the trained black grid recognition model.

7. The credit anti-fraud blacklist region identification system based on grid partitioning according to claim 1, characterized in that, The system also includes a real-time risk warning and response module, which is used to make risk decisions and issue warnings based on the black grid identification results.

8. A method for identifying blacklisted regions in credit fraud prevention based on grid partitioning, characterized in that, The method, applied to the grid-based credit fraud blacklist region identification system according to any one of claims 1 to 7, comprises: Collect and preprocess GPS coordinate data and face recognition images to obtain preprocessed GPS coordinate data and preprocessed image features, as well as collected business data; Spatial discretization and grid encoding are performed on the preprocessed GPS coordinate data to obtain a GPS grid. Based on the business data, multiple indicators within each GPS grid are statistically analyzed. Based on the aforementioned multiple indicators and the preprocessed image features, multiple feature vectors are determined; The multiple feature vectors are input into the trained black grid recognition model to obtain the black grid recognition result. The trained black grid recognition model is obtained by training a training sample set with normal labels or black grid labels.

9. An electronic device, characterized in that, It includes at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor, which, when executed by the at least one control processor, enable the at least one control processor to perform the grid-based credit anti-fraud blacklist region identification method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the grid-based credit fraud blacklist region identification method as described in claim 8.

Images