Method and device for identifying internet credit fraud risk based on GPS track
By collecting and processing the borrower's GPS trajectory data and combining multiple models to identify fraud risks, the problem of difficult borrower behavior characteristics in traditional methods is solved, and efficient identification and dynamic monitoring of Internet credit fraud risks is achieved.
Patent Information
- Application Number
- CN202510683093.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-26
- Publication Date
- 2025-08-08
AI Technical Summary
Traditional Internet credit fraud risk identification methods rely on static information and are difficult to capture the actual behavior and activities of borrowers, resulting in insufficient accuracy in identifying fraud risk.
By collecting the borrower's GPS trajectory data, data cleaning and processing are carried out in the application, loan and post-loan stages, combining logistic regression, XGBoost and Transformer models, GPS trajectory features and basic information and credit information are integrated to identify fraud risks.
It improves the accuracy of identifying Internet credit fraud risks, can promptly detect potential fraud risks, reduce losses, and achieve dynamic monitoring and early warning of risks.
Smart Images
Figure CN120450852A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and in particular to a method and device for identifying Internet credit fraud risks based on GPS tracks. Background Art
[0002] Internet lending, with its convenient and efficient features, provides borrowers with a quick way to obtain funds, while also expanding the business scope and improving service efficiency for financial institutions. However, the rapid growth of internet lending has also brought with it increasingly serious fraud risks.
[0003] In internet lending scenarios, due to information asymmetry between lenders and borrowers, borrowers may engage in fraudulent activities, such as intentionally concealing their true circumstances and providing false information to obtain loans. This fraudulent behavior not only causes direct economic losses to financial institutions, but also disrupts the normal order of the financial market and affects the healthy development of the entire industry. Therefore, how to effectively identify and prevent internet lending fraud risks has become a major challenge for financial institutions.
[0004] Traditional methods for identifying internet credit fraud risk primarily rely on basic information and creditworthiness provided by borrowers, such as ID numbers, contact information, proof of income, and credit reports. While this information can reflect a borrower's creditworthiness to a certain extent, it has significant limitations. For one thing, some borrowers may falsify or tamper with this information, making it difficult for risk identification models based on this information to accurately assess their true risk level. Furthermore, this static information cannot fully reflect the borrower's actual behavior and activities, making it difficult to detect potential fraudulent activity.
[0005] In recent years, with the widespread adoption of Global Positioning System (GPS) technology, borrowers' device GPS data has become a new source of data for risk identification. GPS data can record borrowers' geographic locations in real time, reflecting their daily activities and behavioral patterns. By analyzing borrowers' GPS trajectory data, behavioral characteristics that are not reflected in traditional information can be obtained, such as the borrower's usual location, travel patterns, and hotspots. These characteristics are valuable for identifying fraud risks, as fraudulent behavior is often accompanied by unusual behavioral patterns, such as frequent changes of residence and activities in high-risk areas. Summary of the Invention
[0006] The purpose of the present invention is to propose a method and device for identifying Internet credit fraud risks based on GPS tracks, which can identify fraud risks in combination with GPS tracks.
[0007] To achieve the above objectives, in a first aspect, the present invention provides a method for identifying Internet credit fraud risks based on GPS trajectory, comprising: Collect data from the application stage, loan stage, and post-loan stage, including: When a borrower applies for a loan through an internet credit platform, the GPS data of the borrower's device is collected, along with the borrower's basic information, credit information, and device information; During the loan period from loan disbursement to repayment, GPS trajectory data of the borrower's device is continuously collected, and the collection frequency is dynamically adjusted based on risk assessment; During the repayment phase, GPS data is collected based on repayment progress and risk status; Clean the collected data, including removing duplicate data, filling missing values, and identifying and handling outliers; Processing of GPS coordinates, including coordinate correction and discretization; Extract GPS trajectory features and integrate them with the borrower's basic information, credit information, and device information; Use logistic regression, XGBoost, and Transformer models for training and evaluation, and select the best-performing model as the final fraud risk identification model; Conduct risk identification, including real-time credit risk interception, credit risk assessment during loan issuance, and post-loan fraud risk warning, and adopt corresponding strategies based on the risk identification results.
[0008] The basic solution's beneficial effects: Comprehensive data collection at the application, loan, and post-loan stages covers borrowers' GPS data, basic information, credit information, device information, and other multi-dimensional data. This provides a comprehensive understanding of borrowers' behaviors and characteristics, providing a rich data foundation for accurately identifying fraud risks.
[0009] During the loan phase, the frequency of GPS trajectory data collection is dynamically adjusted based on risk assessment, allowing for more flexible capture of changes in borrower behavior. For higher-risk borrowers, the frequency can be increased to promptly identify unusual behavior; for lower-risk borrowers, the frequency can be appropriately reduced to conserve resources. During the repayment phase, GPS data collection based on repayment progress and risk profiles can also be used to monitor borrower behavior in a targeted manner, improving the accuracy of risk identification.
[0010] Cleaning collected data, removing duplicates, filling missing values, and identifying and processing outliers can improve data quality and reduce the impact of data noise on the model. Correcting and discretizing GPS coordinates helps improve the accuracy and usability of GPS data, making subsequent feature extraction and model training more accurate.
[0011] Integrating GPS trajectory features with the borrower's basic information, credit information, and device information can fully leverage the complementary nature of different data types. GPS trajectory data can reflect the borrower's actual behavior and range of activities. Combined with other information, it can provide a more comprehensive profile of the borrower and improve the accuracy of fraud risk identification.
[0012] Using multiple models such as logistic regression, XGBoost, and Transformer for training and evaluation, and selecting the best-performing model as the final fraud risk identification model, can fully leverage the advantages of different models and find the most suitable model for the problem through comparison and optimization.
[0013] By implementing a fraud risk identification model that intercepts credit risk in real time, assesses credit utilization risks during lending, and provides early warning of post-loan fraud risks, potential fraud risks can be promptly identified at different stages of the business. Implementing appropriate strategies based on risk identification results can effectively mitigate credit risk and reduce losses. For example, real-time interception of high-risk credit applications can prevent loans from being issued to borrowers with a high fraud risk. Risks identified during lending can be promptly controlled, such as adjusting credit limits and strengthening monitoring. Post-loan fraud risk warnings can also help ensure proactive response and asset security.
[0014] As an implementable preferred solution, the collected data is cleaned, including the following: Use the pandas library to process the collected GPS trajectory data, identify and delete duplicate records by judging the data's timestamp, coordinate value, and device ID information; For missing values in GPS trajectory data, if the missing time is short, the coordinates of the missing points are calculated using linear interpolation based on the coordinate values of the previous and next time points; if the missing time is long, the coordinates are estimated by combining the borrower's historical trajectory pattern and surrounding environment information; Anomalies in GPS trajectory data are identified by setting a threshold range. If a record shows that the movement distance exceeds this range and excluding special circumstances, the record is marked as an anomaly.
[0015] As an implementable preferred solution, the GPS coordinates are processed, including the following: Coordinate correction uses commercial map API to correct GPS coordinates. An HTTP request is sent to the commercial map correction interface to correct the coordinates of each collected GPS track data one by one, and the corrected coordinates replace the original coordinates.
[0016] As an implementable preferred solution, processing the GPS coordinates also includes the following: Discretization processing divides the geographic area into fixed-size grids, determines the number of grid rows and columns based on the coordinate range, constructs a grid index, maps the corrected GPS track points to the corresponding grid cells, determines the grid to which the track point belongs based on its latitude and longitude coordinates, and records the track point's timestamp, grid number, and other related information to form discretized track data.
[0017] As an implementable preferred solution, extracting GPS trajectory features includes the following: Extract the dwell time feature and calculate the borrower's dwell time in each grid cell. By analyzing the timestamps of the trajectory points, if multiple consecutive trajectory points belong to the same grid cell, the sum of the time intervals between these points is calculated as the dwell time. Extract the moving speed feature and calculate the moving speed based on the coordinates and time intervals of adjacent trajectory points. The formula is:
[0018] in, and are the coordinates of adjacent trajectory points, and is the corresponding timestamp; Extract the moving direction feature and calculate the moving direction between adjacent trajectory points; express the moving direction d by calculating the azimuth between two points. The formula is: .
[0019] As an implementable and optimal solution, we used logistic regression, XGBoost, and Transformer models for training and evaluation, and selected the best-performing model as the final fraud risk identification model, including the following: Evaluation indicators include Accuracy, Recall and F1-Score, and the calculation formulas are:
[0020]
[0021]
[0022] Among them, Precision is the accuracy rate, that is, the ratio of the number of correctly predicted positive samples to the number of predicted samples.
[0023] As an implementable and preferred solution, risk identification includes the following: Real-time credit risk interception is to input the processed real-time data into a trained fraud risk identification model. The model outputs the risk rating results and adopts corresponding strategies based on the risk rating results.
[0024] As an implementable and preferred solution, risk identification includes the following: The credit risk assessment during loan period is to regularly collect GPS track data of existing loan customers during loan period and conduct correlation analysis with pre-loan data. Based on the historical track analysis results and the risk level assessed by the model, different credit limits and differentiated interest rates are adapted. Calculate the similarity between the current trajectory and the pre-loan trajectory using the following formula:
[0025] in, and For two trajectories, For trajectory points and The distance between is the path weight; the higher the trajectory similarity score, the more stable the trajectory and the lower the risk.
[0026] As an implementable and preferred solution, risk identification includes the following: Post-loan fraud risk warning closely monitors the borrower's repayment behavior and GPS trajectory changes. If the borrower shows signs of overdue repayment and his GPS trajectory shows abnormal behavior, the system triggers a risk warning and takes differentiated early collection intervention measures based on the risk level.
[0027] In a second aspect, the present invention also provides an electronic device that utilizes the above-mentioned method for identifying Internet credit fraud risks based on GPS tracks, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 This is a logical diagram of the method for identifying Internet credit fraud risks based on GPS trajectories.
[0029] Figure 2 FIG. 2 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0030] In order to make the technical solution and advantages of the present application clearer, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings. It will be understood that the specific embodiments described herein are only partial embodiments of the present invention, which are only used to explain the present application, rather than to limit the present application. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered to be isolated, and they can be combined with each other to achieve better technical effects. The same reference numerals appearing in the drawings of the following embodiments represent the same features or components, which can be applied to different embodiments.
[0031] In addition, unless otherwise defined, technical or scientific terms used in the description of the present invention should have the common meanings understood by those skilled in the art in the art to which the present invention belongs.
[0032] The present invention will be further described in detail below with reference to the accompanying drawings: Reference numerals: electronic device 500 , processor 501 , communication interface 502 , memory 503 , bus 504 .
[0033] Reference Figure 1 The present disclosure provides a method for identifying Internet credit fraud risks based on GPS trajectory, including: Step S100, collecting data from the application stage, loan stage, and post-loan stage, includes: In step S101, when a borrower applies for a loan through an online credit platform, the platform app, after obtaining user authorization, collects GPS data from the borrower's device in real time. This data also includes basic borrower information, such as name, ID number, contact information, loan amount, and loan term; credit information, including credit score and historical credit history; and device information, such as device model, device ID, and IP address.
[0034] Step S102: During the loan period, from loan disbursement to repayment, GPS trajectory data from the borrower's device is continuously collected. The frequency of collection can be dynamically adjusted based on risk assessment. If abnormal risk indicators are found, the collection frequency is increased to 100% to promptly monitor changes in the borrower's behavior.
[0035] Step S103: During the repayment phase, GPS data is collected based on repayment progress and risk profile. For borrowers who repay on time, the collection frequency is kept low. For borrowers showing signs of overdue payments, the frequency is increased, closely monitoring their location changes to provide a basis for risk warnings and collection interventions.
[0036] Step S200, data cleaning, includes: Step S201 removes duplicate data. The collected GPS trajectory data is processed using the Pandas library. Duplicate records are identified and deleted by examining information such as the data's timestamp, coordinate values, and device ID. For example, for multiple records from the same device at the same time and coordinate location, only one is retained to ensure data uniqueness.
[0037] Step S202: For missing values in GPS trajectory data, interpolation is used to fill in the missing values. If the missing time is short (no more than 10 minutes), the coordinates of the missing points are calculated using linear interpolation based on the coordinate values of the previous and next time points. If the missing time is long, it is estimated by combining the borrower's historical trajectory pattern and surrounding environment information. For missing values in other data (such as basic information, credit information, etc.), they are processed according to the characteristics of the data. For numerical data, such as credit scores, if the missing ratio is low (for example, less than 5%), the mean or median can be used for filling; if the missing ratio is high, it can be considered to fill in or delete the corresponding records based on the characteristics of similar borrowers. For text data, such as borrower occupation information, if it is missing, it can be marked as "unknown".
[0038] Step S203 identifies outliers in the GPS trajectory data by setting a threshold range. For example, based on common sense and historical data, borrowers' travel distances within a short period of time should normally not exceed a certain range. If a record shows a travel distance outside this range, excluding special circumstances (such as airplane travel), the record is marked as an outlier. Outliers can be corrected or deleted based on the specific circumstances. If the outlier is due to a data collection error, an attempt can be made to correct it based on previous and subsequent data. If the cause cannot be determined, the outlier is deleted.
[0039] Step S300, processing the GPS coordinates, including: Step S301: Coordinate correction. GPS coordinates are corrected using a commercial map API. In the Python code, import the relevant API call library, such as the requests library. According to the API documentation, construct a request parameter containing the original GPS coordinate information and send an HTTP request to the commercial map correction interface. The coordinates of each collected GPS track are corrected one by one, replacing the original coordinates with the corrected coordinates to ensure that subsequent analysis is based on more accurate geographic location information.
[0040] Step S302 involves discretization, dividing the geographic area into fixed-size grids, for example, 1 km x 1 km. The number of rows and columns in the grid is determined based on the coordinate range, and a grid index is constructed. The Python numpy library is used to create a two-dimensional array representing the grid, with each element of the array corresponding to the number or related attributes of a grid cell.
[0041] Map the corrected GPS track points to the corresponding grid cells. Determine the grid cell to which the track point belongs based on its longitude and latitude coordinates. For example, for longitude lon and latitude lat, the grid range is [min_lon, max_lon] for longitude and [min_lat, max_lat] for latitude, and the grid size is grid_size. The grid row index is row = int((lat - min_lat) / grid_size), and the column index is col = int((lon - min_lon) / grid_size). Record the track point's timestamp, grid number, and other relevant information to form the discretized track data.
[0042] Step S400, performing model training, includes: Step S401, extracting GPS trajectory features, including: Extract the dwell time feature and calculate the borrower's dwell time in each grid cell. By analyzing the timestamps of the trajectory points, if multiple consecutive trajectory points belong to the same grid cell, the sum of the time intervals between these points is calculated as the dwell time.
[0043] Extract the moving speed feature and calculate the moving speed based on the coordinates and time intervals of adjacent trajectory points. The formula is:
[0044] in, and are the coordinates of adjacent trajectory points, and The corresponding timestamp is used. The speed of each trajectory segment is calculated, and the average speed, maximum speed, and minimum speed are statistically analyzed.
[0045] Extract the moving direction feature and calculate the moving direction between adjacent trajectory points. The moving direction d is expressed by calculating the azimuth angle between two points. The formula is:
[0046] Step S402 involves feature fusion, integrating GPS trajectory features with the borrower's basic information, credit information, and device information. For example, age and gender can be combined with GPS trajectory features to analyze differences in trajectory behavior among borrowers of different age groups and genders. Credit scores can also be correlated with trajectory features to study the relationship between credit status and behavioral patterns. New feature combinations can be created, such as calculating the proportion of time spent in high-risk areas by borrowers of different credit ratings.
[0047] In step S403, the logistic regression, XGBoost, and Transformer models are trained and evaluated. The preprocessed data is divided into training, validation, and test sets according to a set ratio. Each model is trained on the training set, and the validation set is used to fine-tune model parameters. The trained models are evaluated on the test set, comparing their performance in fraud risk identification, with a focus on their ability to capture GPS time series features. Evaluation metrics include accuracy, recall, and F1-score, calculated using the following formulas:
[0048]
[0049]
[0050] Precision is the accuracy rate, which is the ratio of the number of correctly predicted positive samples to the number of predicted samples. The model with the best performance in each indicator is selected as the final fraud risk identification model.
[0051] Step S500, performing risk identification, includes: Step S501: Real-time credit risk interception, specifically as follows: When a borrower submits a loan application, the system obtains real-time GPS trajectory data, basic information, device information, etc. According to the above data preprocessing process, the real-time data is cleaned, rectified, and discretized to extract the corresponding features.
[0052] The processed real-time data is fed into a trained fraud risk identification model, which outputs a risk rating, categorizing it into three levels: high, medium, and low. The model also labels the GPS points it finds, recording key information such as the grid cell they belong to, their dwell time, and their movement speed. For example, if the model predicts a loan application as high-risk, and the GPS point corresponding to that application is located in a high-risk area (such as a gambling den), the GPS point is marked as high-risk, and the associated trajectory characteristics are recorded in detail.
[0053] Based on the risk rating results, the linked approval system implements appropriate strategies. For high-risk loan applications, the system automatically intercepts and prevents loan disbursement, forwarding the application information to the manual review team. Manual reviewers then make a comprehensive assessment based on GPS tracking information, other borrower information, and system-informed risk points. If the manual review confirms a risk of fraud, the loan application is rejected. If the risk is deemed manageable, the decision to proceed with the approval process can be made based on the actual circumstances. For medium-risk loan applications, partial manual review or enhanced ongoing loan monitoring can be selected based on specific business rules. For low-risk loan applications, rapid approval and loan disbursement are achieved.
[0054] Step S502, credit risk assessment during loan, is as follows: For existing loan customers, we regularly collect GPS trajectory data during their loan period and analyze it in correlation with pre-loan data. We analyze changes in the borrower's trajectory, such as whether there are unusually long periods of stay or frequent cross-regional movements. By comparing trajectory characteristics over different time periods, we calculate trajectory stability indicators, such as trajectory similarity scores. We use the Dynamic Time Warping (DTW) algorithm to calculate the similarity between the current trajectory and the pre-loan trajectory. The formula is:
[0055] in, and For two trajectories, For trajectory points and The distance between is the path weight. The higher the trajectory similarity score, the more stable the trajectory and the lower the risk.
[0056] Based on historical analysis and model-assessed risk levels, we adapt different credit limits and interest rates. For customers whose risk levels rise, we appropriately reduce credit limits and increase loan interest rates to cover the potential for fraud losses. For customers whose risk levels decline, we can consider increasing credit limits and lowering loan interest rates to improve customer satisfaction and loyalty.
[0057] Step S503, issuing a post-loan fraud risk warning, includes: During the repayment phase, the borrower's repayment behavior and GPS tracking are closely monitored. If a borrower shows signs of overdue payments and their GPS tracking shows unusual behavior (such as frequent visits to high-risk areas, prolonged stays in fixed locations, etc.), the system triggers a risk alert. Differentiated early collection intervention measures are implemented based on risk level. For high-risk customers, collection procedures are immediately initiated, requiring repayment through phone calls and door-to-door collection visits. For medium-risk customers, reminders are initially sent via SMS and emails. If the overdue situation persists, further collection measures are implemented.
[0058] Based on the results of post-loan risk warnings and collection interventions, we collect actual fraud cases and risk data and feed it into model training. We continuously optimize the model by adjusting model parameters or adding new features to enhance its ability to identify fraud risks. We also optimize risk assessment strategies and business processes, such as adjusting risk grading standards and improving approval processes, to continuously enhance overall risk control capabilities.
[0059] Those skilled in the art will appreciate that all or part of the process steps in the method for identifying internet credit fraud risks based on GPS tracks can be implemented by instructing related hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of various embodiments of the method for identifying internet credit fraud risks based on GPS tracks. Any reference to memory, storage, database, or other media used in the embodiments provided herein may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0060] This embodiment of the present application also provides an electronic device 500 that utilizes the aforementioned system for identifying internet credit fraud risks based on GPS trajectory. The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the aforementioned method for identifying internet credit fraud risks based on GPS trajectory are implemented. In this embodiment of the present application, the processor serves as the control center of the computer method and can be a processor of a physical machine or a processor of a virtual machine.
[0061] Reference Figure 2 The electronic device 500 includes at least one processor 501, at least one communication interface 502, at least one memory 503, and at least one bus 504. Bus 504 is used to enable communication between these components, communication interface 502 is used to communicate signaling or data with other node devices, and memory 503 stores machine-readable instructions executable by processor 501. When the electronic device 500 is running, processor 501 communicates with memory 503 via bus 504. When the machine-readable instructions are invoked by processor 501, the steps of the method for identifying internet credit fraud risks based on GPS trajectory are executed.
[0062] The above contents are merely embodiments of the present invention. Common knowledge such as the known specific structures and characteristics in the scheme is not described in detail here. A person of ordinary skill in the art is aware of all common technical knowledge in the technical field to which the invention belongs before the filing date or priority date, is able to obtain all existing technologies in the field, and has the ability to apply conventional experimental means before that date. A person of ordinary skill in the art can, under the guidance of this application, improve and implement this scheme in combination with his or her own abilities. Some typical known structures or known methods should not become an obstacle for a person of ordinary skill in the art to implement this application. It should be pointed out that for a person of ordinary skill in the art, several variations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the scope of protection of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection claimed in this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.
Claims
1. A method for identifying Internet credit fraud risks based on GPS trajectory, characterized in that: Collect data from the application stage, loan stage, and post-loan stage, including: When a borrower applies for a loan through an internet credit platform, the GPS data of the borrower's device is collected, along with the borrower's basic information, credit information, and device information; During the loan period from loan disbursement to repayment, GPS trajectory data of the borrower's device is continuously collected, and the collection frequency is dynamically adjusted based on risk assessment; During the repayment phase, GPS data is collected based on repayment progress and risk status; Clean the collected data, including removing duplicate data, filling missing values, and identifying and handling outliers; Processing of GPS coordinates, including coordinate correction and discretization; Extract GPS trajectory features and integrate them with the borrower's basic information, credit information, and device information; Use logistic regression, XGBoost, and Transformer models for training and evaluation, and select the best-performing model as the final fraud risk identification model; Conduct risk identification, including real-time credit risk interception, credit risk assessment during loan issuance, and post-loan fraud risk warning, and adopt corresponding strategies based on the risk identification results.
2. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Clean the collected data, including the following: Use the pandas library to process the collected GPS trajectory data, identify and delete duplicate records by judging the data's timestamp, coordinate value, and device ID information; For missing values in GPS trajectory data, if the missing time is short, the coordinates of the missing points are calculated using linear interpolation based on the coordinate values of the previous and next time points; if the missing time is long, the coordinates are estimated by combining the borrower's historical trajectory pattern and surrounding environment information; Anomalies in GPS trajectory data are identified by setting a threshold range. If a record shows that the movement distance exceeds this range and excluding special circumstances, the record is marked as an anomaly.
3. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 2, characterized in that: Processing of GPS coordinates, including the following: Coordinate correction uses commercial map API to correct GPS coordinates. An HTTP request is sent to the commercial map correction interface to correct the coordinates of each collected GPS track data one by one, and the corrected coordinates replace the original coordinates.
4. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 2, characterized in that: Processing of GPS coordinates also includes the following: Discretization processing divides the geographic area into fixed-size grids, determines the number of grid rows and columns based on the coordinate range, constructs a grid index, maps the corrected GPS track points to the corresponding grid cells, determines the grid to which the track point belongs based on its latitude and longitude coordinates, and records the track point's timestamp, grid number, and other related information to form discretized track data.
5. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Extract GPS trajectory features, Includes the following: Extract the dwell time feature and calculate the borrower's dwell time in each grid cell. By analyzing the timestamps of the trajectory points, if multiple consecutive trajectory points belong to the same grid cell, the sum of the time intervals between these points is calculated as the dwell time. Extract the moving speed feature and calculate the moving speed based on the coordinates and time intervals of adjacent trajectory points. The formula is: in, and are the coordinates of adjacent trajectory points, and is the corresponding timestamp; Extract the moving direction feature and calculate the moving direction between adjacent trajectory points; express the moving direction d by calculating the azimuth between two points. The formula is: 。 6. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Use logistic regression, XGBoost, and Transformer models for training and evaluation, and select the best-performing model as the final fraud risk identification model, including the following: Evaluation indicators include Accuracy, Recall and F1-Score, and the calculation formulas are: Among them, Precision is the accuracy rate, that is, the ratio of the number of correctly predicted positive samples to the number of predicted samples.
7. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Risk identification includes the following: Real-time credit risk interception is to input the processed real-time data into a trained fraud risk identification model. The model outputs the risk rating results and adopts corresponding strategies based on the risk rating results.
8. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Risk identification includes the following: The credit risk assessment during loan period is to regularly collect GPS track data of existing loan customers during loan period and conduct correlation analysis with pre-loan data. Based on the historical track analysis results and the risk level assessed by the model, different credit limits and differentiated interest rates are adapted. Calculate the similarity between the current trajectory and the pre-loan trajectory using the following formula: in, and For two trajectories, For trajectory points and The distance between is the path weight; the higher the trajectory similarity score, the more stable the trajectory and the lower the risk.
9. The method for identifying Internet credit fraud risk based on GPS trajectory according to claim 1, characterized in that: Risk identification includes the following: Post-loan fraud risk warning closely monitors the borrower's repayment behavior and GPS trajectory changes. If the borrower shows signs of overdue repayment and his GPS trajectory shows abnormal behavior, the system triggers a risk warning and takes differentiated early collection intervention measures based on the risk level.
10. An electronic device, characterized in that: The method for identifying Internet credit fraud risks based on GPS tracks according to any one of claims 1 to 9 is used, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor.