Multi-dimensional police-people joint defense fraud data analysis method and platform

Abstract

The invention discloses a multi-dimensional police and civilian joint defense fraud data analysis method and platform, and the method comprises the following steps: training an analysis model, constructing an automatic abnormality detection and early warning mechanism, and building a real-time monitoring and early warning system; carrying out regular fraud model display, and simulating the fraud model to carry out directional training test; obtaining an abnormal access behavior of sensitive data of the real-time monitoring and early warning system to obtain abnormal data; continuously adjusting analysis model parameters based on directional training test results and abnormal data; according to the method, firstly, the model is trained by analyzing the data of the fraud cases, test analysis is carried out, so that more effective public education and propaganda strategies are formulated, and public anti-fraud cognition and consciousness are accurately improved by regularly carrying out training test and accurate positioning training on the public; the development of anti-fraud propaganda is facilitated, the anti-fraud workload is reduced, and the propaganda effect and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of anti-fraud technology, and specifically to a multi-dimensional police-civilian joint anti-fraud data analysis method and platform. Background Art

[0002] Existing anti-fraud data analysis has certain defects in the actual process. It usually relies on a single or limited data source, and the information is incomplete, which may lead to one-sided analysis results. It is impossible to fully grasp the overall picture of fraud activities. It is inefficient, easily affected by human factors, and difficult to process large-scale data. The information is mainly released by the police and the public passively receives it. Targeted training and testing cannot be carried out according to the public's anti-fraud awareness. The efficiency and workload of anti-fraud knowledge promotion will be relatively large. Summary of the invention

[0003] The purpose of the present invention is to provide a multi-dimensional police-civilian joint anti-fraud data analysis method and platform to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a multi-dimensional police-civilian joint anti-fraud data analysis method, comprising the following steps: S1: Obtain multidimensional data and integrate and process it to obtain the data to be analyzed; S2: Collect historical case data information to train analysis models, build automatic anomaly detection and early warning mechanisms, and establish real-time monitoring and early warning systems; S3: Regular fraud model presentation based on real-time monitoring and early warning system, and simulated fraud models for targeted training and testing; S4: Obtain abnormal access behaviors of sensitive data of real-time monitoring and early warning systems, formulate extraction mechanisms, extract data related to abnormal access behaviors, and obtain abnormal data; S5: Continuously adjust the analysis model parameters based on the results of targeted training tests and abnormal data, improve the algorithm performance, and regularly train the analysis model to ensure that the analysis model can adapt to the ever-changing fraud situation; S6: Input the data to be analyzed into the analysis model to predict fraud risks.

[0005] Preferably, obtaining multidimensional data and integrating and processing to obtain data to be analyzed includes: S11: Acquire multi-dimensional data, the multi-dimensional data including communication records, financial transaction data, social media information, user behavior data, and a database of illegal numbers of the communication management department; S12: Connect multi-dimensional data from various sources to the data integration platform through API interface, data export, and crawler program; S13: Clean the multidimensional data to remove duplicate records, correct errors, handle missing values ​​and outliers, convert the data into a unified format, and load the cleaned and converted data into the database as the data to be analyzed.

[0006] Preferably, historical case data information is collected to train the analysis model, and an automatic detection and warning mechanism is constructed to establish a real-time monitoring and warning system, including: S21: Collect historical case information, analyze the data through statistical analysis and visualization technology, and extract features that are helpful in identifying fraudulent behavior from the original data, including statistical features, text features, and network structure features; S22: Build an analytical model based on a deep learning model, use historical data to train the analytical model, and perform model tuning through cross-validation and grid search methods; S23: Use the test data set to evaluate the model performance, calculate the evaluation indicators of accuracy, recall rate, and F1 score, test the model performance, cyclically train the analysis model, and improve the accuracy of model analysis; S24: Build an automatic anomaly detection and early warning mechanism, and establish a real-time monitoring and early warning system in combination with the analysis model to achieve real-time or batch fraud risk analysis and prediction.

[0007] Preferably, the automatic abnormality detection and early warning mechanism includes automatic detection of calls and fund account change information of associated users. When abnormalities occur, the early warning mechanism is automatically triggered and the police are directly contacted.

[0008] Preferably, orientation training tests include: S31: Construct a network fraud simulation test module, divide the test into three levels: elementary, intermediate, and advanced, and match the appropriate level for testing according to the actual situation of the user; S32: Classify test takers according to test results and provide targeted training; S33: Record the user's test results and training status, conduct cyclical testing and training, and enhance the user's anti-fraud awareness.

[0009] Preferably, the primary test is a test question generated based on historical fraud cases; the intermediate test is to design various types of simulated fraud links based on the latest fraud methods, and regularly send different types of simulated fraud links to users; the advanced test is embedded training, embedding fraud links in daily emails, text messages, and APPs to test the public's anti-fraud capabilities.

[0010] Preferably, the user's actual situation is used to form a user-defined personal portrait based on the user's age, occupation, and educational level information, providing the user with tailored educational content for preventing telecommunications and Internet fraud.

[0011] Preferably, the test results are scored in percentage, with excellent: 80-100, good: 60-79, and poor: 0-59. The user scores are classified based on the K-means clustering algorithm to facilitate subsequent targeted training, including the following steps: S321: Select K initial cluster centers and assign each data point to the cluster with the closest cluster center; S322: recalculating the cluster center of each cluster, that is, the mean of all data points in the cluster, and repeating the allocation and updating steps until the cluster center no longer changes significantly or reaches a predetermined number of iterations; S323: When the change of the cluster center is less than a preset threshold, the algorithm terminates and outputs the final clustering result.

[0012] Preferably, the formula for minimizing the sum of squared distances from all data points to the center of the cluster to which they belong in the K-means clustering algorithm is: ,in, is the qth data point, It is a data point The center of the cluster to which it belongs, m is the total number of data points, and the objective function is iteratively optimized to make each data point as close as possible to the center of the cluster to which it belongs, thereby achieving clustering.

[0013] According to another aspect of the present invention, a multi-dimensional police-civilian joint fraud prevention data analysis platform is provided, which is used for use of any of the above-mentioned multi-dimensional police-civilian joint fraud prevention data analysis methods, including: Data analysis module, used to obtain multi-dimensional data and integrate and process to obtain data to be analyzed; Model training module, which is used to collect historical case data information to train analysis models, build automatic abnormality detection and early warning mechanisms, and establish real-time monitoring and early warning systems; Real-time monitoring and early warning system for regular fraud model presentations and targeted training and testing of simulated fraud models; Abnormal data acquisition module, used to obtain abnormal access behaviors of sensitive data of real-time monitoring and early warning systems, formulate extraction mechanisms, extract data related to abnormal access behaviors, and obtain abnormal data; Model optimization training module, which is used to continuously adjust the analysis model parameters based on the results of targeted training tests and abnormal data, improve the algorithm performance, and regularly train the analysis model to ensure that the analysis model can adapt to the ever-changing fraud situation; Prediction module: used to input the data to be analyzed into the analysis model for fraud risk prediction.

[0014] Compared with the prior art, the present invention has the following beneficial effects: (1) By integrating and analyzing data from different sources, such as communication network data, Internet data, and financial transaction data, patterns and trends of fraudulent behavior can be identified earlier. By analyzing the data of fraud cases, we can better understand the changes and characteristics of fraud methods, thereby formulating more effective public education and publicity strategies. Dimensional data analysis requires data sharing and technical collaboration among different departments, which helps to break down information silos and promote cross-departmental cooperation. (2) Through regular training and testing of the public, they can be classified according to their anti-fraud awareness, and then the public can be classified and focused training can be provided to those with low anti-fraud awareness, so as to improve their anti-fraud knowledge and awareness, improve the effectiveness of training, and better carry out anti-fraud work. The training and publicity focus can be placed on those with relatively low awareness, and they can be accurately positioned. For those with high anti-fraud awareness, new fraud methods can be regularly informed, which is conducive to the development of anti-fraud publicity, reducing workload, and improving the effectiveness and efficiency of publicity. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flowchart of the steps of a multi-dimensional police-civilian joint anti-fraud data analysis method proposed by the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] See also Figure 1 , an embodiment provided by the present invention: a multi-dimensional police-civilian joint anti-fraud data analysis method, comprising the following steps: S1: Obtain multidimensional data and integrate and process it to obtain the data to be analyzed, including: S11: Acquire multi-dimensional data, the multi-dimensional data including communication records, financial transaction data, social media information, user behavior data, and a database of illegal numbers of the communication management department; S12: Connect multi-dimensional data from various sources to the data integration platform through API interface, data export, and crawler program; S13: Clean the multidimensional data to remove duplicate records, correct errors, handle missing values ​​and outliers, convert the data into a unified format, and load the cleaned and converted data into the database as the data to be analyzed.

[0018] S2: Collect historical case information to train and analyze models, and build an automatic anomaly detection and early warning mechanism, and establish a real-time monitoring and early warning system. The automatic anomaly detection and early warning mechanism includes automatic detection of related users' calls and financial account change information. When an anomaly occurs, the early warning mechanism will be automatically triggered, directly contacting the police, notifying the user and the police in his jurisdiction to take timely measures to block fraudulent activities.

[0019] S3: Conduct regular fraud model demonstrations based on the real-time monitoring and early warning system, and simulate fraud models for targeted training and testing. Use the early warning platform to regularly release the latest fraud methods and prevention guidelines to the public. Regularly hold online and offline lectures to popularize the latest anti-fraud skills, increase public vigilance, encourage citizens to actively report suspicious situations, set up convenient and easy-to-use online reporting channels, simulate fraud processes for training, carry out various publicity activities, and enhance the whole society's awareness and support for anti-fraud work.

[0020] S4: Obtain abnormal access behaviors of sensitive data of real-time monitoring and early warning systems, formulate extraction mechanisms, extract data related to abnormal access behaviors, and obtain abnormal data. The definition of abnormal access behaviors includes that the public may visit based on their own experiences of being deceived, and fraudsters may repeatedly study fraud models to exploit loopholes and develop new fraud methods. Therefore, statistical analysis should be conducted on abnormal accesses to predict the intentions of abnormal accesses and adjust the models according to possible situations.

[0021] S5: Continuously adjust model parameters based on actual application results, improve algorithm performance, collect user feedback through questionnaires, etc., to ensure that the entire system can adapt to the ever-changing fraud situation.

[0022] S6: Input the data to be analyzed into the analysis model to predict fraud risks.

[0023] Furthermore, multidimensional data can also be collected from public channels such as community reports, social media, and news reports. This information can provide first-hand information about the latest fraud trends, obtain information such as phone records, text message content, and geographic location, which can be used to analyze call patterns and abnormal behaviors, monitor financial activities such as bank transfers and credit card usage, and promptly detect suspicious fund flows.

[0024] Furthermore, historical case data and information are collected to train the analysis model, and an automatic detection and warning mechanism is built to establish a real-time monitoring and warning system, including: S21: Collect historical case information, analyze the data through statistical analysis and visualization technology, and extract features that are helpful in identifying fraudulent behavior from the original data, including statistical features, text features, and network structure features; S22: Build an analytical model based on a deep learning model, use historical data to train the analytical model, and perform model tuning through cross-validation and grid search methods; S23: Use the test data set to evaluate the model performance, calculate the evaluation indicators of accuracy, recall rate, and F1 score, test the model performance, cyclically train the analysis model, and improve the accuracy of model analysis; S24: Build an automatic anomaly detection and early warning mechanism, and establish a real-time monitoring and early warning system in combination with the analysis model to achieve real-time or batch fraud risk analysis and prediction.

[0025] It should be noted that the deep learning model needs to select appropriate machine learning models, such as BERT+BiLCNN, LSTM+Attention, XGBoost, SVM, Random Forest, etc., and select the most suitable model according to different fraud types and data characteristics.

[0026] Use historical data to train the model, adjust model parameters for optimization, use cross-validation, ROC curve, AUC value and other methods to evaluate model performance, ensure the generalization ability of the model on different data sets, and if the model has overfitting or underfitting problems, it can be optimized by increasing the amount of data, adjusting the model complexity, using regularization techniques, etc.

[0027] In response to the ever-changing fraud methods, incremental learning methods such as data replay and semi-supervised learning are adopted to maintain the model's ability to identify new and old fraud methods. The model is updated regularly and retrained using the latest fraud case data to ensure the timeliness and accuracy of the model.

[0028] Extracting useful features from the raw data, such as call duration, frequency, time point, number of contacts, etc., and using advanced data analysis such as K-means or DBSCAN can help identify groups with similar behavior patterns, which may indicate the existence of fraud organizations. By modeling events within a specific time period, understanding the trend of fraud activities over time, and establishing a network of relationships between entities such as people, equipment, and accounts, it helps to reveal complex fraud chains.

[0029] Furthermore, regular fraud model displays are conducted based on the real-time monitoring and early warning system, including displaying real fraud methods and their consequences through live broadcasts, text messages, posters, etc., and based on cooperation with APP manufacturers, fraud methods are displayed when the APP is opened, that is, anti-fraud publicity can be carried out before entering the APP, which can be in the form of posters.

[0030] Further, the orientation training test includes: S31: Construct a network fraud simulation test module, divide the test into three levels: elementary, intermediate, and advanced, and match the appropriate level for testing according to the actual situation of the user; S32: Classify test takers according to test results and provide targeted training; S33: Record the user's test results and training status, conduct cyclical testing and training, and enhance the user's anti-fraud awareness.

[0031] For S31, it can be specifically: S311: Determine the target user groups, such as college students and the general public, analyze the level of awareness and acceptance of online fraud by different user groups, and determine the difficulty grading standard of the test; S312: Use WeChat developer tools or other development platforms to design and develop a network fraud simulation test system, and design knowledge base cases, including primary, intermediate, and advanced fraud cases, to ensure the diversity and authenticity of the cases; S313: Implement the real-time push function of knowledge base cases, push corresponding fraud cases according to the level selected by the user, and implement various fraud simulation form testing functions, such as telephone fraud, SMS fraud, network fraud, etc.

[0032] For S32, it can be specifically: S321: Collect the user's performance data in the simulation test, including the accuracy of answering questions, reaction time, etc., and use machine learning algorithms to analyze the data to identify the user's level of fraud awareness and susceptibility; S322: Based on the analysis results, users are divided into three levels: primary, intermediate and advanced, and classification standards are designed to ensure the accuracy and scientificity of the classification; S323: Based on the user classification results, formulate targeted education strategies and design educational content and forms of different levels, such as focusing on explaining basic concepts for beginners, adding case analysis for intermediate users, and conducting simulated practical exercises for advanced users.

[0033] Furthermore, for the Internet fraud simulation test module, the training test is divided into three levels: elementary, intermediate and advanced. The elementary test is a test question edited in combination with historical fraud cases. The fraud cases can be edited and selected for test questions or displayed in the form of patterns in a similar way to an exam, which is convenient for training assessment tests; the intermediate test is to design various types of simulated fraud links based on the latest fraud methods, and regularly send different types of simulated fraud links to registered users. During the test, the public can be informed of the hedgehog simulated fraud to facilitate user experience; the advanced test is an embedded training system, which embeds fraud links in daily emails, text messages, and APPs to test the public's anti-fraud capabilities. The front-end links, personal information filling, etc. are the same as the real fraud steps. According to the real simulation, the anti-fraud performance is improved, and when the tester implements the bank card addition, the test is terminated to remind the tester to ensure security. This advanced test is suitable for people who have passed the intermediate test.

[0034] Furthermore, for training and testing, existing social platform tools such as WeChat applets can be used to develop an Internet fraud simulation test system. This system can push knowledge base cases in real time and enhance users' awareness of prevention through various fraud simulation tests. By designing and implementing game-based learning systems, such as "Anti-Phishing Phil", users' ability to identify fraudulent websites can be improved, and user-defined personal portraits can be formed based on user information such as age, occupation, and educational level, providing users with customized educational content to prevent telecommunications network fraud.

[0035] For the online fraud simulation test module, it can be based on the development of WeChat applet: using WeChat developer tools, develop a WeChat applet for online fraud simulation testing for colleges and universities, which realizes multiple functions such as real-time push of knowledge base cases and testing of various fraud simulation forms.

[0036] For the fraud model display, the real-time push function of the knowledge base case is implemented to help users learn and identify online fraud, and a variety of fraud simulation tests are provided to enable users to experience different types of online fraud. Through the simulated fraud 2D paper man game, users can deeply experience the various forms of network and communication fraud, and use reverse cases to warn the general public.

[0037] At the same time, the front-end users of the system can consult customer service and register, log in and post comments on the system's own small forum for learning and communication, provide personalized retrieval functions, provide users with a certain degree of early warning information, and mine the fraudulent behavior process contained in the big data of online fraud through statistical analysis.

[0038] Furthermore, the test results are scored in percentage, with excellent: 80-100, good: 60-79, and poor: 0-59. The specific classification can be set according to the actual situation. The crowd can be classified according to anti-fraud knowledge, and then those with relatively low cognition can be trained to improve their anti-fraud ability. For those with relatively high anti-fraud awareness, new fraud methods can be regularly informed to ensure that anti-fraud knowledge is updated in real time. After training, people with relatively low anti-fraud awareness can be re-trained and tested regularly to improve their anti-fraud awareness. Precise and targeted training can be carried out to improve the effect of anti-fraud publicity and reduce the workload of training. The scored public is classified based on the K-means clustering algorithm to facilitate subsequent targeted training, including the following steps: S321: Select K initial cluster centers and assign each data point to the cluster with the closest cluster center; S322: recalculating the cluster center of each cluster, that is, the mean of all data points in the cluster, and repeating the allocation and updating steps until the cluster center no longer changes significantly or reaches a predetermined number of iterations; S323: When the change of the cluster center is less than a preset threshold, the algorithm terminates and outputs the final clustering result.

[0039] Furthermore, the formula for minimizing the sum of squared distances from all data points to the center of the cluster to which they belong in the K-means clustering algorithm is: ,in, is the qth data point, It is a data point The center of the cluster to which it belongs, m is the total number of data points, and the objective function is iteratively optimized to make each data point as close as possible to the center of the cluster to which it belongs, thereby achieving clustering.

[0040] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A multi-dimensional police-civilian joint fraud prevention data analysis method, Features: The following steps are involved: S1: Obtain multidimensional data and integrate and process it to obtain the data to be analyzed; S2: Collect historical case data information to train analysis models, build automatic anomaly detection and early warning mechanisms, and establish real-time monitoring and early warning systems; S3: Regular fraud model presentation based on real-time monitoring and early warning system, and simulated fraud models for targeted training and testing; S4: Obtain abnormal access behaviors of sensitive data of real-time monitoring and early warning systems, formulate extraction mechanisms, extract data related to abnormal access behaviors, and obtain abnormal data; S5: Continuously adjust the analysis model parameters based on the results of targeted training tests and abnormal data, improve the algorithm performance, and regularly train the analysis model to ensure that the analysis model can adapt to the ever-changing fraud situation; S6: Input the data to be analyzed into the analysis model to predict fraud risks.

2. According to the multi-dimensional police-civilian joint fraud prevention data analysis method of claim 1, Features: Acquiring multidimensional data and integrating and processing the data to be analyzed includes: S11: Acquire multi-dimensional data, the multi-dimensional data including communication records, financial transaction data, social media information, user behavior data, and a database of illegal numbers of the communication management department; S12: Connect multi-dimensional data from various sources to the data integration platform through API interface, data export, and crawler program; S13: Clean the multidimensional data to remove duplicate records, correct errors, handle missing values ​​and outliers, convert the data into a unified format, and load the cleaned and converted data into the database as the data to be analyzed.

3. According to the multi-dimensional police-civilian joint fraud prevention data analysis method of claim 1, Features: Collect historical case data information to train analysis models, build automatic detection of anomalies and early warning mechanisms, and establish real-time monitoring and early warning systems including: S21: Collect historical case information, analyze the data through statistical analysis and visualization technology, and extract features that are helpful in identifying fraudulent behavior from the original data, including statistical features, text features, and network structure features; S22: Build an analytical model based on a deep learning model, use historical data to train the analytical model, and perform model tuning through cross-validation and grid search methods; S23: Use the test data set to evaluate the model performance, calculate the evaluation indicators of accuracy, recall rate, and F1 score, test the model performance, cyclically train the analysis model, and improve the accuracy of model analysis; S24: Build an automatic anomaly detection and early warning mechanism, and establish a real-time monitoring and early warning system in combination with the analysis model to achieve real-time or batch fraud risk analysis and prediction.

4. According to the multi-dimensional police-civilian joint fraud prevention data analysis method of claim 1, Features: The automatic anomaly detection and early warning mechanism includes automatic detection of related users' calls and financial account change information. When an anomaly occurs, the early warning mechanism will be automatically triggered and the police will be directly contacted.

5. According to the multi-dimensional police-civilian joint anti-fraud data analysis method of claim 1, Features: Orientation training tests include: S31: Construct a network fraud simulation test module, divide the test into three levels: elementary, intermediate, and advanced, and match the appropriate level for testing according to the actual situation of the user; S32: Classify test takers according to test results and provide targeted training; S33: Record the user's test results and training status, conduct cyclical testing and training, and enhance the user's anti-fraud awareness.

6. According to the multi-dimensional police-civilian joint anti-fraud data analysis method of claim 5, Features: The primary test is a test question generated based on historical fraud cases; the intermediate test is to design various types of simulated fraud links based on the latest fraud methods, and regularly send different types of simulated fraud links to users; the advanced test is embedded training, embedding fraud links in daily emails, text messages, and APPs to test the public's anti-fraud capabilities.

7. According to claim 5, a multi-dimensional police-civilian joint fraud prevention data analysis method, Features: The user's actual situation is used to form a customized personal portrait based on the user's age, occupation, and educational level, providing the user with tailored educational content to prevent telecommunications and Internet fraud.

8. According to claim 5, a multi-dimensional police-civilian joint fraud prevention data analysis method, Features: The test results are scored in percentage, with excellent: 80-100, good: 60-79, and poor: 0-59. The user ratings are classified based on the K-means clustering algorithm to facilitate subsequent targeted training, including the following steps: S321: Select K initial cluster centers and assign each data point to the cluster where the nearest cluster center is located; S322: recalculating the cluster center of each cluster, that is, the mean of all data points in the cluster, and repeating the allocation and updating steps until the cluster center no longer changes significantly or reaches a predetermined number of iterations; S323: When the change of the cluster center is less than a preset threshold, the algorithm terminates and outputs the final clustering result.

9. A multi-dimensional police-civilian joint fraud prevention data analysis method according to claim 8, Features: The formula for minimizing the sum of the squares of the distances from all data points to the centers of their clusters in the K-means clustering algorithm is: ,in, is the qth data point, It is a data point The center of the cluster to which it belongs, m is the total number of data points, and the objective function is iteratively optimized to make each data point as close as possible to the center of the cluster to which it belongs, thereby achieving clustering.

10. A multi-dimensional police-civilian joint anti-fraud data analysis platform, used to implement a multi-dimensional police-civilian joint anti-fraud data analysis method as described in any one of claims 1-9, It is characterized in that include: Data analysis module, used to obtain multi-dimensional data and integrate and process to obtain data to be analyzed; Model training module, which is used to collect historical case data information to train analysis models, build automatic abnormality detection and early warning mechanisms, and establish a real-time monitoring and early warning system; Real-time monitoring and early warning system for regular fraud model presentations and targeted training and testing of simulated fraud models; Abnormal data acquisition module, used to obtain abnormal access behavior of sensitive data of real-time monitoring and early warning system, formulate extraction mechanism, extract data related to abnormal access behavior, and obtain abnormal data; Model optimization training module, which is used to continuously adjust the analysis model parameters based on the results of targeted training tests and abnormal data, improve the algorithm performance, and regularly train the analysis model to ensure that the analysis model can adapt to the ever-changing fraud situation; Prediction module: used to input the data to be analyzed into the analysis model for fraud risk prediction.

Images