A risk assessment system, method, computer storage medium, and electronic device

Abstract

The application discloses a risk discrimination system, method, computer storage medium and electronic equipment. The system comprises a risk management platform and a risk discrimination platform. The risk management platform is used for receiving risk marking data fed back by the risk discrimination platform for marking transaction data, distributing the risk marking data as a risk discrimination task, initiating a risk discrimination request to the risk discrimination platform, and obtaining a discrimination audit conclusion fed back by the risk discrimination platform. The risk discrimination platform is used for marking transaction data, determining risk marking data, determining risk marking data and transaction data corresponding to the risk marking data as to-be-discriminated data in response to a risk discrimination request, discriminating the to-be-discriminated data by using a multi-task risk automatic discrimination model trained according to risk types, determining a discrimination audit conclusion of the to-be-discriminated data, and sending the discrimination audit conclusion to the risk management platform, so that the accuracy of risk discrimination can be ensured.

Description

Technical Field

[0001] This application relates to the field of information security technology, specifically to a risk assessment system, as well as a method and apparatus for such a system. This application also relates to a risk data output method and apparatus, a computer storage medium, and an electronic device. Background Technology

[0002] E-commerce (electronic commerce) refers to business activities that use information network technology to facilitate the exchange of goods. It can also be understood as transactions and related services conducted electronically using technologies such as the Internet and computers. It represents another form of digitalization, networking, and informatization of various aspects of traditional business activities. Electronic transactions include online transactions between consumers and merchants, online electronic payments, and various business, trading, financial, and related integrated service activities. It is a new business operation model. Nowadays, transactions conducted via the Internet have become a routine activity in daily life.

[0003] E-commerce platforms (transaction software platforms) serve as a bridge connecting buyers and sellers, greatly promoting the convenience and efficiency of transactions, but also bringing various risks and challenges.

[0004] First, the risks to cybersecurity cannot be ignored. E-commerce transactions involve the transmission of a large amount of sensitive information, such as personal identification information, bank account details, and transaction data. If platform security measures are inadequate, they are highly vulnerable to security threats such as hacker attacks and data breaches, which not only damage user privacy but may also lead to financial losses and a crisis of trust.

[0005] Secondly, transaction fraud is another major hidden danger. The anonymity and remote transaction characteristics of e-commerce platforms provide opportunities for criminals to commit fraudulent acts such as providing false product information, phishing websites, and malicious refund requests. These can cause economic losses to consumers and merchants and damage the platform's reputation.

[0006] Furthermore, legal compliance risks are also a crucial factor to consider. Different countries and regions have varying laws and regulations regarding e-commerce, including those concerning consumer rights protection, data protection, and taxation. Platforms that fail to comply with relevant laws and regulations may face legal action, fines, or even business bans.

[0007] Therefore, paying attention to, identifying, and assessing risks is crucial for the healthy operation of e-commerce software platforms. Summary of the Invention

[0008] This application provides a risk assessment system to solve the problems of poor accuracy and high cost caused by relying on manual risk review in the prior art.

[0009] This application provides a risk assessment system, characterized in that it includes: a risk management platform and a risk assessment platform;

[0010] The risk management platform is used to receive risk labeling data for risk labeling of transaction data fed back by the risk identification platform, and to distribute the risk labeling data as a risk identification task, initiate a risk identification request to the risk identification platform, and obtain the identification and review conclusion corresponding to the risk identification request fed back by the risk identification platform.

[0011] The risk assessment platform is used to perform risk labeling on the transaction data to determine the risk-labeled data of the transaction data; and to respond to the risk assessment request initiated by the risk management platform for the risk assessment task, to determine the risk-labeled data in the risk assessment request and the transaction data corresponding to the risk-labeled data as data to be assessed; to perform risk assessment on the data to be assessed using a multi-task automatic risk assessment model trained separately according to risk type, to determine the assessment review conclusion of the data to be assessed, and to send the assessment review conclusion to the risk management platform.

[0012] In some embodiments, the risk assessment platform is used to acquire the transaction data through a first acquisition link and / or a second acquisition link according to the set automatic tagging trigger conditions; and to perform risk tagging on the transaction data according to the constructed risk assessment rule set to determine the risk tagging data corresponding to the transaction data.

[0013] In some embodiments, the risk discrimination platform is used to train a first classification sub-model according to a first risk type to determine a target first classification sub-model; to train a second classification sub-model according to a second risk type to determine a target second classification sub-model; and to train a third classification sub-model according to a third risk type to determine a target third classification sub-model.

[0014] By inputting the data to be judged into a multi-task risk automatic judgment model fitted according to the first classification sub-model, the second classification sub-model and the third classification sub-model, the judgment and review conclusion of the data to be judged is determined.

[0015] In some embodiments, the risk management platform further includes:

[0016] Based on the obtained discrimination and review conclusions, the system filters target discrimination and review conclusions that meet the discrimination requirements, and pushes the target discrimination and review conclusions, along with the transaction data corresponding to the target discrimination and review conclusions, as progress data to the data processing stage to perform corresponding processing operations.

[0017] In some embodiments, the risk management platform is configured to: determine the data features of the risk tagging data extracted from the data to be judged as a first feature; determine the selection features of the transaction data extracted from the data to be judged as a second feature; determine the statistical features of the selected features of the transaction data extracted from the data to be judged as a third feature; and determine the risk profile data features of the transaction data extracted from the data to be judged as a fourth feature.

[0018] One or more of the first feature, second feature, third feature, and fourth feature are input into the automatic discrimination model to determine the risk type, and the risk type discrimination result is determined as the discrimination and review conclusion of the data to be judged.

[0019] This application also provides a risk assessment method, including:

[0020] In response to a risk assessment request initiated for a risk assessment task, the risk labeling data in the risk assessment request and the transaction data corresponding to the risk labeling data are determined as data to be assessed.

[0021] Based on a multi-task automatic risk discrimination model trained separately according to risk type, risk discrimination is performed on the data to be discriminated, and the discrimination and review conclusion of the data to be discriminated is determined.

[0022] Send the aforementioned judgment and review conclusion.

[0023] In some embodiments, it also includes:

[0024] The acquired transaction data is risk-tagged to determine the risk-tagged data of the transaction data.

[0025] In some embodiments, the step of risk-tagging the acquired transaction data and determining the risk-tagging data of the transaction data includes:

[0026] According to the set automatic tagging trigger conditions, the transaction data is obtained through the first acquisition link and / or the second acquisition link;

[0027] The transaction data is risk-tagged according to the constructed risk assessment rule set to determine the risk-tagged data corresponding to the transaction data.

[0028] In some embodiments, the step of acquiring transaction data through a first acquisition link and / or a second acquisition link according to the set automatic tagging trigger conditions includes:

[0029] Based on the set selection criteria, the transaction data snapshots of the transaction participants that meet the risk assessment criteria are selected as the first acquisition link, and the transaction data is acquired when the automatic tagging trigger condition is met.

[0030] And / or,

[0031] Based on the set selection criteria, the integrated data of the transaction data of the selected transaction participants that meet the risk assessment criteria is used as the second acquisition link. When the automatic tagging trigger condition is met, the transaction data is acquired.

[0032] In some embodiments, the step of performing risk assessment on the data to be assessed based on a multi-task automatic risk assessment model trained separately according to risk type, and determining the assessment and review conclusion of the data to be assessed, includes:

[0033] The first classification sub-model is trained based on the first risk type to determine the target first classification sub-model;

[0034] The second classification sub-model is trained based on the second risk type to determine the target second classification sub-model;

[0035] The third classification sub-model is trained based on the third risk type to determine the target third classification sub-model;

[0036] The first target classification sub-model, the second target classification sub-model, and the third target classification sub-model are fitted into a multi-task risk automatic discrimination model;

[0037] The data to be judged is input into the multi-task risk automatic judgment model to determine the judgment and review conclusion of the data to be judged.

[0038] In some embodiments, the step of inputting the data to be judged into the multi-task risk automatic judgment model and determining the judgment and review conclusion of the data to be judged includes:

[0039] The data features of the risk labeling data extracted from the data to be judged are determined as the first feature;

[0040] The selected features of the transaction data in the extracted data to be judged are determined as the second feature;

[0041] The statistical features of the selected features of the transaction data in the extracted data to be judged are determined as the third feature;

[0042] The risk profile data features of the transaction data extracted from the data to be judged are determined as the fourth feature;

[0043] One or more of the first feature, second feature, third feature, and fourth feature are input into the multi-task risk automatic discrimination model to determine the risk type;

[0044] The risk type identification result is determined as the identification and review conclusion of the data to be identified.

[0045] In some embodiments, sending the judgment and review conclusion includes:

[0046] Based on the comparison between the judgment and audit conclusion and the set judgment threshold, the data to be judged that meets the advancement requirements is advanced to the final processing stage.

[0047] This application also provides a method for outputting risk data, including:

[0048] The received risk labeling data is distributed as a risk assessment task;

[0049] Based on the viewing request for the risk assessment task, a risk assessment request for the target transaction data is generated and sent, wherein the risk assessment request carries the target transaction data and the target risk labeling data corresponding to the target transaction data;

[0050] Receive the multi-task automatic risk discrimination model trained separately based on risk type, and output the discrimination and review conclusion corresponding to the target transaction data.

[0051] In some embodiments, the step of receiving the judgment and review conclusion corresponding to the target transaction data from a multi-task risk automatic discrimination model trained separately based on risk type includes:

[0052] The system receives the target transaction data and the target risk labeling data corresponding to the target transaction data as input data to the multi-task risk automatic judgment model, and outputs the judgment and review conclusion on the risk type of the target transaction data.

[0053] The judgment and review conclusion is stored and displayed on the viewing page corresponding to the viewing request.

[0054] In some embodiments, it also includes:

[0055] The area displaying the review conclusion on the viewing page displays a second prompt message corresponding to the review conclusion.

[0056] In some embodiments, it also includes:

[0057] Based on the received risk labeling data, add a first prompt message corresponding to the risk labeling data;

[0058] Output the risk labeling data and the first prompt information.

[0059] This application also provides a computer storage medium for storing data generated by a network platform, and a program for processing the data generated by the network platform.

[0060] When the program is read and executed, it performs the logical steps as described in the risk assessment system, or the steps as described in the risk assessment method, or the steps as described in the risk data output method.

[0061] This application also provides an electronic device, including:

[0062] processor;

[0063] The memory is used to store programs that process data generated by the network platform. When the program is read and executed by the processor, it performs the logical steps as performed in the risk assessment system described above, or the steps as performed in the risk assessment method described above, or the steps as performed in the risk data output method described above.

[0064] Compared with the prior art, this application has the following advantages:

[0065] This application provides a risk assessment system that uses a risk assessment platform to label acquired transaction data for risk, and sends the labeled risk-labeled data to a risk management platform. The risk management platform initiates a risk assessment request to the risk assessment platform based on a risk assessment task. This allows the system to perceive risk points in transaction data immediately upon initiating the risk assessment task through the risk-labeled data. The risk assessment platform identifies the risk-labeled data carried in the risk assessment request, along with the corresponding transaction data, as data to be assessed. Based on a multi-task automatic risk assessment model trained separately for each risk type, the system performs risk assessment on the data to be assessed, determines the assessment review conclusion, and sends the review conclusion to the risk management platform. This ensures the accuracy of risk assessment of transaction data and avoids omissions or misjudgments of risk points. Furthermore, it can advance to the final stage of risk assessment if the risk assessment conclusion meets the requirements, improving the efficiency of risk assessment. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the structure of a risk assessment system provided in this application;

[0067] Figure 2This is a timing diagram of an embodiment of a risk assessment system performing a assessment task provided in this application;

[0068] Figure 3 This is a schematic diagram illustrating an embodiment of the training process of a multi-task automatic risk discrimination model in a risk discrimination system provided in this application;

[0069] Figure 4 This is a schematic diagram of an automatic discrimination embodiment in a risk discrimination system provided in this application;

[0070] Figure 5 This is a flowchart of a risk assessment method provided in this application;

[0071] Figure 6 This is a schematic diagram of the structure of a risk assessment device provided in this application;

[0072] Figure 7 This is a flowchart of a risk data output method provided in this application;

[0073] Figure 8 This is a schematic diagram of the structure of a risk data output device provided in this application;

[0074] Figure 9 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0075] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0076] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The descriptive terms used in this application and the appended claims, such as "a," "first," and "second," are not intended to limit quantity or sequence, but rather to distinguish information of the same type from one another.

[0077] While significant progress has been made in online financial transaction risk detection, existing technologies still face a series of challenges and problems. First, the rapid pace of technological development is not keeping pace with financial innovation, making it difficult for risk detection methods to comprehensively cover all emerging risks. For example, with the application of emerging technologies such as blockchain and artificial intelligence, the methods and forms of online financial transactions are constantly evolving, and existing risk detection systems may not be able to adapt to these changes in a timely manner, resulting in detection blind spots. Second, data privacy and information security remain major challenges in risk detection. Online financial transactions involve a large amount of sensitive information, such as personal identification information and transaction records, which are vulnerable to threats such as hacking and internal leaks during transmission and storage. Although data encryption and firewalls are widely used, there is still a risk of them being cracked or bypassed, leading to frequent data breaches. Furthermore, the accuracy and timeliness of risk detection systems also face challenges. Due to the complexity and diversity of online financial transactions, risk detection systems need to process large amounts of data and identify potential risk signals. However, due to limitations in algorithm design and data processing capabilities, the accuracy and timeliness of these systems often fall short of ideal levels. This could lead to some genuine risk events being missed or delayed in detection, especially in processes where the accuracy of risk probabilities is determined through manual review based on potential risk signals. This could further result in missed detections, false positives, or misjudgments. Finally, updating and maintaining the risk detection system also presents challenges. With continuous technological advancements and market changes, risk detection systems require constant updates and optimizations to adapt to new risk types and trading patterns. However, updating and maintaining the system demands significant human, material, and financial resources and may face challenges related to technical compatibility and stability.

[0078] In summary, existing technologies for risk detection in online financial transactions still face numerous problems and challenges. To further improve the accuracy and timeliness of risk detection and enhance the system's intelligence and automation, this application provides a risk assessment system that can improve the accuracy of risk assessment while avoiding missed detections, false detections, or misjudgments through automated execution. It also increases the speed of risk assessment, improving work efficiency.

[0079] Therefore, it can be understood that the inventive concept of this application stems from the need for risk control in existing online transactions, especially in scenarios involving a large number of pre-purchase and post-payment transactions, where risk control becomes even more crucial. Before providing a detailed description of the risk assessment system provided in this application, the relevant technical terms involved in the embodiments provided in this application will be explained as follows:

[0080] Anti-fraud refers to the measures and methods employed in financial lending to prevent and combat various fraudulent activities. Fraud is defined as the act of obtaining loans, deceiving others, fraudulently claiming insurance payouts, misappropriating credit lines, and evading repayment obligations through deception, false statements, or other illegal means during the application, approval, disbursement, and repayment processes for loans or insurance.

[0081] Fraud prevention assessment (case adjudication): refers to the process of early warning, analysis, evaluation and handling of fraudulent activities, in order to promptly detect and respond to fraudulent activities, reduce fraud risks and protect the interests of enterprises and customers.

[0082] Anti-fraud analysis center: refers to a one-stop financial anti-fraud analysis platform that supports the entire lifecycle of anti-fraud cases, from case creation, case merging, case allocation, team queuing, case analysis and assessment, to case handling and closure.

[0083] Risk tagging refers to the process of identifying, classifying, and marking potential risk factors during risk management and control; this facilitates a clearer understanding and analysis of cases.

[0084] Transaction data refers to data generated from financial transactions conducted via the internet. It can involve buyers and sellers, as well as third parties, and includes data on completed payments and / or incomplete payments.

[0085] Purchase first, pay later: This is a scenario based on a trading platform where users acquire goods first and then make payment. The advance payment can be made by the trading platform or a third party, or it can be stipulated in the agreement that the goods must be acquired before payment.

[0086] The risks in this embodiment can be abnormal behaviors that occur during the transaction process. To make it easier to understand, this embodiment provides some examples of abnormal behaviors, such as fraud and cash-out, but it is not limited to these. The purpose is to ensure the security of transactions and maintain a good network environment.

[0087] like Figure 1 As shown, Figure 1 This is a schematic diagram of a risk assessment system provided in this application. The system includes a risk management platform 101 and a risk assessment platform 102.

[0088] The risk management platform 101 is used to receive risk labeling data for risk labeling of transaction data fed back by the risk identification platform, and to distribute the risk labeling data as a risk identification task, initiate a risk identification request to the risk identification platform, and obtain the identification and review conclusion corresponding to the risk identification request fed back by the risk identification platform.

[0089] The risk assessment platform 102 is used to perform risk labeling on the transaction data and determine the risk labeling data of the transaction data; in response to the risk assessment request initiated by the risk management platform for the risk assessment task, the risk labeling data carried in the risk assessment request and the transaction data corresponding to the risk labeling data are determined as data to be assessed; according to a multi-task automatic risk assessment model trained separately based on risk type, the data to be assessed is risk assessed, the assessment review conclusion of the data to be assessed is determined, and the assessment review conclusion is sent to the risk management platform.

[0090] like Figure 2 As shown, Figure 2This is a timing diagram illustrating the execution of a risk assessment task by a risk assessment system provided in this application. The risk assessment platform 102 needs to perform risk labeling on transaction data. In this embodiment, to avoid omissions, errors, and mislabeling that occur during manual labeling, and also to improve the efficiency of the labeling operation, an automatic risk labeling logic can be adopted. This process requires first acquiring transaction data. In this embodiment, the transaction data can be acquired in at least two ways: through a first acquisition link and / or through a second acquisition link. That is, through a transaction data snapshot and / or through the data source of the transaction data. Based on the acquired transaction data, the risk assessment platform is driven to perform risk labeling on the transaction data. In this embodiment, the risk tagging of the transaction data can be automatically driven according to the set automatic tagging trigger conditions. For example, T+15mins is the automatic tagging trigger condition, that is, the automatic tagging operation is driven by the transaction generation date (T: the risk review requirement generation date or the transaction date) + 15 minutes (wherein, the risk review requirement generation date can be the generation date of the risk review form generated for the transaction data, that is, not all transaction data is reviewed, but risk review forms are selectively generated. The specific selection method can be combined with different scenarios. Of course, risk review can also be performed on all transaction data. Therefore, T can also represent the transaction date, that is, the transaction generation date or the transaction occurrence date. Of course, the meaning of T can be determined according to different needs and is not limited to the example given in this embodiment). Of course, the automatic tagging trigger conditions can also be set according to the risk judgment scenario requirements or the transaction data amount, etc., and are not limited to the examples given above. In this embodiment, the first acquisition link is a transaction data snapshot. This snapshot can be selected based on set selection criteria, choosing transaction data snapshots from participants that meet risk assessment criteria. For example, selection criteria could be T+1D (the date the risk audit requirement is generated or the transaction date + 1 day), T+15mins, etc. Transaction data can be understood as order data, and the transaction data snapshot can be understood as an order data snapshot. The order data includes data of completed transactions and data of cancelled transactions. The second acquisition link is a transaction data data source. This source can be integrated data from participants that meet risk assessment criteria, selected based on set selection criteria. For example, it can integrate the transaction data of both buyers and sellers as the transaction data source. Updates to the data source can also follow set update criteria, such as T+15mins, thus ensuring the real-time nature of the transaction data in the data source. In this embodiment, both the transaction data snapshot and the data source can be stored in a database or data warehouse; the specific storage format and method are not limited. In this embodiment, the integration of the transaction data can be performed on the risk management platform, and the integrated data can be stored in a database, data warehouse, or server.

[0091] As described above, the risk assessment platform needs to perform risk labeling on transaction data. This embodiment uses automatic labeling. Therefore, the automatic labeling logic can be activated according to the set automatic labeling trigger conditions. These trigger conditions can be set based on specific risk assessment scenario requirements or according to general risk assessment requirements. This embodiment uses T+15mins as an example of the automatic labeling trigger condition, but it is not limited to this. When the automatic labeling trigger condition is triggered, transaction data can be acquired through the first acquisition link and / or the second acquisition link, and the automatic labeling logic can be executed on the transaction data. In this embodiment, the transaction data can be risk-labeled according to the constructed risk judgment rule set to determine the risk-labeled data corresponding to the transaction data. The risk judgment rule set can also be understood as a labeling strategy package, which can be determined through historically generated risk data or data with potential risks. For example, for buyers, the risk assessment rule could be to identify whether their shopping habits have changed abruptly, and to query the buyer's highest transaction change score in the last 15 days using a transaction change risk level algorithm model to determine if it falls into a risk category. If it does, the algorithm would output "Buyer shopping habit change = true". In other words, based on the shopping habit-related data characteristics identified in the buyer's transaction data, the algorithm model determines whether a risk exists; if so, a risk label is applied. Risk labeling data can include risk description information and risk values, which can include judgment values ​​such as yes, no, and uncertain. These values ​​are modifiable. Risk can also be determined by identifying the buyer's shipping address, contact information, and transaction IP address. Therefore, the risk assessment rule set for buyers can be constructed using any one or more pieces of information from transaction data such as buyer transaction behavior, buyer transaction attributes, and buyer profiles. Of course, different risk strategies can be adopted for different transaction items, such as setting targeted risk strategies based on the value and attributes of the transaction items. Therefore, the risk assessment rule set in this embodiment is only illustrated by the above example and is not limited to the above methods. For sellers, the risk assessment rules can identify whether the transaction data related to the seller includes high-risk categories, whether it is on a negative list, or whether it has been blacklisted, etc. These will not be listed exhaustively, and can also be detected through algorithmic models. The risk assessment rule set can be constructed based on relevant data such as transaction scenarios, transaction items, and transaction methods. The risk identification platform uses the risk assessment rule set to label the risk points in the acquired transaction data, determining the corresponding risk-labeled data for the transaction data.The risk labeling data is used to label the transaction data, so there can be a correlation between the two. For example, the corresponding transaction data can be found through the risk labeling data, or the risk labeling data of the transaction data can be obtained through the transaction data.

[0092] To facilitate intuitive capture of risk points and reduce processing operations in scenarios requiring table lookup, statistics, and analysis, in this embodiment, the risk labeling data may further include: a first prompt information (icon). In this embodiment, the first prompt information may be a "recommended" marker displayed in the risk judgment value area, which can serve as recommended data for subsequent judgment.

[0093] like Figure 2 As shown, after the risk identification platform completes the labeling, it sends the risk labeling data to the risk management platform 101. The risk management platform then distributes the risk labeling data to the risk identification task for subsequent execution. In this embodiment, the execution of the risk identification task can be achieved by the risk management platform initiating a risk identification request to the risk identification platform. This risk identification request carries the acquired risk labeling data and the corresponding transaction data. The risk identification request can be generated by the risk management platform based on operation requests such as viewing the risk identification task. When a risk identification request is generated, the risk management platform can acquire the risk labeling data and the corresponding transaction data in real time and send them together to the risk identification platform. The risk assessment platform determines the received risk-labeled data and the corresponding transaction data as the data to be assessed and executes automatic risk assessment logic. Specifically, it uses a multi-task automatic risk assessment model trained separately for each risk type to assess the risk in the data to be assessed, determines the assessment conclusion, and sends the assessment conclusion to the risk management platform. In this embodiment, the multi-task automatic risk assessment model can be a sub-model trained separately for different risk types, and the multi-task automatic risk assessment model is obtained by fitting multiple sub-models together.

[0094] like Figure 3 As shown, Figure 3This is a schematic diagram illustrating an embodiment of the training process of a multi-task automatic risk discrimination model in a risk discrimination system provided in this application. In this embodiment, the risk type targeted by the model training can be determined based on the specific risk discrimination scenario. This may include a first risk type, a second risk type, a third risk type, etc., and the three risk types are trained separately. The model can be a binary classification model, such as a decision tree model. Specifically, the training may involve training a first classification sub-model based on the first risk type to determine a target first classification sub-model; training a second classification sub-model based on the second risk type to determine a target second classification sub-model; and training a third classification sub-model based on the third risk type to determine a target third classification sub-model. The target first classification sub-model, the target second classification sub-model, and the target third classification sub-model are then fitted into a multi-task automatic risk discrimination model. For different risk types, the corresponding weights of the sub-models can be set according to the risk discrimination scenario and requirements. The training data for the sub-models can be historical data or existing datasets. In this embodiment, the first, second, and third risk types can be types of risk (whether it is a risk, whether it is money laundering risk, and whether it is fraud risk), and can also include other risk types, such as: fraudulent transaction risk, malicious buyer / seller risk, credit risk, etc. There can be multiple different risk types for different risk assessment scenarios, and the determination of specific risk types is not limited to the above examples. In a transaction scenario, a risk type can be understood as a potential abnormal transaction data. This abnormal transaction data can be abnormal data existing after the transaction participants have completed the relevant transaction agreement, abnormal data existing during the transaction process, or abnormal data appearing in a particular participant. In this embodiment, risk assessment can be performed on abnormal data existing in the "buy first, pay later" transaction process. Transaction data can be transaction data for a specific transaction type, or it can be all transaction data, specifically determined according to the risk assessment requirements; there is no specific limitation in this embodiment. The transaction participants include the buyer and seller, and can also include third parties. Similarly, there is no specific limitation on the transaction participants; anyone related to the transaction behavior is considered a transaction participant.

[0095] like Figure 4 As shown, Figure 4This is a schematic diagram of an automatic risk assessment embodiment provided in this application. The data to be assessed is input into a trained multi-task automatic risk assessment model to determine the output result, i.e., the assessment and review conclusion. Specifically, the data features of the risk-labeled data extracted from the data to be assessed are determined as the first feature; the selection features of the transaction data extracted from the data to be assessed are determined as the second feature; the statistical features of the selected features of the transaction data extracted from the data to be assessed are determined as the third feature; the risk profile data features of the transaction data extracted from the data to be assessed are determined as the fourth feature; one or more of the first, second, third, and fourth features are input into the multi-task automatic assessment model for risk type assessment; and the risk type assessment result is determined as the assessment and review conclusion of the data to be assessed. The data features of the risk-labeled data include, for example, easily cashed-out goods, sudden changes in shopping habits, etc. The selected features can refer to the transaction data currently undergoing risk assessment, and the selected features for risk assessment include, for example, small-amount split orders, abnormal payment duration, etc. The statistical features of the selected features can refer to the proportion of risky transactions over the past month, the number of successful risk assessments, etc. The risk profile data features of the transaction data can be historical risk profile data, such as fraud score, stability score, credit score, etc. The extracted features are input into a multi-task automatic risk assessment model to determine the input results corresponding to the risk type of the transaction data, and the risk assessment review conclusion. For example: if the multi-task automatic risk assessment model outputs has_risk = 0, it recommends no risk, and the corresponding risk assessment review conclusion is "no risk"; if the multi-task automatic risk assessment model outputs has_risk = 1 & risk_detail = CASH OUT, it recommends risk, and the risk type is cash-out risk, and the corresponding risk assessment review conclusion is "confirmed cash-out"; if the multi-task automatic risk assessment model outputs has_risk = 1 & risk_detail = NOTSURE, it recommends risk, and the risk type is uncertain; if the multi-task automatic risk assessment model outputs has_risk = -1, it is uncertain whether there is risk. Because the risk type and the existence of risk are uncertain, there is no risk assessment review conclusion.

[0096] In this embodiment, the multi-task risk discrimination model can output a score or a probability value for a risk type. By comparing the value of the risk type discrimination review conclusion with a set discrimination threshold, the data to be discriminated that meets the advancement requirements is advanced to the end processing stage. For example, when the value of the discrimination review conclusion is greater than or equal to the discrimination threshold, the transaction data corresponding to the discrimination review conclusion can be identified as data with a high discrimination accuracy, and the transaction data can be advanced to the end processing stage to avoid risk and ensure the transaction security of the transaction participants.

[0097] In such Figure 2 As shown, the risk assessment platform sends the output assessment conclusion to the risk management platform. The risk management platform can display the assessment conclusion, for example, on the viewing page corresponding to the viewing request. In this embodiment, a second prompt message corresponding to the assessment conclusion can also be displayed in the assessment conclusion display area on the viewing page. The second prompt message can be an audit prompt message, such as: the current transaction data has been identified as risky, and the risk category is the second risk type (such as fraud risk). In this embodiment, when the assessment conclusion is the second risk type (such as fraud risk) and / or the third risk type (such as money laundering risk), a first prompt message (icon), such as a "recommended" label, can also be displayed in the assessment conclusion display area.

[0098] The above describes an embodiment of a risk assessment system provided in this application. This embodiment uses a risk assessment platform to risk-label acquired transaction data and sends the labeled risk-labeled data to a risk management platform. The risk management platform initiates a risk assessment request to the risk assessment platform based on a risk assessment task. This allows the risk points in the transaction data to be perceived immediately upon initiating the risk assessment task through the risk-labeled data. The risk assessment platform identifies the risk-labeled data carried in the risk assessment request and the corresponding transaction data as data to be assessed. Based on a multi-task automatic risk assessment model trained separately according to risk type, the platform performs risk assessment on the data to be assessed, determines the assessment review conclusion, and sends the review conclusion to the risk management platform. This ensures the accuracy of risk assessment of transaction data and avoids omissions or misjudgments of risk points. Furthermore, it can advance to the end stage of risk assessment if the risk assessment conclusion meets the requirements, improving the efficiency of risk assessment.

[0099] It is understood that the risk identification platform and risk management platform involved in the risk identification system provided in this application can be located on the same execution side to complete the corresponding data processing and interaction functions, or they can be distributed on different end sides to realize the corresponding data processing and interaction, such as: local end and server end.

[0100] Based on the above, this application also provides a risk assessment method, which is described using the risk assessment platform as the execution end. For example... Figure 5 As shown, Figure 5 This is a flowchart of a risk assessment method provided in this application, which may include:

[0101] Step S501: In response to a risk assessment request initiated for a risk assessment task, the risk labeling data carried in the risk assessment request and the transaction data corresponding to the risk labeling data are determined as data to be assessed. The risk assessment request may be invoked during the execution of a risk assessment task, which may include viewing or other related processing operations. In this embodiment, the acquired transaction data also needs to be risk-labeled to determine the risk-labeled data. The specific implementation process may include: acquiring the transaction data through a first acquisition link and / or a second acquisition link according to the set automatic labeling trigger conditions; performing risk labeling on the transaction data according to the constructed risk judgment rule set to determine the risk-labeled data corresponding to the transaction data. In this embodiment, the risk-labeled data may include risk description information and risk values, etc. The risk values ​​may include judgment values ​​such as yes, no, and cannot be determined, and these values ​​have modifiable permissions. The presence of risk can also be determined by identifying the buyer's shipping address, contact information, transaction IP, etc. In other words, the risk assessment rule set for buyers can be constructed using any one or more pieces of information from various transaction data, such as buyer transaction behavior, buyer transaction attributes, and buyer profiles. Of course, different risk strategies can be adopted for different transaction items corresponding to the transaction data; for example, targeted risk strategies can be set based on the value and attributes of the transaction items. Therefore, the risk assessment rule set in this embodiment is only illustrated as an example and is not limited to the above methods. For sellers, the risk assessment rules can identify whether the transaction data related to the seller includes high-risk categories, whether they are on a negative list, or whether they have been blacklisted, etc. These will not be listed in detail here, and can also be detected through algorithmic models. The risk assessment rule set can be constructed based on relevant data such as transaction scenarios, transaction items, and transaction methods. The risk identification platform uses the risk assessment rule set to label the risk points of the acquired transaction data, determining the risk labeling data corresponding to the transaction data. The risk tagging data is applied to the transaction data, thus there can be a correlation between the two. For example, the corresponding transaction data can be found through the risk tagging data, or the risk tagging data of the transaction data can be obtained through the transaction data. To facilitate intuitive capture of risk points and reduce processing operations in scenarios requiring table lookups, statistics, and analysis, the risk tagging data may also include: a first prompt information (icon). In this embodiment, the first prompt information may be a "recommended" marker displayed in the risk judgment value area, which can serve as recommended data for subsequent judgments.

[0102] In this embodiment, the transaction data can be acquired through the first acquisition link and / or the second acquisition link according to the set automatic tagging trigger conditions in the following manner:

[0103] Based on the set selection criteria, the transaction data snapshots of the transaction participants that meet the risk assessment criteria are selected as the first acquisition link, and the transaction data is acquired when the automatic tagging trigger condition is met.

[0104] And / or,

[0105] Based on the set selection criteria, the integrated data of the transaction data of the selected transaction participants that meet the risk assessment criteria is used as the second acquisition link. When the automatic tagging trigger condition is met, the transaction data is acquired.

[0106] Step S502: Based on the multi-task automatic risk discrimination model trained separately according to risk type, perform risk discrimination on the data to be discriminated and determine the discrimination review conclusion of the data to be discriminated; the specific implementation process may include: training a first classification sub-model according to a first risk type to determine a target first classification sub-model; training a second classification sub-model according to a second risk type to determine a target second classification sub-model; training a third classification sub-model according to a third risk type to determine a target third classification sub-model; fitting the target first classification sub-model, the target second classification sub-model, and the target third classification sub-model into a multi-task automatic risk discrimination model. The process of inputting the data to be judged into the multi-task automatic risk judgment model to determine the judgment and review conclusion of the data to be judged may include: determining the data features of the risk tagging data extracted from the data to be judged as the first feature; determining the selection features of the transaction data extracted from the data to be judged as the second feature; determining the statistical features of the selected features of the transaction data extracted from the data to be judged as the third feature; determining the risk profile data features of the transaction data extracted from the data to be judged as the fourth feature; inputting one or more of the first, second, third, and fourth features into the multi-task automatic risk judgment model for risk type judgment; and determining the risk type judgment result as the judgment and review conclusion of the data to be judged.

[0107] Step S503: Send the judgment and review conclusion. Specifically, the risk judgment platform may send the output judgment and review conclusion to the risk management platform, which may then display the judgment and review conclusion, for example, on the viewing page corresponding to the viewing request. In this embodiment, a second prompt message corresponding to the judgment and review conclusion may also be displayed in the judgment and review conclusion display area on the viewing page. The second prompt message may be a review prompt message, such as: the current transaction data is identified as risky, and the risk category is the second risk type (such as fraud risk). In this embodiment, when the judgment and review conclusion is the second risk type (such as fraud risk) and / or the third risk type (such as money laundering risk), a first prompt message (icon), such as a "recommendation" marker, may be displayed in the judgment and review conclusion display area. Based on the comparison between the judgment and review conclusion and the set judgment threshold, the data to be judged that meets the advancement requirements is advanced to the end processing operation stage. That is, sending the judgment and review conclusion may include directly advancing the data that meets the recommendation requirements to the end processing operation stage, i.e., recognizing the risk fact and executing subsequent related processing operations.

[0108] The description of the above-described risk assessment method embodiments can be found in the description of the risk assessment platform in the aforementioned risk assessment system. It is understood that the risk assessment method provided in this application primarily focuses on the risk assessment platform to achieve automatic risk labeling and automatic risk assessment. Automatic risk labeling facilitates the immediate detection of risk information in transaction data during risk assessment tasks, thereby preventing potential risks from being overlooked and improving the efficiency of subsequent processing operations related to risk assessment tasks. Automatic risk assessment obtains risk assessment conclusions regarding transaction data from the output of a multi-task automatic risk assessment model. This reduces the probability of misjudgments, improves the accuracy of risk assessment and the efficiency of related processing, and avoids losses to transaction participants due to the time-consuming risk assessment process when transaction data is generated.

[0109] In other words, the core of the risk identification method provided in this application lies in the processing of automated risk labeling and automated risk identification. The identification and review conclusion obtained after the automated risk identification process does not concern itself with how the risk labeling data and the identification and review conclusion are output and displayed, or how the transaction data is sent.

[0110] It is understood that the risk identification method provided in this application is an example described with a risk identification platform as the execution side. The risk identification platform is only the name of the execution side in the execution of a method or system. The corresponding execution process is not limited to the risk identification platform, but can also be other execution parties that can implement the risk identification method process.

[0111] The above is a description of an embodiment of a risk identification method provided in this application. Corresponding to the aforementioned embodiment of a risk identification method, this application also discloses an embodiment of a risk identification device. Please refer to [link / reference]. Figure 6 Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0112] like Figure 6 As shown, Figure 6 This is a schematic diagram of a risk assessment device provided in this application. The device includes: a first determining unit 601, a second determining unit 602, and a sending unit 603.

[0113] The first determining unit 601 is used to determine the risk labeling data carried in the risk labeling request and the transaction data corresponding to the risk labeling data as data to be identified in response to the risk identification request initiated for the risk identification task;

[0114] The second determining unit 602 is used to perform risk judgment on the data to be judged based on a multi-task risk automatic judgment model trained separately based on risk type, and determine the judgment review conclusion of the data to be judged;

[0115] The sending unit 603 is used to send the judgment and review conclusion.

[0116] It also includes a tagging unit, used to perform risk tagging on the acquired transaction data and determine the risk tagging data of the transaction data.

[0117] The tagging unit may specifically include an acquisition subunit and a determination subunit. The acquisition subunit is used to acquire the transaction data through a first acquisition link and / or a second acquisition link according to the set automatic tagging trigger conditions. The determination subunit is used to perform risk tagging on the transaction data according to the constructed risk judgment rule set and determine the risk tagging data corresponding to the transaction data.

[0118] The acquisition subunit may specifically include: selecting a snapshot of the transaction data of a transaction participant that meets the risk assessment criteria as the first acquisition link according to the set selection criteria, and acquiring the transaction data when the automatic tagging trigger condition is met; and / or, selecting the integrated data of the transaction data of the transaction participants that meet the risk assessment criteria as the second acquisition link according to the set selection criteria, and acquiring the transaction data when the automatic tagging trigger condition is met.

[0119] The second determining unit 602 can be specifically implemented by including: a first determining subunit, a second determining subunit, a third determining subunit, a fitting subunit, and a conclusion determining subunit. The first determining subunit is used to train a first classification submodel based on a first risk type to determine a target first classification submodel; the second determining subunit is used to train a second classification submodel based on a second risk type to determine a target second classification submodel; the third determining subunit is used to train a third classification submodel based on a third risk type to determine a target third classification submodel; the fitting subunit is used to fit the target first classification submodel, the target second classification submodel, and the target third classification submodel into a multi-task automatic risk discrimination model; and the conclusion determining subunit is used to input the data to be discriminated into the multi-task automatic risk discrimination model to determine the discrimination and review conclusion of the data to be discriminated.

[0120] The specific implementation process of the conclusion determination subunit may include: a first feature determination subunit, a second feature determination subunit, a third feature determination subunit, a fourth feature determination subunit, a discrimination subunit, and a determination subunit. The first feature determination subunit is used to determine the data features of the risk-labeled data extracted from the data to be discriminated as the first feature; the second feature determination subunit is used to determine the selection features of the transaction data extracted from the data to be discriminated as the second feature; the third feature determination subunit is used to determine the statistical features of the selected features of the transaction data extracted from the data to be discriminated as the third feature; the fourth feature determination subunit is used to determine the risk profile data features of the transaction data extracted from the data to be discriminated as the fourth feature; the discrimination subunit is used to input one or more of the first, second, third, and fourth features into the multi-task automatic risk discrimination model for risk type discrimination; the determination subunit is used to determine the risk type discrimination result as the discrimination review conclusion of the data to be discriminated.

[0121] The specific implementation process of the sending unit 603 may include: a push subunit, which is used to compare the judgment review conclusion with the set judgment threshold, and push the data to be judged that meets the push requirements to the end processing operation stage.

[0122] The above is a description of an embodiment of a risk identification device provided in this application. The specific implementation process of the device can be referred to the description of the risk identification system and method above, and will not be described in detail here.

[0123] Based on the above, this application also provides a method for outputting risk data, such as... Figure 7 As shown, Figure 7This is a flowchart illustrating a risk data output method provided in this application. The method is described using the risk management platform in the risk assessment system as an example. The core of this method lies in how to provide transaction data to the risk assessment platform and how to output the acquired risk labeling data and assessment conclusions. It does not focus on how the risk labeling data or assessment conclusions are completed; regardless of the method used to complete the risk labeling data and assessment conclusions, this risk data output method can be used. The risk data output method may include:

[0124] Step S701: Distribute the received risk labeling data as a risk identification task; in this embodiment, the distribution can be distributed to the task execution side, whichever is more specific.

[0125] Step S702: Based on the viewing request for the risk assessment task, generate and send a risk assessment request for the target transaction data, wherein the risk assessment request carries the target transaction data and the target risk labeling data corresponding to the target transaction data.

[0126] Step S703: Receive the judgment and review conclusions corresponding to the target transaction data from the multi-task automatic risk discrimination model trained separately based on risk type; the specific implementation process may include:

[0127] Step S703-1: Receive the target transaction data and the target risk tagging data corresponding to the target transaction data as input data for the multi-task risk automatic judgment model, and output the judgment and review conclusion on the risk type of the target transaction data;

[0128] Step S703-2: Store the judgment and review conclusion and display it on the viewing page corresponding to the viewing request.

[0129] It may also include:

[0130] Based on the received risk labeling data, add a first prompt message corresponding to the risk labeling data;

[0131] Output the risk labeling data and the first prompt information.

[0132] It may also include:

[0133] The area displaying the review conclusion on the viewing page displays a second prompt message corresponding to the review conclusion.

[0134] The above-mentioned risk data output method can be described in conjunction with the above-mentioned risk identification system and risk identification method embodiments, and will not be elaborated here.

[0135] The above is a detailed description of an embodiment of a risk data output method provided in this application. Corresponding to the aforementioned embodiment of a risk data output method, this application also discloses an embodiment of a risk data output device. Please refer to [link / reference]. Figure 8 Since the device embodiments are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0136] like Figure 8 As shown, Figure 8 This is a schematic diagram of a risk data output device provided in this application. The embodiment of the device includes: a distribution unit 801, a generation unit 802, and a receiving unit 803.

[0137] The distribution unit 801 is used to distribute the received risk labeling data as a risk discrimination task;

[0138] The generation unit 802 is used to generate and send a risk assessment request for target transaction data according to the viewing request for the risk assessment task, wherein the risk assessment request carries the target transaction data and target risk labeling data corresponding to the target transaction data;

[0139] The receiving unit 803 is used to receive the judgment and review conclusion corresponding to the target transaction data output by the multi-task risk automatic judgment model trained separately based on risk type. Specifically, it includes a receiving subunit and a display subunit. The receiving subunit is used to receive the judgment and review conclusion for the risk type of the target transaction data, which is output by the multi-task risk automatic judgment model, using the target transaction data and the target risk tagging data corresponding to the target transaction data as input data. The display subunit is used to store the judgment and review conclusion and display it on the viewing page corresponding to the viewing request.

[0140] It also includes: an adding unit and a tag output unit, wherein the adding unit is used to add a first prompt message corresponding to the received risk tagging data; and the tag output unit is used to output the risk tagging data and the first prompt message. It also includes: a display unit, used to display a second prompt message corresponding to the judgment and review conclusion in the judgment and review conclusion display area displayed on the viewing page.

[0141] The above is a description of a risk data output method provided in this application. For the specific implementation process of this risk data output method, please refer to the above-mentioned risk identification system, risk identification method and risk data output method. It will not be repeated here.

[0142] Based on the above, this application also provides a computer storage medium for storing data generated by a network platform, and a program for processing the data generated by the network platform.

[0143] When the program is read and executed, it performs the logical steps as described in the risk assessment system, or the steps as described in the risk assessment method, or the steps as described in the risk data output method.

[0144] Based on the above, this application also provides an electronic device, such as... Figure 9 As shown, Figure 9 This is a schematic diagram of the structure of an electronic device provided in this application, which may include:

[0145] Processor 901;

[0146] The memory 902 is used to store a program for processing data generated by the network platform. When the program is read and executed by the processor, it performs the logical steps as performed in the risk assessment system described above, or the steps as performed in the risk assessment method described above, or the steps as performed in the risk data output method described above.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0148] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0149] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0150] 1. Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0151] 2. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A risk discrimination system, characterized by, Comprise: A risk management platform and a risk discrimination platform; The risk management platform is configured to receive risk labeling data fed back by the risk discrimination platform for risk labeling of transaction data, distribute the risk labeling data as a risk discrimination task, initiate a risk discrimination request to the risk discrimination platform, and obtain a discrimination audit conclusion fed back by the risk discrimination platform corresponding to the risk discrimination request; The risk discrimination platform is configured to perform risk labeling on the transaction data, and determine the risk labeling data of the transaction data; In response to the risk discrimination request initiated by the risk management platform for the risk discrimination task, the risk labeling data in the risk discrimination request and the transaction data corresponding to the risk labeling data are determined as to-be-discriminated data; A multi-task risk automatic discrimination model trained according to risk types is used to discriminate the to-be-discriminated data, determine a discrimination audit conclusion of the to-be-discriminated data, and send the discrimination audit conclusion to the risk management platform.

2. The risk discrimination system according to claim 1, characterized in that, The risk discrimination platform is configured to acquire the transaction data through a first acquisition link and / or a second acquisition link according to a set automatic labeling trigger condition, and perform risk labeling on the transaction data according to a constructed risk judgment rule set to determine the risk labeling data corresponding to the transaction data.

3. The risk discrimination system according to claim 1, characterized in that, The risk discrimination platform is configured to train a first classification sub-model according to a first risk type to determine a target first classification sub-model; Train a second classification sub-model according to a second risk type to determine a target second classification sub-model; Train a third classification sub-model according to a third risk type to determine a target third classification sub-model; The to-be-discriminated data is input into a multi-task risk automatic discrimination model fitted according to the first classification sub-model, the second classification sub-model, and the third classification sub-model to determine a discrimination audit conclusion of the to-be-discriminated data.

4. The risk discrimination system of claim 1, wherein, The risk management platform further comprises: According to the acquired discrimination audit conclusion, a target discrimination audit conclusion meeting the discrimination requirement is screened, and the target discrimination audit conclusion and the transaction data corresponding to the target discrimination audit conclusion are pushed to a data processing stage as promotion data to perform corresponding processing operations.

5. The risk discrimination system of claim 1, wherein, The risk management platform is configured to determine data features of the risk labeling data in the to-be-discriminated data as first features; Determine selected features of the transaction data in the to-be-discriminated data as second features; Determine statistical features of the selected features of the transaction data in the to-be-discriminated data as third features; Determine risk portrait data features of the transaction data in the to-be-discriminated data as fourth features; 6. A risk discrimination method characterized by comprising: One or more features of the first features, the second features, the third features, and the fourth features are input into the automatic discrimination model for risk type discrimination, and the risk type discrimination result is determined as the discrimination audit conclusion of the to-be-discriminated data. Comprise: In response to a risk discrimination request initiated for a risk discrimination task, determine the risk labeling data in the risk discrimination request and the transaction data corresponding to the risk labeling data as to-be-discriminated data; According to the multi-task risk automatic discrimination model trained based on the risk types, perform risk discrimination on the to-be-discriminated data to determine a discrimination audit conclusion of the to-be-discriminated data. Send the discrimination audit conclusion.

7. The risk discrimination method according to claim 6, characterized in that, Also includes: Risk labeling of the obtained transaction data to determine risk labeling data of the transaction data.

8. The risk discrimination method according to claim 7, characterized in that, The risk labeling of the obtained transaction data to determine the risk labeling data of the transaction data includes: According to the set automatic labeling trigger condition, obtain the transaction data through the first acquisition link and / or the second acquisition link; According to the constructed risk judgment rule set, perform risk labeling on the transaction data to determine the risk labeling data corresponding to the transaction data.

9. The risk discrimination method according to claim 8, characterized in that, The transaction data obtained according to the set automatic labeling trigger condition includes: According to the set selection condition, the transaction data snapshot of the transaction participant selected to meet the risk discrimination is taken as the first acquisition link, and the transaction data is obtained when the automatic labeling trigger condition is met; And / or, According to the set selection condition, the integrated data of the transaction data integrated by selecting the transaction participant meeting the risk discrimination is taken as the second acquisition link, and the transaction data is obtained when the automatic labeling trigger condition is met.

10. The risk discrimination method according to claim 6, characterized in that, The risk discrimination on the to-be-discriminated data according to the multi-task risk automatic discrimination model trained based on the risk types includes: Train the first classification sub-model according to the first risk type to determine a target first classification sub-model; Train the second classification sub-model according to the second risk type to determine a target second classification sub-model; Train the third classification sub-model according to the third risk type to determine a target third classification sub-model; Fit the target first classification sub-model, the target second classification sub-model, and the target third classification sub-model into a multi-task risk automatic discrimination model; Input the to-be-discriminated data into the multi-task risk automatic discrimination model to determine a discrimination audit conclusion of the to-be-discriminated data.

11. The risk discrimination method according to claim 10, characterized in that, The input of the to-be-discriminated data into the multi-task risk automatic discrimination model to determine the discrimination audit conclusion of the to-be-discriminated data includes: Determine the data features of the risk labeling data in the to-be-discriminated data extracted as first features; Determine the selected features of the transaction data in the to-be-discriminated data extracted as second features; Determine the statistical features of the selected features of the transaction data in the to-be-discriminated data extracted as third features; Determine the risk portrait data features of the transaction data in the to-be-discriminated data extracted as fourth features; Input one or more of the first features, second features, third features, and fourth features into the multi-task risk automatic discrimination model for risk type discrimination; The risk type discrimination result is determined as a discrimination audit conclusion of the to-be-discriminated data.

12. The risk discrimination method according to claim 6, characterized in that, The sending of the discrimination audit conclusion comprises: According to the comparison between the discrimination audit conclusion and the set discrimination threshold, the to-be-discriminated data meeting the promotion requirement is promoted to an end processing operation stage.

13. A risk data output method characterized by, Comprise: Distribute the received risk labeling data as a risk discrimination task; According to the viewing request for the risk discrimination task, generate and send a risk discrimination request of target transaction data, wherein the risk discrimination request carries the target transaction data and target risk labeling data corresponding to the target transaction data; Receive a multi-task risk automatic discrimination model trained based on a risk type, and output a discrimination audit conclusion corresponding to the target transaction data.

14. The risk data output method according to claim 13, wherein, The receiving of the multi-task risk automatic discrimination model trained based on the risk type and outputting of the discrimination audit conclusion corresponding to the target transaction data comprises: Receive the target transaction data and the target risk labeling data corresponding to the target transaction data as input data of the multi-task risk automatic discrimination model, and output a discrimination audit conclusion of the risk type of the target transaction data; Store and display the discrimination audit conclusion on a viewing page corresponding to the viewing request.

15. A computer storage medium for storing network platform generated data and a program for processing the network platform generated data; The program, when read and executed, performs the logical steps performed in any one of the risk discrimination systems of claims 1 to 5, or performs the steps of any one of the risk discrimination methods of claims 6 to 12, or performs the steps of the risk data output method of claim 13 or 14.

16. An electronic device comprising: A processor; A memory for storing a program for processing network platform generated data, the program, when read and executed by the processor, performs the logical steps performed in any one of the risk discrimination systems of claims 1 to 5, or performs the steps of any one of the risk discrimination methods of claims 6 to 12, or performs the steps of the risk data output method of claim 13 or 14.

Images