Object processing method and system, computing device and storage medium

Abstract

Embodiments of the present specification provide an object processing method and system, a computing device and a storage medium, the object processing method comprising: determining to-be-detected data and a risk detection result of a target object, the risk detection result being obtained by using a risk detection unit to perform risk detection on the to-be-detected data; determining a risk identification problem corresponding to the target object according to the to-be-detected data and the risk detection result by using a risk identification model; providing the risk identification question to an information interaction unit corresponding to the target object, and receiving an answer result for the risk identification question returned by the information interaction unit; and performing risk identification on the target object by using the risk identification model according to the risk identification question, the answer result, the to-be-detected data and the risk detection result to obtain a risk identification result corresponding to the target object.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of risk control technology, and in particular to an object processing method; one or more embodiments of this specification also relate to an object processing system, a credit processing method, an object processing device, a computing device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous development of computer and internet technologies, many organizations or enterprises provide services to specific groups to meet their needs.

[0003] However, in the existing technology, when a specific object makes a service application, it may commit fraud by providing false data, resulting in significant losses. In order to avoid significant losses caused by fraud, it is necessary to identify risks based on the data provided by the specific object. Therefore, how to accurately identify risks of a specific object has become an urgent problem to be solved. Summary of the Invention

[0004] In view of the above, embodiments of this specification provide an object processing method. One or more embodiments of this specification also relate to an object processing system, a credit processing method, an object processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, an object processing method is provided, including:

[0006] The target object's data to be detected and the risk detection results are determined, wherein the risk detection results are obtained by performing risk detection on the data to be detected using a risk detection unit;

[0007] Using a risk identification model, the risk identification problem corresponding to the target object is determined based on the data to be detected and the risk detection results.

[0008] The risk identification question is provided to the information interaction unit corresponding to the target object, and the answer result returned by the information interaction unit for the risk identification question is received, wherein the answer result is generated by the information interaction unit based on the target object's answer behavior for the risk identification question;

[0009] Using the risk identification model, the target object is identified based on the risk identification question, the answer result, the data to be detected, and the risk detection result, thereby obtaining the risk identification result corresponding to the target object.

[0010] According to a second aspect of the embodiments of this specification, an object processing apparatus is provided, comprising:

[0011] The information determination module is configured to determine the data to be detected and the risk detection result of the target object, wherein the risk detection result is obtained by performing risk detection on the data to be detected using a risk detection unit;

[0012] The problem determination module is configured to use a risk identification model to determine the risk identification problem corresponding to the target object based on the data to be detected and the risk detection results.

[0013] The response receiving module is configured to provide the risk identification question to the information interaction unit corresponding to the target object, and receive the response result returned by the information interaction unit for the risk identification question, wherein the response result is generated by the information interaction unit based on the target object's response behavior to the risk identification question;

[0014] The risk identification module is configured to use the risk identification model to identify risks of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtain the risk identification result corresponding to the target object.

[0015] According to a third aspect of the embodiments of this specification, an object processing system is provided, the system comprising an information interaction unit, a risk detection unit, and a risk identification unit, wherein...

[0016] The information interaction unit is configured to determine the data to be detected of the target object, provide the data to be detected to the risk detection unit and the risk identification unit, send the risk identification question provided by the risk identification unit to the target object, and return the answer result generated based on the target object's response to the risk identification question to the risk identification unit.

[0017] The risk detection unit is configured to perform risk detection on the data to be detected, obtain the risk detection result of the target object, and provide the risk detection result to the risk identification unit;

[0018] The risk identification unit is configured to determine the target object's data to be detected and the risk detection result, use a risk identification model to determine the risk identification question corresponding to the target object based on the data to be detected and the risk detection result, provide the risk identification question to the information interaction unit corresponding to the target object, receive the answer result returned by the information interaction unit for the risk identification question, and use the risk identification model to perform risk identification on the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result to obtain the risk identification result corresponding to the target object.

[0019] According to a fourth aspect of the embodiments of this specification, a credit processing method is provided, comprising:

[0020] The data to be tested for credit users and the credit risk detection results are determined, wherein the credit risk detection results are obtained by using a credit risk detection unit to perform credit risk detection on the data to be tested;

[0021] Using a credit risk identification model, the credit risk identification problem corresponding to the credit user is determined based on the data to be detected and the credit risk detection results.

[0022] The credit risk identification question is provided to the information interaction unit corresponding to the credit user, and the answer result returned by the information interaction unit for the credit risk identification question is received, wherein the answer result is generated by the information interaction unit based on the credit user's answer behavior for the credit risk identification question;

[0023] Using the credit risk identification model, the credit risk of the credit user is identified based on the credit risk identification question, the answer result, the data to be detected, and the credit risk detection result, thereby obtaining the credit risk identification result corresponding to the credit user.

[0024] According to a fifth aspect of the embodiments of this specification, a computing device is provided, comprising:

[0025] Memory and processor;

[0026] The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the above-described object processing method or credit processing method.

[0027] According to a sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores a computer program / instructions that, when executed by a processor, implement the steps of the object processing method or credit processing method described above.

[0028] According to a seventh aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the object processing method or credit processing method described above.

[0029] The object processing method in one or more embodiments provided in this specification, during the object processing process, after the risk detection unit performs a first-stage risk detection on the data to be detected and obtains the risk detection result, in order to ensure the accuracy of risk identification, can use a risk identification model to determine the corresponding risk identification question for the target object based on the data to be detected and the risk detection result. Then, based on the target object's answer to the risk identification question, the risk identification question, the data to be detected, and the risk detection result, a second-stage risk identification is performed to obtain an accurate risk identification result. This achieves accurate risk identification of the target object based on the data information provided by the target object, avoiding significant losses caused by the target object committing fraud by providing false data information. Attached Figure Description

[0030] Figure 1 This is a schematic diagram illustrating the application of an object processing method provided in one embodiment of this specification;

[0031] Figure 2 This is a flowchart of an object processing method provided in one embodiment of this specification;

[0032] Figure 3 This is a schematic diagram of model training in an object processing method provided in one embodiment of this specification;

[0033] Figure 4 This is a schematic diagram illustrating the application of a model in an object processing method provided in one embodiment of this specification;

[0034] Figure 5 This is a flowchart illustrating the processing procedure of an object processing method provided in one embodiment of this specification.

[0035] Figure 6 This is a schematic diagram of the structure of an object processing system provided in one embodiment of this specification;

[0036] Figure 7 This is a schematic diagram of the structure of an object processing device provided in one embodiment of this specification;

[0037] Figure 8 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0038] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0039] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0040] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0041] Furthermore, 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, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0042] In one or more embodiments of this specification, a large model refers to a deep learning model with a large number of model parameters, typically containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even tens of trillions of model parameters. A large model can also be called a foundational model (Foundation Model 1). It is pre-trained using large-scale unlabeled corpora to produce a pre-trained model with hundreds of millions of parameters. Such models can adapt to a wide range of downstream tasks and have good generalization ability. Examples include Large Language Models (LLMs) and Multimodal Pre-training Models (MLMs).

[0043] In practical applications, large models only require a small number of samples to fine-tune the pre-trained model before they can be applied to different tasks. Large models can be widely used in fields such as Natural Language Processing (NLP) and Computer Vision. Specifically, they can be applied to computer vision tasks such as Visual Question Answering (VQA), Image Captioning (IC), and Image Generation, as well as NLP tasks such as text-based sentiment classification, text summarization, and machine translation. The main application scenarios for large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design.

[0044] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0045] Large-scale models (LLMs) refer to language models with a large number of parameters and capabilities. These models are trained to handle natural language processing tasks such as text generation, text classification, and machine translation. LLMs are typically composed of deep neural networks with billions or even tens of billions of parameters, enabling them to capture a vast amount of linguistic features and contextual information, and exhibiting impressive language understanding and generation capabilities.

[0046] Anti-fraud: This can refer to credit anti-fraud, which means taking measures and technologies in credit services to identify and prevent fraudulent activities, including identifying false identity information, preventing identity theft, detecting false financial information provided by applicants, and detecting whether applicants have been the victim of telecommunications fraud.

[0047] Credit assessors are professionals who perform credit assessments and approvals at financial institutions or other lending organizations. They are responsible for reviewing a loan applicant's credit history, income, debt situation, and other relevant information to determine whether to approve the loan application, as well as the specific details regarding the loan amount, interest rate, and terms.

[0048] With the continuous development of computer and internet technologies, many institutions and enterprises provide services tailored to specific groups to meet their needs. For example, in the credit sector, financial institutions commonly use online credit loan approval. Through digital platforms, applicants can easily submit loan applications from home and quickly obtain approval results. This trend not only saves customers time and effort but also enables financial institutions to process a large number of loan applications in a short period, greatly expanding their scale. On the other hand, by querying various types and levels of third-party data, borrowers can obtain more accurate user profile information, thereby making accurate credit assessments.

[0049] However, certain individuals may engage in fraudulent activities by providing false data during the service application process, leading to significant losses. For example, in credit scenarios, while online credit loan approval is convenient, this convenience comes at the cost of security. On one hand, a group of black market intermediaries specializing in malicious loan fraud will specifically target loopholes in credit systems to help unqualified customers obtain excessive credit. On the other hand, telecommunications fraud remains widespread. After gaining the trust of customers, fraudsters often guide them to loan platforms to withdraw cash and then transfer it to the fraudsters in order to extract as much cash as possible. Currently, there are no particularly good solutions for these two issues. Some financial institutions have introduced manual credit checks to verify the authenticity of users and their genuine borrowing intentions; however, this sacrifices the convenience of digital lending and increases labor costs.

[0050] To address the aforementioned issues, this manual provides two solutions. The first solution is a manual credit review solution. In this solution, if the system determines that a user may be risky, the credit approval process is suspended and transferred to a human review process. This human review can take the form of a phone call or an online interview. However, this solution has significant drawbacks: First, because the process is suspended and requires human intervention, credit reviewers need to be online 24 / 7; otherwise, the credit approval process cannot return results in real time, thus significantly impacting the user experience. Second, credit reviewers themselves represent additional labor costs. Finally, the skill levels of credit reviewers vary, and those lacking sufficient ability are unlikely to be effective in fraud detection. In other words, manual credit review requires experienced credit reviewers, and human intervention significantly increases product costs and reduces customer experience, thus possessing considerable shortcomings.

[0051] The second approach avoids manual credit checks and instead relies on accessing as much external third-party data (all structured data) as possible, establishing numerous rules and models to identify fraudulent users at the data level. However, this approach has significant drawbacks: First, the structured data lacks real-time responsiveness and effectiveness. For example, a user who has been scammed by a telecom fraudster might apply for a loan within twenty minutes of receiving the call, making it difficult for any third-party data to identify the fraud risk in such a short time. Even if it is identified the next day, the loan has already been obtained, rendering the data meaningless. Second, if a user decides to borrow from an illegal intermediary, and the user has no prior criminal record, third-party data may struggle to assess this type of risk. In other words, while this anti-fraud method can identify some simple fraudulent activities, the underlying data and data-based modeling methods involved are relatively mature and fixed. More sophisticated fraudsters can often bypass these steps, making this approach significantly flawed.

[0052] Based on this, this specification provides an object processing method, and also relates to an object processing system, a credit processing method, an object processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0053] See Figure 1 , Figure 1 This diagram illustrates an application illustration of an object processing method provided according to an embodiment of this specification, based on... Figure 1 As can be seen, the user sends the data to be detected and the risk detection result to the server 104 through the terminal 102. The server 104 processes the data to be detected and the risk detection result using the risk identification model to obtain the risk identification question. Then, the server 104 sends the risk identification question to the terminal 102. After the terminal 102 displays the risk identification question to the user, it obtains the user's answer based on the user's answer to the risk identification question and sends the answer to the server 104 again for risk identification. After receiving the answer, the server 104 uses the risk identification model to identify the risk in the answer, the data to be detected, and the risk detection result to obtain the risk identification result, and sends the risk identification result to the terminal 102 to display to the user.

[0054] Through the above steps, the risk of the target can be accurately identified based on the data provided by the target, thus avoiding significant losses caused by the target committing fraud by providing false data.

[0055] See Figure 2 , Figure 2 A flowchart of an object processing method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0056] Step 202: Determine the data to be detected and the risk detection results of the target object, wherein the risk detection results are obtained by performing risk detection on the data to be detected using a risk detection unit.

[0057] The target object can be understood as the object that needs to be risk-detected. The target object can be a user, enterprise, company, or institution. For example, the target object can be a user or company applying for a loan.

[0058] The data to be tested can be understood as data corresponding to the target object that needs to be risk-detected; by conducting risk detection on the data to be tested, it is possible to identify whether the target object poses a risk.

[0059] The risk detection unit can be understood as a unit used to perform risk detection on a target object. For example, the risk detection unit is a risk control system, a core risk system, a risk control unit, etc.; the risk detection can also be understood as fraud risk detection; the risk detection unit can be a server used to perform risk detection. The server can be a server, virtual machine, container, or other device, without specific limitations.

[0060] The risk detection result can be understood as the result obtained by the risk detection unit through risk detection of the data to be detected; when the risk detection unit is a risk control system, the risk detection result can be understood as the fraud detection result of whether the target object has a fraud risk; the risk detection result can be expressed in text or semantic form, for example, the risk detection result can be "the risk control system determines that the user may have a black intermediary risk" or "the risk control system determines that the user may have a fraud risk", that is to say, the risk detection result can be a risk detection text result.

[0061] In one or more embodiments provided in this specification, the risk detection unit can acquire the data to be detected of the target object, perform risk detection on the data to be detected, and obtain the risk detection result of the target object.

[0062] Specifically, the step of performing risk detection on the data to be detected and obtaining the risk detection result for the target object includes:

[0063] Based on the data to be detected, the target object is subjected to qualification detection, and if the target object is determined to have passed the qualification detection, the target object is subjected to risk detection, and the risk detection result of the target object is obtained.

[0064] For example, in a credit scenario, the risk control system can use the data to be tested to perform credit qualification checks on the credit applicant (i.e., the target object) (i.e., the qualification check mentioned above). If the credit applicant meets the credit qualification requirements, the risk control system will continue to perform a check to determine whether the credit applicant has a fraud risk (i.e., risk detection), thereby obtaining the possible types of fraud by the user (i.e., risk detection results).

[0065] It should be noted that if the target object fails the qualification test, the result of failing the qualification test will be used as the risk identification result for the target object. This risk identification result can then be sent to the target object via the information interaction unit.

[0066] For example, if it is determined that a credit applicant does not meet the credit eligibility criteria, the system will send the rejection result (i.e., risk identification result) to the credit applicant through the interactive system, thereby rejecting the user's credit application.

[0067] In one or more embodiments provided in this specification, the step of performing risk detection on the data to be detected and obtaining the risk detection result of the target object further includes:

[0068] The target object is identified using the data to be detected to determine the object type corresponding to the target object; for example, the above data (i.e., the data to be detected) is input into the risk control system (i.e., the risk detection unit), and the risk control system uses the risk control decision engine to perform customer segmentation (i.e., customer type) based on the above data.

[0069] Based on the object type and the data to be detected, the target object is subject to eligibility testing. For example, after the risk control system uses the risk control decision engine to determine the customer segment (e.g., personal loan user type) corresponding to the user, it conducts eligibility review based on the eligibility review strategy corresponding to the customer segment and the aforementioned data to determine whether the user meets the credit eligibility requirements. If so, the system performs subsequent operations to determine whether the user has a fraud risk. If not, the system returns a result of credit rejection through the interactive system.

[0070] If the target object passes the qualification test, a risk test is performed on the target object, and the risk test result of the target object is obtained.

[0071] For example, when a risk control system determines that a user meets the creditworthiness requirements, it performs an operation to assess whether the user (i.e., the target) poses a fraud risk (i.e., conducts risk detection). This assessment includes:

[0072] In the process of determining whether a user poses a fraud risk, if it is determined that the user does not pose a fraud risk, then the user meets the criteria for applying for a loan. Therefore, the operation of calculating credit elements such as credit limit and interest is performed, and the credit approval result (which includes credit elements) is returned to the user through the interactive system. If it is determined that the user may pose a fraud risk, then the fraud risk type (i.e., risk detection result) is sent to the large model system for risk identification.

[0073] It should be noted that if the large model system determines that the user has been granted credit during the subsequent risk identification process, the risk control system can perform operations to calculate credit elements such as credit limit and interest based on the results returned by the large model system (i.e., risk identification results), and return the credit approval result to the user through the interactive system.

[0074] In one or more embodiments provided in this specification, if it is determined that the target object fails the qualification test, the target object is identified as a risk object, and the risk identification result corresponding to the risk object is determined. The risk identification result is then provided to the information interaction unit corresponding to the target object, so that the information interaction unit displays the risk identification result to the target object. For example, in the process of the risk control system reviewing the qualifications of users and determining whether users meet the credit granting requirements, if the user does not meet the credit granting qualifications, the system will not perform the subsequent operation of determining whether the user has fraud risk; instead, it will provide the result of refusing credit (i.e., the risk identification result) to the user (i.e., the target object) through the interaction system.

[0075] In one or more embodiments provided in this specification, determining the target object's data to be detected and the risk detection results includes:

[0076] The system receives the detection data of the target object determined by the information interaction unit, wherein the detection data includes object data of the target object and risk detection data, the object data is generated by the information interaction unit based on the information upload operation of the target object, and the risk detection data is determined by the information interaction unit from the data storage unit based on the object data;

[0077] Receive the risk detection result of the target object determined by the risk detection unit, wherein the risk detection result is obtained by the risk detection unit through risk detection of the data to be detected.

[0078] The information interaction unit can be understood as a unit used to obtain the data to be detected through interaction. In one or more embodiments provided in this specification, the information interaction unit can be understood as a unit used to obtain the data to be detected. The unit can be a server, a client, an application deployed on the client, a webpage displayed on the client, a calling interface, etc.; that is to say, the information interaction unit can be a calling interface, a webpage, an application deployed on the client, etc.

[0079] In one or more embodiments provided in this specification, the information interaction unit may include an object interaction subunit and / or a data acquisition subunit.

[0080] The object interaction subunit can be understood as a unit that interacts with the target object. Through the object interaction subunit, the object data of the target object can be obtained. For example, the object interaction subunit can be an application deployed in the client, a webpage displayed in the client, a call interface for obtaining the object data, an interaction system, etc.

[0081] The data acquisition subunit can be understood as a unit that interacts with the data storage unit. Through this data acquisition subunit, risk detection data can be obtained from the data storage unit. For example, the data acquisition subunit can be a calling interface for obtaining the risk detection data, a data transmission channel, a third-party data query system, etc.

[0082] The object data can be data related to the target object. For example, if the target object is a user applying for a loan, the object data can be the user's personal information submitted during the loan application process and / or other related information involved in the loan application process. The user's personal information includes, but is not limited to, the user's personal identity information, personal asset information, etc. The other related information can be the device information corresponding to the device used by the user to apply for a loan online, the location information provided by the device, etc.

[0083] The data storage unit can be understood as a unit that stores risk detection data. This data storage unit can be a local disk, a database, a third-party institution, etc. Correspondingly, the risk detection data includes, but is not limited to, third-party data and local historical data. The third-party data can be understood as data stored in a third-party institution, including, but not limited to, the target object's credit data, debt data, etc. The local historical data can be understood as historical data to be detected, historical risk detection results, historical response results, historical risk identification results, etc., provided by the target object and stored on a local disk or in a database.

[0084] Taking the application of the object processing method provided in this manual in an online credit anti-fraud scenario as an example, this method will be explained. The target object is the user applying for a loan; the object data is the personal information provided by the user during the online credit application process; and the risk detection data can be local historical data and third-party data. The object interaction subunit is the interaction system, the data acquisition subunit is the third-party data query system, and the risk detection unit is the risk control system.

[0085] Based on this, the interactive system can be understood as a user interaction module. The interaction between this system and the user is as follows: First, the user logs into the credit system (i.e., the interactive system) and submits personal information (i.e., object data). Furthermore, in the object processing method, after the internal system makes a risk decision, the user can directly obtain the credit granting result through the interactive system, or the user can answer a series of questions (i.e., risk identification questions) through the interactive system. Based on the effectiveness of the answers (i.e., the answer results), this object processing method returns the final credit granting result through the interactive system.

[0086] For third-party data query systems, after obtaining user information (i.e., object data) and obtaining user authorization, it is necessary to query a series of internal information (i.e., local historical data) and / or external information (i.e., third-party data) of the user.

[0087] Internal information includes, but is not limited to: whether the user has a history of credit or other products with this institution, the user's past application information, and the user's environmental information for this application (GPS, IP, Wi-Fi, etc.).

[0088] External information includes, but is not limited to: user credit information, third-party multiple information, third-party overdue information, blacklist information, and personal asset information such as bank statements and housing provident fund.

[0089] The above data can be input into the risk control system and the large model system respectively to obtain the corresponding processing results.

[0090] Regarding the risk control system, this system is used to determine user qualifications and stratify risks. The specific execution method is as follows:

[0091] First, the above data is input into the risk control system, which uses a risk control decision engine to segment users based on the above data.

[0092] Secondly, after determining the customer segment corresponding to the user (e.g., personal loan user), the above data is reviewed based on the review strategy corresponding to the customer segment to determine the user's qualifications and whether the user meets the credit execution requirements. If so, the user is risk-stratified to determine whether the user has fraud risk; if not, the result of credit rejection is returned through the interactive system.

[0093] Finally, in the process of risk stratification of users and determining whether users have fraud risks, if it is determined that the user does not have fraud risks, then the user is determined to meet the criteria for applying for a loan. Therefore, based on the results returned by the large model system (i.e., the risk identification results), the operation of calculating credit elements such as credit limit and interest is performed, and the credit approval result is returned to the user through the interactive system. Here, risk stratification can be understood as determining the types of fraud risks that users may have.

[0094] If a user is determined to be at risk of fraud, the fraud risk type (i.e., the risk detection result) is sent to the large model system for risk detection.

[0095] Based on the above steps, it can be seen that the risk control system is the core risk control result unit. This risk control system needs to make a comprehensive judgment by combining various types of user information and the results of the large model. When the large model is required to perform automated credit review on users, the core risk control system needs to combine existing data to give the user's suspected risk category. This suspected risk category includes, but is not limited to: risk of being defrauded by telecommunications, risk of being a black intermediary, risk of using false information, etc.

[0096] In the above embodiments, the object processing method uses the information interaction unit and the risk detection unit to determine the target object's data to be detected and the risk detection results, which facilitates the subsequent risk identification model to identify the target object based on comprehensive and accurate data, thereby improving the accuracy of the risk identification results.

[0097] Step 204: Using a risk identification model, determine the risk identification problem corresponding to the target object based on the data to be detected and the risk detection results.

[0098] The risk identification model can be understood as a model used for risk identification. For example, the risk identification model can be a model used for fraud risk identification in online credit anti-fraud scenarios. The risk identification model can be a large model. In one or more embodiments provided in this specification, the risk identification model can have the function of generating risk identification questions and identifying risks of target objects based on the answer results.

[0099] In one or more embodiments provided in this specification, the object processing method can be applied to a risk identification unit, which can be understood as a unit that performs risk identification on a target object. This unit can be a server, a software module in a server, a virtual machine, a cloud server, a container, etc., without specific limitations. In one or more embodiments provided in this specification, the risk identification unit can be a large model system. For example, a large model system can be understood as a system for background information used in subsequent credit review. The core system of the large model (i.e., the large model system) deploys a large model (i.e., a risk identification model). The large model includes a question generator (i.e., a question generation network layer) and an answer discriminator (i.e., a risk identification network layer). The question generator is used to generate user questions (i.e., risk identification questions), and the answer discriminator is used to determine the user's answer to a specific question.

[0100] The question generator continuously generates new questions based on the user's answers until the large model is sufficiently confident that the user is either a fraudulent or non-fraudulent user, at which point it terminates.

[0101] It should be noted that the question generator or answer discriminator can be one or more network layers in a large model; or, the question generator or answer discriminator can be a sub-model in a large model; in one or more embodiments provided in this specification, the core system of the large model can include a question generation model and an answer discriminator model, and the large model can be a large model framework composed of the question generation model and the answer discriminator model.

[0102] Risk identification questions can be understood as questions used to identify risks to a target object. These questions include, but are not limited to, questions such as "What city are you currently in?" and "What is your current age?". Correspondingly, the answer can be understood as the target object's response to the risk identification question. For example, the answer includes, but is not limited to, "I am currently in Beijing" or "I am currently 30 years old." It should be noted that the answer can be presented in the form of text, video, audio, etc., that is, the answer can be a text answer, a video answer, an audio answer, etc.

[0103] Following the previous example, after the third-party data query system, interaction system, and risk control system provide their respective data to the large model system, the large model system will input the above data into the large model and use the large model to generate questions based on the above data, thereby obtaining user questions.

[0104] In one or more embodiments provided in this specification, determining the risk identification problem corresponding to the target object using a risk identification model based on the data to be detected and the risk detection result includes:

[0105] The data to be detected and the risk detection results are input into the risk identification model;

[0106] By utilizing the question generation network layer in the risk identification model, a question generation operation is performed based on the data to be detected and the risk detection results to obtain the risk identification question corresponding to the target object.

[0107] The problem generation network layer can be understood as one or more network layers in the risk identification model used to generate risk identification problems; the problem generation network layer can also be understood as a sub-model in the risk identification model used to generate risk identification problems; in one or more embodiments provided in this specification, the problem generation network layer can be a problem generation model, the risk identification network layer can be a risk identification model, and correspondingly, the risk identification model can be a risk identification model composed of the problem generation model and the risk identification model.

[0108] Following the previous example, after the third-party data query system, interaction system, and risk control system provide their respective data to the large model system, the large model system will input the above data into the large model. The large model contains a question generator, which can generate questions based on the above data to obtain user questions.

[0109] In the above embodiments, the question generation network layer in the risk identification model performs a question generation operation based on the data to be detected and the risk detection results to obtain the risk identification question corresponding to the target object, thereby facilitating accurate risk identification of the target object based on the risk identification question.

[0110] In one or more embodiments provided in this specification, before determining the risk identification problem corresponding to the target object based on the data to be detected and the risk detection result using the risk identification model, the method further includes:

[0111] Determine the risk identification model to be trained, as well as the training samples and sample labels corresponding to the risk identification model to be trained, wherein the training samples include risk identification question samples, answer result samples, data samples to be detected, and risk detection result samples of the sample object;

[0112] Using the risk identification model to be trained, the risk identification problem corresponding to the sample object is determined based on the data sample to be detected and the risk detection result sample.

[0113] Using the risk identification model to be trained, the risk of the sample object is identified based on the risk identification question sample, the answer result sample, the data sample to be detected, and the risk detection result sample, so as to obtain the risk identification result corresponding to the sample object;

[0114] Based on the risk identification problem corresponding to the sample object, the risk identification result corresponding to the sample object, and the sample label, the model parameters of the risk identification model to be trained are adjusted to obtain the trained risk identification model.

[0115] Here, the sample object can be understood as the target object used as a training sample; the risk identification problem sample can be understood as the risk identification problem used as a training sample; the answer result sample can be understood as the answer result used as a training sample; the data to be detected sample can be understood as the data to be detected used as a training sample; and the risk detection result sample can be understood as the risk detection result used as a training sample.

[0116] Following the example above, Figure 3 This is a schematic diagram of model training in an object processing method provided in one embodiment of this specification; based on Figure 3It can be seen that the training process for the data recognition model includes: 1. Selecting a base model; 2. Preparing question-and-answer format corpus; 3. Model training (fine-tuning); 4. Testing the model using test corpus; 5. Manually evaluating the model's performance; 6. Optimizing model parameters and adding corpus; 7. Iteration.

[0117] Based on the above, it can be seen that the core system of this large model involves two stages: model training and model generation. Specifically, regarding the model training stage, based on... Figure 3 As can be seen, the model training can be understood as Supervised Fine-tuning (SFT). Specifically, it can be understood as pre-selecting an initial large model, which can be understood as a model that has been pre-trained using sample data, such as a base model. Then, by providing a pre-provided accurate question-and-answer corpus (i.e., risk identification question samples, answer result samples, data samples to be detected, and risk detection result samples of the sample objects), the model is trained, thereby completing the fine-tuning of the base model and obtaining the trained risk identification model.

[0118] The preparation of the question-and-answer corpus (i.e., training data) used for model training can include the following methods: 1. Simulate customer information as realistically as possible, and ensure the information is as comprehensive as possible. It should be noted that this question-and-answer corpus needs to be assembled into a prompt and input into the model (i.e., the initial large model). Specific assembly methods can be found in the generation section. 2. The question portion of the corpus can be determined based on question-and-answer information from historical lending data. This part is crucial for inputting lending service knowledge; the questions should be as comprehensive, rich, and professional as possible. 3. Design user responses to the questions from the perspectives of both fraudulent and legitimate users. 4. The response portion of the corpus is also based on question-and-answer information from historical lending data to identify the reliability of user responses.

[0119] In the above embodiments, during model training, it is necessary to determine a pre-trained initial large model (i.e., the risk identification model to be trained), and fine-tune the initial large model using training samples and sample labels. This reduces the computational resources consumed by model training while making the model more closely resemble specific real-world application scenarios (such as lending scenarios), avoiding the problem of incompatibility between the trained risk identification model and real-world application scenarios. Furthermore, in the large model creation (model training) stage, by combining and fine-tuning the base large model with specific corpora, the generalization ability of the base large model trained on a large amount of corpus can be fully utilized, and combined with specific knowledge, the final usability can be achieved.

[0120] The risk identification model includes a problem generation network layer and a risk identification network layer;

[0121] The step of adjusting the model parameters of the risk identification model based on the risk identification problem corresponding to the sample object, the risk identification result corresponding to the sample object, and the sample label to obtain a trained risk identification model includes:

[0122] From the sample labels, determine a first sample label for training the problem generation network layer and a second sample label for training the risk identification network layer;

[0123] Based on the risk identification question corresponding to the sample object and the first sample label, the model parameters of the question generation network layer are adjusted, and based on the risk identification result corresponding to the sample object and the second sample label, the model parameters of the risk identification network layer are adjusted to obtain the trained risk identification model.

[0124] Here, the first sample label can be understood as the sample label used for training the question generation network layer. For example, the first sample label can be the risk identification question used as the sample label. Subsequently, the risk identification question output by the question generation network layer based on the training samples and the first sample label can be used to calculate the loss function, and the model parameters of the question generation network layer can be adjusted using the loss function obtained from the calculation.

[0125] The second sample label can be understood as the sample label used for training the risk identification network layer. For example, the second sample label can be the risk identification result used as the sample label. Subsequently, the risk identification network layer can calculate the loss function based on the risk identification result output by the training samples and the second sample label, and adjust the model parameters of the risk identification network layer through the loss function obtained by the calculation.

[0126] Following the previous example, after processing the training data using a question generator and an answer discriminator to obtain the corresponding questions and discriminant results, the questions and discriminant results used as sample labels are compared with the questions and discriminant results output by the model to calculate the loss function, thereby obtaining the loss function corresponding to the question generator and the loss function corresponding to the answer discriminator. Based on the two loss functions, the model parameters of the question generator and the answer discriminator are adjusted respectively to obtain the large model that has been trained.

[0127] In the above embodiments, the initial large model is fine-tuned using the loss function calculated from the training samples and sample labels, thereby reducing the computational resources consumed in model training while ensuring model performance, thus obtaining a better performing model.

[0128] Step 206: Provide the risk identification question to the information interaction unit corresponding to the target object, and receive the answer result returned by the information interaction unit for the risk identification question, wherein the answer result is generated by the information interaction unit based on the target object's answer behavior for the risk identification question.

[0129] The answering behavior operation can be understood as the operation of answering risk identification questions. This answering behavior operation can be voice answering behavior operation, text input answering behavior operation, video recording answering behavior operation, etc., and the risk identification questions can be answered through the above operations.

[0130] Specifically, after the risk identification unit uses the risk identification model to determine the risk identification problem, it sends the risk identification problem to the information interaction unit corresponding to the target object.

[0131] After receiving a risk identification question, the information interaction unit will provide the risk identification question to the target object. For example, the risk identification question can be displayed to the target object through an information display page in the information interaction unit; the information display page can be an application interface, a webpage, etc.

[0132] After the target object generates a response to the risk identification question, the information interaction unit returns the response to the risk identification unit.

[0133] Step 208: Using the risk identification model, perform risk identification on the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtain the risk identification result corresponding to the target object.

[0134] Specifically, the risk identification question, the answer result, the data to be tested, and the risk detection result are input into the risk identification model. The risk identification model can perform risk identification on the target object based on the risk identification question, the answer result, the data to be tested, and the risk detection result, thereby obtaining the risk identification result corresponding to the target object.

[0135] In one or more embodiments provided in this specification, the step of using the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtaining the risk identification result corresponding to the target object, includes:

[0136] The answer result, the data to be detected, and the risk detection result are input into the risk identification model;

[0137] Using the risk identification network layer in the risk identification model, a question generation operation is detected based on the risk identification question, the answer result, the data to be detected, and the risk detection result, to obtain the question generation operation detection result.

[0138] If it is determined that the problem generation operation will not be performed based on the problem generation operation detection result, the risk identification result corresponding to the target object is obtained based on the problem generation operation detection result.

[0139] The question generation operation detection result can be understood as a detection result used to determine whether to continue the question generation operation. This detection result can be a detection result obtained by the risk identification model based on semantic analysis of the risk identification question, the answer result, the data to be detected, and the risk detection result. For example, the question generation operation detection can be understood as a discriminative operation to determine whether to continue asking questions to the target object, and the question generation operation detection result can be understood as the discriminative result of the risk identification model for this discriminative operation.

[0140] Continuing with the previous example, the core system of the large model involves two stages: model training and model generation. The model generation stage can be understood as assembling a prompt by integrating user-generated data from three parties and the risk types returned by the risk control system, and then inputting this prompt into the model to obtain a question. After the user answers the question, the prompt is further assembled and input into the model again to obtain the judgment result; for details, please refer to... Figure 4 , Figure 4 This is a schematic diagram illustrating the application of a model in an object processing method provided in one embodiment of this specification, based on... Figure 4 As can be seen, on the information input side (i.e. the information input part), it is necessary to assemble various types of information, such as information actively filled in by the user, information on third-party data queried by the user, information on the user's environment, such as geographical location, information on potential fraud risks of the user as determined by the risk control system, information on the user's historical Q&A records, and prompt information, such as what aspects of questions can be asked. The assembly method of the Prompt can be seen in the example of "first question" below.

[0141] For the "Initial Question," the assembled Prompt could be: The user claims to be 25 years old and residing in Beijing. They have a father named Zhang Moumou, whose phone number is 133***. Third-party data shows the user has had three previous jobs: A, B, and C. Monthly housing provident fund contributions are 2000 yuan. The user's IP address is located in Shanghai. The risk control system determines the user may pose a risk of being a fraudulent intermediary. There are currently no records of the user's past questions. Please output: 1. Can you confirm whether the user poses a risk of being a fraudulent intermediary? 2. If you cannot confirm, please ask the user a question based on the above information. Hint: If the user is at risk of falsifying information, you can address details in the user's information or have the user answer; if the user is at risk of being a victim of telecommunications fraud, you can ask the user to clearly state the purpose of the loan and how they learned about this platform, etc.

[0142] After the initial question is asked, the user's answers will be assembled into a Prompt. This Prompt might contain the following information: The user claims to be 25 years old and residing in Beijing. They have a father named Zhang Moumou with the phone number 133***. Third-party data shows the user has had three previous jobs: A, B, and C. Their monthly housing provident fund contribution is 2000 yuan. The user's IP address is located in Shanghai. The risk control system determines the user may pose a risk of working for a fraudulent agency. The user's past Q&A history is: Question 1: What is your current city? Answer: I am currently in Beijing. Please output: 1. Can you confirm whether the user poses a risk of working for a fraudulent agency? 2. If you cannot confirm, please ask the user a question based on the above information.

[0143] After the above prompt is input into the large model (i.e., the core large model), the answer discriminator in the large model will make a judgment. If a question needs to be asked again, the large model will generate a corresponding question based on the input data and perform a question-and-answer operation with the user again; otherwise, the large model will output the corresponding judgment result to achieve risk identification of the user. For details on the output of the large model, please refer to [link to relevant documentation]. Figure 4 The model output section includes a prompt asking "Do you need further questions?" and the model's decision result.

[0144] Based on this, through multiple rounds of iteration, the final large model outputs a judgment result based on user feedback.

[0145] In the above embodiments, the object processing method provided in one or more embodiments of this specification proposes an automated anti-fraud identification system based on a large model. Since the large model itself has the ability to summarize data and analyze and reason, and can be continuously iterated, it can effectively and efficiently identify the fraudulent characteristics of users, making up for the shortcomings of the current stage of relying solely on structured data for anti-fraud. Furthermore, through the design and adjustment of the prompt layer, the content and method of the questions can be continuously changed, thereby providing a variety of countermeasures in the process of attacking and defending against fraudulent users.

[0146] In one or more embodiments provided in this specification, after the step of using the risk identification network layer in the risk identification model to perform question generation operation detection based on the answer result, the data to be detected, and the risk detection result, and obtaining the question generation operation detection result, the method further includes:

[0147] If it is determined to execute a question generation operation based on the detection results of the question generation operation, the risk identification model is used to determine the risk identification question corresponding to the target object based on the answer result, the data to be detected, and the risk detection results.

[0148] Continue executing the steps of providing the risk identification question to the information interaction unit corresponding to the target object and receiving the answer result returned by the information interaction unit for the risk identification question, until it is determined that the question generation operation will not be executed based on the detection result of the question generation operation.

[0149] Using the previous example, after inputting the above prompt into the large model, the answer discriminator in the large model will make a judgment. If the output of the answer discriminator in the large model is "further questions are needed", then it is necessary to ask questions again. By using the large model to generate corresponding questions based on the input data, the question-and-answer operation is performed on the user again. Based on this, through multiple rounds, the large model finally outputs the judgment result based on the user feedback.

[0150] In the above embodiments, the object processing method provided in one or more embodiments of this specification can effectively and efficiently identify the fraudulent characteristics of users because the large model itself has the ability to summarize data and analyze and reason, and can be continuously iterated, thus making up for the shortcomings of the current stage of relying solely on structured data for anti-fraud.

[0151] In one or more embodiments provided in this specification, after using the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtaining the risk identification result corresponding to the target object, the method further includes:

[0152] The risk identification result is provided to the information interaction unit corresponding to the target object, so that the information interaction unit can display the risk identification result.

[0153] The risk identification result can be understood as the result obtained after identifying whether there is any risk to the target object. The risk identification result can include high-risk identification results and low-risk identification results. For example, in the online credit anti-fraud scenario, a high-risk identification result can be understood as the user having a high fraud risk and the result of refusing credit; a low-risk identification result can be the user having a low fraud risk and the result of approving credit.

[0154] Specifically, after obtaining the risk identification result, the risk identification unit can provide the risk identification result to the information interaction unit corresponding to the target object. After receiving the risk identification result, the information interaction unit displays the risk identification result to the target object through the information interaction page, and / or displays the risk identification result to the risk handling object that performs risk handling for the target object through the information interaction page. The risk handling object can be credit risk detection personnel, credit risk review personnel, etc. in the credit scenario.

[0155] In addition, after obtaining the risk identification result, the risk identification unit can provide the risk identification result to the object interaction subunit corresponding to the target object; after receiving the risk identification result, the object interaction subunit displays the risk identification result to the target object and / or the risk handling object through the information interaction page.

[0156] In the above embodiments, the risk identification results are displayed through the information interaction unit, thereby achieving the purpose of timely and rapid display of risk identification results to the target object or the risk handling object.

[0157] In one or more embodiments provided in this specification, after using the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtaining the risk identification result corresponding to the target object, the method further includes:

[0158] The risk identification result is provided to the risk detection unit so that the risk detection unit generates an object processing result based on the risk identification result and displays the object processing result through the information interaction unit.

[0159] The act of displaying the object processing result through the information interaction unit can be understood as displaying the object processing result to the target object and / or the risk processing object through the information interaction unit.

[0160] Following the previous example, after the large model obtains the judgment result (i.e., the risk identification result), the judgment result is sent directly to the risk control core system (i.e., the risk control system). The risk control core system calculates information such as credit limit and interest rate based on the judgment result, and then displays the credit limit, interest rate and other information (i.e., the object processing result) to the user through the interactive system; or, the risk control core system generates a detailed result of credit rejection based on the judgment result (i.e., the object processing result), and then displays it to the user through the interactive system.

[0161] In the above embodiments, the object processing results are displayed through the risk detection unit and the information interaction unit, thereby realizing a detailed display of the results obtained after risk identification to the target object and / or the risk processing object.

[0162] In one or more embodiments provided in this specification, the target object is a credit application user, the data to be detected is credit application detection data, the risk detection result is a credit risk detection result, and the risk identification result is a credit risk identification result;

[0163] The step of using the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtaining the risk identification result corresponding to the target object, includes:

[0164] Using the risk identification model, based on the risk identification question, the credit applicant's answer, the credit application detection data, and the credit risk detection result, the credit applicant is identified for credit risk, and the corresponding credit risk identification result is obtained.

[0165] Among them, credit application detection data can be understood as data used to detect the risk of credit applicants during the credit application process; this credit application detection data can include credit application materials submitted by the user (i.e., the credit applicant), the user's historical credit application data, and the user's third-party data, etc.

[0166] Credit risk detection results can be understood as the results obtained by the risk detection unit after conducting risk detection on credit application data.

[0167] The risk identification question can be understood as a risk identification question determined for the credit applicant by using a risk identification model based on credit application detection data and credit risk detection results; correspondingly, the answer result is the result generated by the credit applicant's answer behavior in response to the risk identification question.

[0168] Credit risk identification results can be understood as the results obtained after identifying the credit risk of credit applicants. These results can be high-risk or low-risk.

[0169] Following the previous example, after obtaining the risk identification question and the answer, the risk identification question, the answer, the credit application information (such as information actively filled in by the user), the user's historical credit application data (such as historical Q&A records), the user's third-party data, and the fraud risk type (i.e., the credit risk detection result) are input into the large model. The large model is then used to identify the credit risk of the credit applicant based on the above data, and obtain the corresponding credit risk identification result for the credit applicant.

[0170] The object processing method provided in one or more embodiments of this specification in the above embodiments offers an online credit anti-fraud solution based on large-scale model-based automated credit review. Since large-scale models inherently possess long text generation and multi-turn dialogue capabilities, and their reasoning abilities are constantly improving, they can act as an automated, experienced, and continuously evolving advanced credit reviewer, significantly enhancing the anti-fraud identification capabilities of credit products, provided they are trained with appropriate sample data. By introducing a foundational large-scale model (LLM) and training it with training data, the large-scale model can replace humans as a 24 / 7 "credit reviewer," and by incorporating sufficient professional credit service knowledge, its professionalism surpasses that of most human credit reviewers, thereby solving this type of fraud problem.

[0171] In one or more embodiments provided in this specification, determining the target object's data to be detected and the risk detection results includes:

[0172] The client sends the target object's data to be detected and the risk detection result, wherein the target object's data to be detected and the risk detection result are sent by the information interaction unit in the client based on the target object's information upload operation.

[0173] It should be noted that the object processing methods provided in this manual can be applied to the server side, which is connected to the client side.

[0174] Specifically, the information interaction unit in this client can send the data to be detected and the risk detection results uploaded by the target object to the server based on the information upload operation of the target object;

[0175] After receiving the data to be detected and the risk detection results, the server will perform risk identification operations based on the data to be detected and the risk detection results to obtain the risk identification results of the target object.

[0176] After using the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtaining the risk identification result corresponding to the target object, the method further includes:

[0177] The risk identification result is sent to the client, so that the client sends the risk identification result to the target object.

[0178] Specifically, after the server uses the risk identification model to determine the risk identification result, it will send the risk identification result to the client for display, so that the client can display the risk identification result and achieve the purpose of quickly and timely displaying the risk identification result to the target object.

[0179] Sending the risk identification results to the client for display can be understood as sending the risk identification results to the client so that the client can display the risk identification results to the target object and / or the risk treatment object.

[0180] In the object processing method provided in one or more embodiments of this specification, during the object processing process, after the risk detection unit performs a first-stage risk detection on the data to be detected and obtains the risk detection result, in order to ensure the accuracy of risk identification, a risk identification model can be used to determine the corresponding risk identification question for the target object based on the data to be detected and the risk detection result. Then, based on the target object's answer to the risk identification question, the risk identification question, the data to be detected, and the risk detection result, a second-stage risk identification is performed to obtain an accurate risk identification result. This achieves accurate risk identification of the target object based on the data information provided by the target object, avoiding significant losses caused by the target object committing fraud by providing false data information.

[0181] The following is in conjunction with the appendix Figure 5 Taking the online credit anti-fraud application of the object processing method provided in this specification as an example, the object processing method will be further explained. Figure 5 This specification illustrates a flowchart of an object processing method according to an embodiment, based on... Figure 5 As can be seen, the object processing method in this specification is applied to an anti-fraud system, which consists of an interactive system, a third-party data query system, a large model core system, and a risk control core system. The anti-fraud processing method of this anti-fraud system specifically includes the following steps.

[0182] Step 502: The user submits the credit application materials online.

[0183] Specifically, users can apply for credit online through an interactive system, and in the process, they need to submit credit application materials online.

[0184] After receiving the credit application materials, the interactive system will provide them to the third-party data query system, the big model core system, and the risk control core system.

[0185] Step 504: Obtain the user's historical information and third-party data.

[0186] Specifically, after receiving the credit application materials submitted by the user, the third-party data query system will obtain the user's corresponding historical information (such as historical Q&A information, historical credit application information, etc.) from the local storage unit (such as local disk, local database) based on the user's personal information in the credit application materials.

[0187] Furthermore, the third-party data query system will obtain the corresponding third-party data (such as social security information, credit information, personal asset information, etc.) from third-party institutions based on the user's personal information in the loan application materials.

[0188] After obtaining the institution's historical information and third-party data, the third-party data query system will provide it to the core system of the big data model and the core system of risk control.

[0189] Step 506: The risk control decision engine segments customers.

[0190] Specifically, after receiving the user's credit application materials, the institution's historical information, and third-party data, the core risk control system will use the risk control decision engine to segment the user based on the data and determine the user's customer type (such as personal credit application type).

[0191] Step 508: Does the credit qualification meet the requirements?

[0192] Specifically, after segmenting customers, the core risk control system determines whether a user meets the credit eligibility criteria based on the user's credit application materials, the institution's historical information, and third-party data; if yes, proceed to step 512; if no, proceed to step 510.

[0193] Step 510: Deny credit and return the result.

[0194] Specifically, the core risk control system will notify users of the rejection of credit granting results through the interactive system if it determines that a user does not meet the credit granting qualifications.

[0195] Step 512: Is there a risk of fraud?

[0196] Specifically, the core risk control system, after determining that the user meets the credit qualification requirements, will judge whether the user has fraud risk based on the user's credit application materials, the institution's historical information, and third-party data; if yes, then proceed to step 514; if no, then proceed to step 530.

[0197] Step 514: Obtain the specific suspected fraud type.

[0198] Specifically, the core risk control system will identify specific suspected fraud types when it determines that a user is at risk of fraud; these fraud types include, but are not limited to: the risk of being a victim of telecommunications fraud, the risk of being a black market intermediary, and the risk of using false information.

[0199] Furthermore, this risk control core system will provide the fraud type to the largest model core system.

[0200] Step 516: Generate problems using the large model problem generator.

[0201] Specifically, the core system of this large model, after receiving the user's credit application materials, the institution's historical information, third-party data, fraud type, etc., assembles this data into a Prompt and inputs it into the large model. For example, the Prompt might be: The user claims to be 25 years old and residing in Beijing. They have a father named Zhang Moumou with the phone number 133***. Third-party data shows the user has had three previous jobs: A, B, and C. Monthly housing provident fund contributions are 2000 yuan. The user's IP address is in Shanghai. The risk control system determines the user may pose a risk of being a fraudulent intermediary. There are currently no records of historical issues with the user. Please output: 1. Can you confirm whether the user poses a risk of being a fraudulent intermediary? 2. If you cannot confirm, please ask the user a question based on the above information. Hint: If the user is at risk of falsifying documents, you can focus on details in the user's information or have the user answer; if the user is at risk of being a victim of telecommunications fraud, you can ask the user to clearly state the purpose of the loan and how they learned about this platform, etc.

[0202] Using the question generator in the large model, a first question is generated for the user based on the data, and this first question is then provided to the interactive system.

[0203] Step 518: Answer Question 1.

[0204] Specifically, after receiving the first question, the interactive system will display the first question to the user; the user will then perform a question-answering operation in the interactive system based on the first question (such as entering text, entering voice, or recording video), so that the interactive system can obtain the user's answer to the first question.

[0205] This interactive system will provide the answer to question one to the core system of the large model.

[0206] Step 520: The large model response discriminator makes a judgment.

[0207] Specifically, the core system of this large model, after receiving the user's answer to the first question, assembles the answer into a Prompt and inputs it into the large model. For example, the Prompt after a question might be: The user claims to be 25 years old and living in Beijing. They have a father named Zhang Moumou, whose phone number is 133***. Third-party data shows the user has had three previous jobs: A, B, and C. Monthly housing provident fund contributions are 2000 yuan. The user's IP address is in Shanghai. The risk control system determines the user may pose a risk of being a black market intermediary. The user's past Q&A is: Question 1: What is your current city? Answer: I am currently in Beijing. Please output: 1. Can you confirm whether the user poses a risk of being a black market intermediary? 2. If not, please ask the user a question based on the above information.

[0208] Using the answer discriminator in the large model, a judgment is made based on the Prompt to determine whether a second question needs to be generated for the user.

[0209] If it is determined that a second question needs to be generated for the user, the question generator in the large model is used to generate a corresponding second question for the user based on the data, and the second question is then provided to the interactive system.

[0210] Step 522: Answer Question 2.

[0211] Specifically, after receiving the second question, the interactive system will display the second question to the user; the user will then perform a question-answering operation in the interactive system based on the second question (such as entering text, voice, or recording video), so that the interactive system can obtain the user's answer to the second question.

[0212] This interactive system will provide the answer to question two to the core system of the largest model.

[0213] Step 524: Do you need to ask any further questions?

[0214] Specifically, after receiving the user's answer to the second question, the core system of the large model will assemble the answer into a Prompt and input the Prompt into the large model.

[0215] Using the answer discriminator in the large model, a judgment is made based on the Prompt to determine whether a third question needs to be generated for the user.

[0216] If it is determined that a third question needs to be generated for the user, the question generator in the large model is used to generate the corresponding third question for the user based on the data, and the interaction system is provided with the third question; through multiple rounds, the large model finally outputs the judgment result based on the user feedback.

[0217] Step 526: Pass or fail?

[0218] Specifically, using the answer discriminator in the large model, if it is determined that there is no need to generate a third question for the user, the result of whether the credit application is approved is output; if the result is approved, step 530 is executed; if the result is not approved, step 528 is executed.

[0219] Step 528: Deny credit and return result.

[0220] Specifically, the core system of the large model will notify the user of the credit rejection result through the interactive system if it determines that the user's credit application will not be approved.

[0221] Step 530: Calculate credit elements such as credit limit and interest rate.

[0222] Specifically, the core system of the large model will send the judgment result to the core risk control system when it determines that the user's credit application has been approved.

[0223] This core risk control system calculates credit limits, interest rates, and other credit elements for users based on the judgment result, and provides the credit limit, interest rate, and other credit elements, as well as the credit approval result, to the interactive system.

[0224] Step 532: Credit approval approved, return credit result.

[0225] Specifically, after receiving credit elements such as credit limit and interest rate, the interactive system will send these elements, along with the credit approval result, to the user.

[0226] The object processing methods in one or more embodiments provided in this specification provide an automated anti-fraud identification system based on a large model. Since the large model itself has the ability to summarize data and analyze and reason, and can be continuously iterated, it can effectively and efficiently identify the fraudulent characteristics of users, making up for the shortcomings of the current stage of relying solely on structured data for anti-fraud.

[0227] Because large models inherently possess the ability to generate long texts and engage in multi-turn dialogues, and their reasoning capabilities are constantly improving, they can serve as an automated, experienced, and continuously evolving senior credit reviewer, provided they are trained with appropriate data. This significantly enhances the anti-fraud identification capabilities of credit products.

[0228] Corresponding to the above method embodiments, this specification also provides object processing system embodiments. Figure 6 A schematic diagram of the structure of an object processing system provided in one embodiment of this specification is shown. Figure 6 As shown, the system includes: an information interaction unit 602, a risk detection unit 604, and a risk identification unit 606, wherein...

[0229] The information interaction unit 602 is configured to determine the data to be detected of the target object, provide the data to be detected to the risk detection unit 604 and the risk identification unit 606, send the risk identification question provided by the risk identification unit 606 to the target object, and return the answer result generated based on the target object's answer behavior to the risk identification question to the risk identification unit 606.

[0230] The risk detection unit 604 is configured to perform risk detection on the data to be detected, obtain the risk detection result of the target object, and provide the risk detection result to the risk identification unit 606.

[0231] The risk identification unit 606 is configured to determine the data to be detected and the risk detection result of the target object, use a risk identification model to determine the risk identification question corresponding to the target object based on the data to be detected and the risk detection result, provide the risk identification question to the information interaction unit 602 corresponding to the target object, and receive the answer result returned by the information interaction unit 602 for the risk identification question. Using the risk identification model, the risk identification question, the answer result, the data to be detected, and the risk detection result are used to identify the risk of the target object, and obtain the risk identification result corresponding to the target object.

[0232] Optionally, the information interaction unit 602 includes an object interaction subunit and a data acquisition subunit; the data to be detected includes object data and risk detection data;

[0233] The object interaction subunit is configured to obtain the object data of the target object based on the object data upload operation of the target object;

[0234] The data acquisition subunit is configured to determine the risk detection data corresponding to the target object from the data storage unit based on the object data.

[0235] Optionally, the risk identification unit 606 is further configured to:

[0236] The system receives the detection data of the target object determined by the information interaction unit 602, wherein the detection data includes object data of the target object and risk detection data, the object data is generated by the information interaction unit 602 based on the information upload operation of the target object, and the risk detection data is determined by the information interaction unit 602 from the data storage unit based on the object data;

[0237] The risk detection result of the target object determined by the risk detection unit 604 is received, wherein the risk detection result is obtained by the risk detection unit 604 through risk detection of the data to be detected.

[0238] Optionally, the risk identification unit 606 is further configured to:

[0239] The data to be detected and the risk detection results are input into the risk identification model;

[0240] By utilizing the question generation network layer in the risk identification model, a question generation operation is performed based on the data to be detected and the risk detection results to obtain the risk identification question corresponding to the target object.

[0241] Optionally, the risk identification unit 606 is further configured to:

[0242] The answer result, the data to be detected, and the risk detection result are input into the risk identification model;

[0243] Using the risk identification network layer in the risk identification model, a question generation operation is detected based on the risk identification question, the answer result, the data to be detected, and the risk detection result, to obtain the question generation operation detection result.

[0244] If it is determined that the problem generation operation will not be performed based on the problem generation operation detection result, the risk identification result corresponding to the target object is obtained based on the problem generation operation detection result.

[0245] Optionally, the risk identification unit 606 is further configured to:

[0246] If it is determined to execute a question generation operation based on the detection results of the question generation operation, the risk identification model is used to determine the risk identification question corresponding to the target object based on the answer result, the data to be detected, and the risk detection results.

[0247] Continue executing the steps of providing the risk identification question to the information interaction unit 602 corresponding to the target object and receiving the answer result returned by the information interaction unit 602 for the risk identification question, until it is determined that the question generation operation will not be executed based on the detection result of the question generation operation.

[0248] Optionally, the risk identification unit 606 is further configured to:

[0249] The risk identification result is provided to the information interaction unit 602 corresponding to the target object, so that the information interaction unit 602 can display the risk identification result.

[0250] Optionally, the risk identification unit 606 is further configured to:

[0251] The risk identification result is provided to the risk detection unit 604, so that the risk detection unit 604 generates an object processing result based on the risk identification result, and displays the object processing result through the information interaction unit 602.

[0252] Optionally, the target object is a credit application user, the data to be detected is credit application detection data, the risk detection result is a credit risk detection result, and the risk identification result is a credit risk identification result;

[0253] Optionally, the risk identification unit 606 is further configured to:

[0254] Using the risk identification model, based on the risk identification question, the credit applicant's answer, the credit application detection data, and the credit risk detection result, the credit applicant is identified for credit risk, and the corresponding credit risk identification result is obtained.

[0255] Optionally, the risk identification unit 606 is further configured to:

[0256] Determine the risk identification model to be trained, as well as the training samples and sample labels corresponding to the risk identification model to be trained, wherein the training samples include risk identification question samples, answer result samples, data samples to be detected, and risk detection result samples of the sample object;

[0257] Using the risk identification model to be trained, the risk identification problem corresponding to the sample object is determined based on the data sample to be detected and the risk detection result sample.

[0258] Using the risk identification model to be trained, the risk of the sample object is identified based on the risk identification question sample, the answer result sample, the data sample to be detected, and the risk detection result sample, so as to obtain the risk identification result corresponding to the sample object;

[0259] Based on the risk identification problem corresponding to the sample object, the risk identification result corresponding to the sample object, and the sample label, the model parameters of the risk identification model to be trained are adjusted to obtain the trained risk identification model.

[0260] Optionally, the risk identification model includes a problem generation network layer and a risk identification network layer;

[0261] Optionally, the risk identification unit 606 is further configured to:

[0262] From the sample labels, determine a first sample label for training the problem generation network layer and a second sample label for training the risk identification network layer;

[0263] Based on the risk identification question corresponding to the sample object and the first sample label, the model parameters of the question generation network layer are adjusted, and based on the risk identification result corresponding to the sample object and the second sample label, the model parameters of the risk identification network layer are adjusted to obtain the trained risk identification model.

[0264] Optionally, the risk identification unit 606 is further configured to:

[0265] The client sends the target object's data to be detected and the risk detection result, wherein the data to be detected and the risk detection result are sent by the information interaction unit 602 in the client based on the target object's information upload operation;

[0266] Optionally, the risk identification unit 606 is further configured to:

[0267] The risk identification results are sent to the client for display.

[0268] The object processing system provided in one or more embodiments of this specification, during the object processing process, after the risk detection unit performs a first-stage risk detection on the data to be detected and obtains the risk detection results, in order to ensure the accuracy of risk identification, can use a risk identification model to determine the corresponding risk identification question for the target object based on the data to be detected and the risk detection results. Then, based on the target object's answer to the risk identification question, the risk identification question, the data to be detected, and the risk detection results, a second-stage risk identification is performed to obtain an accurate risk identification result. This enables accurate risk identification of the target object based on the data information provided by the target object, avoiding significant losses caused by the target object committing fraud by providing false data information.

[0269] The above is an illustrative scheme of an object processing system according to this embodiment. It should be noted that the technical solution of this object processing system and the technical solution of the object processing method described above belong to the same concept. For details not described in detail in the technical solution of the object processing system, please refer to the description of the technical solution of the object processing method described above.

[0270] This specification provides a credit processing method according to one or more embodiments, which specifically includes the following steps.

[0271] The data to be tested for credit users and the credit risk detection results are determined, wherein the credit risk detection results are obtained by using a credit risk detection unit to perform credit risk detection on the data to be tested;

[0272] Using a credit risk identification model, the credit risk identification problem corresponding to the credit user is determined based on the data to be detected and the credit risk detection results.

[0273] The credit risk identification question is provided to the information interaction unit corresponding to the credit user, and the answer result returned by the information interaction unit for the credit risk identification question is received, wherein the answer result is generated by the information interaction unit based on the credit user's answer behavior for the credit risk identification question;

[0274] Using the credit risk identification model, the credit risk of the credit user is identified based on the credit risk identification question, the answer result, the data to be detected, and the credit risk detection result, thereby obtaining the credit risk identification result corresponding to the credit user.

[0275] Wherein, the credit user can be understood as the target object in the above-described object processing method embodiment, the credit risk detection result can be understood as the risk detection result in the above-described object processing method embodiment, the credit risk identification model can be understood as the risk identification model in the above-described object processing method embodiment, the credit risk identification problem can be understood as the risk identification problem in the above-described object processing method embodiment, and the credit risk identification result can be understood as the risk identification result in the above-described object processing method embodiment.

[0276] The credit processing method in one or more embodiments provided in this specification, during the credit processing process, after the credit risk detection unit performs a first-stage credit risk detection on the data to be detected and obtains the credit risk detection result, in order to ensure the accuracy of credit risk identification, can use a credit risk identification model to determine the corresponding credit risk identification question for the credit user based on the data to be detected and the credit risk detection result. Then, based on the credit user's answer to the credit risk identification question, the credit risk identification question, the data to be detected, and the credit risk detection result, a second-stage credit risk identification is performed to obtain an accurate credit risk identification result. This achieves accurate credit risk identification for credit users based on the data information provided by the credit user, avoiding significant losses caused by credit users committing fraud by providing false data information.

[0277] The above is an illustrative scheme of a credit processing method according to this embodiment. It should be noted that the technical solution of this credit processing method and the technical solution of the object processing method described above belong to the same concept. For details not described in detail in the technical solution of the credit processing method, please refer to the description of the technical solution of the object processing method described above.

[0278] Corresponding to the above method embodiments, this specification also provides embodiments of an object processing apparatus. Figure 7 A schematic diagram of an object processing apparatus according to one embodiment of this specification is shown. Figure 7 As shown, the device includes:

[0279] The information determination module 702 is configured to determine the data to be detected and the risk detection result of the target object, wherein the risk detection result is obtained by performing risk detection on the data to be detected using a risk detection unit;

[0280] Problem determination module 704 is configured to use a risk identification model to determine the risk identification problem corresponding to the target object based on the data to be detected and the risk detection result.

[0281] The response receiving module 706 is configured to provide the risk identification question to the information interaction unit corresponding to the target object, and receive the response result returned by the information interaction unit for the risk identification question, wherein the response result is generated by the information interaction unit based on the target object's response behavior to the risk identification question;

[0282] The risk identification module 708 is configured to use the risk identification model to identify the risk of the target object based on the risk identification question, the answer result, the data to be detected, and the risk detection result, and obtain the risk identification result corresponding to the target object.

[0283] Optionally, the information determination module 702 is further configured to:

[0284] The system receives the detection data of the target object determined by the information interaction unit, wherein the detection data includes object data of the target object and risk detection data, the object data is generated by the information interaction unit based on the information upload operation of the target object, and the risk detection data is determined by the information interaction unit from the data storage unit based on the object data;

[0285] Receive the risk detection result of the target object determined by the risk detection unit, wherein the risk detection result is obtained by the risk detection unit through risk detection of the data to be detected.

[0286] Optionally, the problem determination module 704 is further configured to:

[0287] The data to be detected and the risk detection results are input into the risk identification model;

[0288] By utilizing the question generation network layer in the risk identification model, a question generation operation is performed based on the data to be detected and the risk detection results to obtain the risk identification question corresponding to the target object.

[0289] Optionally, the risk identification module 708 is further configured to:

[0290] The answer result, the data to be detected, and the risk detection result are input into the risk identification model;

[0291] Using the risk identification network layer in the risk identification model, a question generation operation is detected based on the risk identification question, the answer result, the data to be detected, and the risk detection result, to obtain the question generation operation detection result.

[0292] If it is determined that the problem generation operation will not be performed based on the problem generation operation detection result, the risk identification result corresponding to the target object is obtained based on the problem generation operation detection result.

[0293] Optionally, the risk identification module 708 is further configured to:

[0294] If it is determined to execute a question generation operation based on the detection results of the question generation operation, the risk identification model is used to determine the risk identification question corresponding to the target object based on the answer result, the data to be detected, and the risk detection results.

[0295] Continue executing the steps of providing the risk identification question to the information interaction unit corresponding to the target object and receiving the answer result returned by the information interaction unit for the risk identification question, until it is determined that the question generation operation will not be executed based on the detection result of the question generation operation.

[0296] Optionally, the object processing apparatus further includes a result sending module, configured to:

[0297] The risk identification result is provided to the information interaction unit corresponding to the target object, so that the information interaction unit can display the risk identification result.

[0298] Optionally, the result sending module is further configured to:

[0299] The risk identification result is provided to the risk detection unit so that the risk detection unit generates an object processing result based on the risk identification result and displays the object processing result through the information interaction unit.

[0300] Optionally, the target object is a credit application user, the data to be detected is credit application detection data, the risk detection result is a credit risk detection result, and the risk identification result is a credit risk identification result;

[0301] The risk identification module 708 is also configured to:

[0302] Using the risk identification model, based on the risk identification question, the credit applicant's answer, the credit application detection data, and the credit risk detection result, the credit applicant is identified for credit risk, and the corresponding credit risk identification result is obtained.

[0303] Optionally, the object processing apparatus further includes a model training module, configured to:

[0304] Determine the risk identification model to be trained, as well as the training samples and sample labels corresponding to the risk identification model to be trained, wherein the training samples include risk identification question samples, answer result samples, data samples to be detected, and risk detection result samples of the sample object;

[0305] Using the risk identification model to be trained, the risk identification problem corresponding to the sample object is determined based on the data sample to be detected and the risk detection result sample.

[0306] Using the risk identification model to be trained, the risk of the sample object is identified based on the risk identification question sample, the answer result sample, the data sample to be detected, and the risk detection result sample, so as to obtain the risk identification result corresponding to the sample object;

[0307] Based on the risk identification problem corresponding to the sample object, the risk identification result corresponding to the sample object, and the sample label, the model parameters of the risk identification model to be trained are adjusted to obtain the trained risk identification model.

[0308] Optionally, the risk identification model includes a problem generation network layer and a risk identification network layer;

[0309] The model training module is also configured as follows:

[0310] From the sample labels, determine a first sample label for training the problem generation network layer and a second sample label for training the risk identification network layer;

[0311] Based on the risk identification question corresponding to the sample object and the first sample label, the model parameters of the question generation network layer are adjusted, and based on the risk identification result corresponding to the sample object and the second sample label, the model parameters of the risk identification network layer are adjusted to obtain the trained risk identification model.

[0312] Optionally, the information determination module 702 is further configured to:

[0313] The client sends the target object's data to be detected and the risk detection result, wherein the target object's data to be detected and the risk detection result are sent by the information interaction unit in the client based on the target object's information upload operation;

[0314] The result sending module is further configured to:

[0315] The risk identification results are sent to the client for display.

[0316] The object processing apparatus in one or more embodiments provided in this specification, during the object processing process, after the risk detection unit performs a first-stage risk detection on the data to be detected and obtains the risk detection result, in order to ensure the accuracy of risk identification, can use a risk identification model to determine the corresponding risk identification question for the target object based on the data to be detected and the risk detection result. Then, based on the target object's answer to the risk identification question, the risk identification question, the data to be detected, and the risk detection result, a second-stage risk identification is performed to obtain an accurate risk identification result. This enables accurate risk identification of the target object based on the data information provided by the target object, avoiding significant losses caused by the target object committing fraud by providing false data information.

[0317] The above is an illustrative scheme of an object processing apparatus according to this embodiment. It should be noted that the technical solution of this object processing apparatus and the technical solution of the object processing method described above belong to the same concept. For details not described in detail in the technical solution of the object processing apparatus, please refer to the description of the technical solution of the object processing method described above.

[0318] Figure 8 A structural block diagram of a computing device 800 according to one embodiment of this specification is shown. The components of the computing device 800 include, but are not limited to, a memory 810 and a processor 820. The processor 820 is connected to the memory 810 via a bus 830, and a database 850 is used to store data.

[0319] The computing device 800 also includes an access device 840, which enables the computing device 800 to communicate via one or more networks 860. Examples of such networks include a Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the Internet. Access device 840 may include one or more of any type of wired or wireless network interface (e.g., network interface card (NIC)), such as IEEE 802.11 Wireless Local Area Network (WLAN) interface, Wi-MAX (Worldwide Interoperability for Microwave Access) interface, Ethernet interface, Universal Serial Bus (USB) interface, cellular network interface, Bluetooth interface, and Near Field Communication (NFC).

[0320] In one embodiment of this specification, the above-described components of the computing device 800 and Figure 8 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 8 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0321] The computing device 800 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 800 can also be a mobile or stationary server.

[0322] The processor 820 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described object processing method or credit processing method.

[0323] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the computing device embodiments are basically similar to the object processing method or credit processing method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the object processing method or credit processing method embodiments.

[0324] An embodiment of this specification also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the object processing method or credit processing method described above.

[0325] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the computer-readable storage medium embodiments are relatively simple in description because they are fundamentally similar to the object processing method or credit processing method embodiments; relevant parts can be referred to in the descriptions of the object processing method or credit processing method embodiments.

[0326] An embodiment of this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described object processing method or credit processing method.

[0327] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product belongs to the same concept as the technical solution of the object processing method or credit processing method described above. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the object processing method or credit processing method described above.

[0328] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0329] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0330] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0331] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0332] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A subject processing method, comprising: determining to-be-detected data of a target subject and a risk detection result, wherein the risk detection result is obtained by performing risk detection on the to-be-detected data by using a risk detection unit; determining, by using a risk identification model, a risk identification question corresponding to the target subject according to the to-be-detected data and the risk detection result; providing the risk identification question to an information interaction unit corresponding to the target subject, and receiving a response result returned by the information interaction unit for the risk identification question, wherein the response result is generated by the information interaction unit based on a response behavior operation of the target subject for the risk identification question; performing risk identification on the target subject according to the risk identification question, the response result, the to-be-detected data and the risk detection result by using the risk identification model, and obtaining a risk identification result corresponding to the target subject.

2. The subject processing method of claim 1, wherein the determining to-be-detected data of a target subject and a risk detection result comprises: receiving the to-be-detected data of the target subject determined based on the information interaction unit, wherein the to-be-detected data comprises subject data of the target subject and risk detection data, the subject data is generated by the information interaction unit based on an information uploading operation of the target subject, and the risk detection data is determined by the information interaction unit from a data storage unit based on the subject data; receiving the risk detection result of the target subject determined based on the risk detection unit, wherein the risk detection result is obtained by the risk detection unit by performing risk detection on the to-be-detected data.

3. The subject processing method of claim 1, wherein the determining, by using a risk identification model, a risk identification question corresponding to the target subject according to the to-be-detected data and the risk detection result comprises: inputting the to-be-detected data and the risk detection result into the risk identification model; performing question generation operation according to the to-be-detected data and the risk detection result by using a question generation network layer in the risk identification model, and obtaining the risk identification question corresponding to the target subject.

4. The subject processing method of claim 1, wherein the performing risk identification on the target subject according to the risk identification question, the response result, the to-be-detected data and the risk detection result by using the risk identification model, and obtaining a risk identification result corresponding to the target subject comprises: inputting the response result, the to-be-detected data and the risk detection result into the risk identification model; performing question generation operation detection according to the risk identification question, the response result, the to-be-detected data and the risk detection result by using a risk identification network layer in the risk identification model, and obtaining a question generation operation detection result; in a case where it is determined not to perform question generation operation according to the question generation operation detection result, obtaining the risk identification result corresponding to the target subject according to the question generation operation detection result.

5. The object processing method of claim 4, further comprising: in a case where it is determined to perform the question generation operation according to the question generation operation detection result, determining, by using the risk identification model, a risk identification question corresponding to the target object according to the answer result, the to-be-detected data, and the risk detection result; continuing to perform the step of providing the risk identification question to the information interaction unit corresponding to the target object and receiving the answer result returned by the information interaction unit for the risk identification question until it is determined not to perform the question generation operation according to the question generation operation detection result.

6. The object processing method of claim 1, further comprising: providing the risk identification result to the information interaction unit corresponding to the target object, so that the information interaction unit displays the risk identification result.

7. The object processing method of claim 1, after the step of determining, by using the risk identification model, a risk identification result corresponding to the target object according to the risk identification question, the answer result, the to-be-detected data, and the risk detection result, further comprising: providing the risk identification result to the risk detection unit, so that the risk detection unit generates an object processing result based on the risk identification result and displays the object processing result through the information interaction unit.

8. The object processing method of claim 1, wherein the target object is a credit application user, the to-be-detected data is credit application detection data, the risk detection result is a credit risk detection result, and the risk identification result is a credit risk identification result; the step of determining, by using the risk identification model, a risk identification result corresponding to the target object according to the risk identification question, the answer result, the to-be-detected data, and the risk detection result, comprises: performing, by using the risk identification model, credit risk identification on the credit application user according to the risk identification question, the answer result of the credit application user, the credit application detection data, and the credit risk detection result, to obtain a credit risk identification result corresponding to the credit application user.

9. The object processing method of claim 1, before the step of determining, by using the risk identification model, a risk identification question corresponding to the target object according to the to-be-detected data and the risk detection result, further comprising: determining a to-be-trained risk identification model, and a training sample and a sample label corresponding to the to-be-trained risk identification model, wherein the training sample comprises a risk identification question sample, an answer result sample, a to-be-detected data sample, and a risk detection result sample of a sample object; determining, by using the to-be-trained risk identification model, a risk identification question corresponding to the sample object according to the to-be-detected data sample and the risk detection result sample. obtaining a risk identification result corresponding to the sample object by using the risk identification model to be trained to perform risk identification on the sample object according to the risk identification question sample, the answer result sample, the data sample to be detected, and the risk detection result sample; adjusting model parameters of the risk identification model to be trained based on the risk identification question corresponding to the sample object, the risk identification result corresponding to the sample object, and the sample label, to obtain the trained risk identification model.

10. The object processing method of claim 9, wherein the risk identification model comprises a question generation network layer and a risk identification network layer. The method further comprises: determining a first sample label for training the question generation network layer and a second sample label for training the risk identification network layer from the sample label; adjusting model parameters of the question generation network layer based on the risk identification question corresponding to the sample object and the first sample label, and adjusting model parameters of the risk identification network layer based on the risk identification result corresponding to the sample object and the second sample label, to obtain the trained risk identification model.

11. The object processing method of claim 1, wherein the method further comprises: determining the data to be detected and the risk detection result of the target object sent by the client, wherein the data to be detected and the risk detection result are sent by the information interaction unit in the client based on an information uploading operation of the target object; after the risk identification result of the target object is obtained by using the risk identification model to perform risk identification on the target object according to the risk identification question, the answer result, the data to be detected, and the risk detection result, the method further comprises: sending the risk identification result to the client for display.

12. An object processing system, comprising an information interaction unit, a risk detection unit, and a risk identification unit, wherein: the information interaction unit is configured to determine data to be detected of a target object, provide the data to be detected to the risk detection unit and the risk identification unit, send a risk identification question provided by the risk identification unit to the target object, and return an answer result generated based on an answer behavior operation of the target object to the risk identification question to the risk identification unit; the risk detection unit is configured to perform risk detection on the data to be detected, obtain a risk detection result of the target object, and provide the risk detection result to the risk identification unit; and the risk identification unit is configured to perform risk identification on the target object based on the risk identification question, the risk detection result, and the sample label, to obtain a risk identification result corresponding to the target object. The risk identification unit is configured to determine to-be-detected data of a target object and a risk detection result, determine a risk identification question corresponding to the target object according to the to-be-detected data and the risk detection result by using a risk identification model, provide the risk identification question to an information interaction unit corresponding to the target object, receive a returned answer result to the risk identification question returned by the information interaction unit, perform risk identification on the target object according to the risk identification question, the answer result, the to-be-detected data and the risk detection result by using the risk identification model, and obtain a risk identification result corresponding to the target object.

13. The object processing system of claim 12, wherein the information interaction unit comprises an object interaction subunit and a data acquisition subunit; and the to-be-detected data comprises object data and risk detection data. The object interaction subunit is configured to acquire the object data of the target object based on an object data uploading operation of the target object. The data acquisition subunit is configured to determine the risk detection data corresponding to the target object from a data storage unit based on the object data.

14. A credit processing method, comprising: determining to-be-detected data of a credit user and a credit risk detection result, wherein the credit risk detection result is obtained by performing credit risk detection on the to-be-detected data by using a credit risk detection unit; determining a credit risk identification question corresponding to the credit user according to the to-be-detected data and the credit risk detection result by using a credit risk identification model; providing the credit risk identification question to an information interaction unit corresponding to the credit user, and receiving a returned answer result to the credit risk identification question returned by the information interaction unit, wherein the answer result is generated by the information interaction unit based on an answer behavior operation of the credit user to the credit risk identification question; performing credit risk identification on the credit user according to the credit risk identification question, the answer result, the to-be-detected data and the credit risk detection result by using the credit risk identification model, and obtaining a credit risk identification result corresponding to the credit user.

15. A computing device, comprising: a memory and a processor; the memory is configured to store computer programs / instructions, and the processor is configured to execute the computer programs / instructions, so as to implement the steps of the object processing method of any one of claims 1 to 11 or the credit processing method of claim 14.

16. A computer readable storage medium, which stores computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the steps of the object processing method of any one of claims 1 to 11 or the credit processing method of claim 14.

17. A computer program product, which comprises computer programs / instructions, and the computer programs / instructions are executed by a processor to implement the steps of the object processing method of any one of claims 1 to 11 or the credit processing method of claim 14.

Images