Financial account processing method and device based on large model, and electronic equipment

Abstract

The invention discloses a financial account processing method and device based on a large model and electronic equipment, and relates to the field of financial science and technology, and the method comprises the steps: determining a data collection strategy based on a historical risk score of a target account; after authorization information of a user of the target account is received, first data is collected based on a data collection strategy, and the first data is at least one of a verification code, a contact information verification result, a face image and transaction verification data of the user; performing feature extraction on the first data through the target model to obtain a first feature of the target account; determining a target risk score of the target account based on the first feature and the second feature of the target account; and under the condition that the target risk score is smaller than a preset risk threshold value, performing unfreezing processing on the target account. According to the method and the device, the technical problem of low user experience caused by the condition that the risk detection accuracy of the financial account is low and the financial account is frozen by mistake in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more specifically, to a method, apparatus, and electronic device for processing financial accounts based on a large model. Background Technology

[0002] With the rapid development of digital finance, financial institutions are facing unprecedented risk challenges. As the hub of fund transfers, the security of financial accounts directly relates to the safety of users' assets. Existing traditional risk detection models primarily rely on expert rules to examine structured data of financial accounts. For example:

[0003] Example 1: Score the data features corresponding to structured data. If the score exceeds the threshold, a risk warning is triggered, and the financial account is frozen.

[0004] Example 2: If a user has overdue loan payments ≥ 3 times in the past 3 months, the user's loan application function in their financial account will be temporarily frozen.

[0005] While existing technologies can identify and warn of account risks to some extent, the quality of expert rules is limited by the level of the experts, making it difficult to cover long-tail scenarios and unknown risks. Furthermore, the maintenance of expert rules relies on manual operation, resulting in long iteration cycles and limited accuracy in risk detection. This can lead to the accidental freezing of users' financial accounts, affecting their daily financial transactions and reducing user experience. The process for unfreezing frozen accounts is complex and time-consuming, easily leading to customer complaints and further degrading user experience.

[0006] There is currently no effective solution to the above problems. Summary of the Invention

[0007] This application provides a financial account processing method, apparatus, and electronic device based on a large model, to at least solve the technical problems in the prior art of low accuracy in risk detection of financial accounts, the occurrence of false freezing of financial accounts, and thus a poor user experience.

[0008] According to one aspect of this application, a financial account processing method based on a large model is provided, comprising: determining a data collection strategy based on the historical risk score of a target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected; after receiving authorization information from the user of the target account, collecting first data based on the data collection strategy, wherein the first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data, and the contact information verification result is the risk detection result of the operator on the user's contact information; extracting features from the first data through a target model to obtain a first feature of the target account, wherein the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, where X is a positive integer; determining a target risk score of the target account based on the first feature and a second feature of the target account, wherein the second feature is used to characterize the user profile and the user's operational behavior sequence on the target account; and unfreezing the target account if the target risk score is less than a preset risk threshold.

[0009] Optionally, before determining the data collection strategy based on the target account's historical risk score and historical reasons for freezing, the financial account processing method based on the large model also includes: scanning the account status of financial accounts in a preset database based on a preset time period; and if the financial account is in a frozen state, using the financial account as the target account.

[0010] Optionally, the step of determining the data collection strategy based on the historical risk score of the target account includes: if the historical risk score is less than or equal to a first preset score, using the first preset strategy as the data collection strategy, wherein the first preset strategy is used to collect the user's facial image; if the historical risk score is greater than the first preset score and less than or equal to a second preset score, using the second preset strategy as the data collection strategy, wherein the second preset strategy is used to collect the user's verification code and contact information verification results, and the second preset score is greater than the first preset score; if the historical risk score is greater than the second preset score, using the third preset strategy as the data collection strategy, wherein the third preset strategy is used to collect the user's verification code, contact information verification results, facial image, and transaction verification data.

[0011] Optionally, after determining the data collection strategy based on the target account's historical risk score and historical reasons for account freezing, the financial account processing method based on the large model further includes: filtering the target account's account identifier in a preset database to obtain the user identifier of the user to whom the target account belongs; filtering the user identifier in the preset database to obtain the user's contact information; generating a prompt message through the data collection strategy, and sending the prompt message to the user through the contact information, wherein the prompt message is used to prompt the user to re-provide the user data corresponding to the data collection strategy.

[0012] Optionally, the step of extracting features from the first data using a target model to obtain the first feature of the target account includes: performing data cleaning and data desensitization on the text data in the first data using the target model to obtain first text data, wherein data cleaning is used to remove noise data from the text data, and data desensitization is used to desensitize sensitive data in the text data; performing image cropping and grayscale processing on the image data in the first data to obtain first image data; extracting features from the first text data and the first image data to obtain first text features and first image features, wherein the first text features are used to characterize the semantic information of keywords in the first text data, and the first image features are used to characterize the pixel information of key points in the first image data; and fusing the first text features and the first image features to obtain the first feature of the target account.

[0013] Optionally, before determining the target risk score of the target account based on the first and second features of the target account, the financial account processing method based on the large model further includes: generating a target statement based on the account identifier of the target account and the user identifier of the user, wherein the target statement is a database operation statement; after receiving the user's authorization information, filtering from a preset database based on the target statement to obtain the user's historical operation records on the target account within a preset historical time period, as well as the user's user data and asset data; extracting features from the user data and asset data to obtain user profile features, and extracting features from the historical operation records to obtain device fingerprint features and user temporal behavior features; and using the user profile features, device fingerprint features, and user temporal behavior features as the second feature.

[0014] Optionally, the step of determining the target risk score of the target account based on the first and second characteristics of the target account includes: determining the first risk score corresponding to the first characteristic and the second risk score corresponding to the second characteristic based on a preset scoring rule, wherein the risk score is used to characterize the similarity between the characteristics of the target account and the characteristics of historical financial accounts with transaction risks; and performing a weighted summation of the first risk score and the second risk score to obtain the target risk score.

[0015] According to another aspect of this application, a financial account processing device based on a large model is also provided, comprising: a strategy determination unit, configured to determine a data collection strategy based on the historical risk score of a target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected; a data collection unit, configured to collect first data based on the data collection strategy after receiving authorization information from the user of the target account, wherein the first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data, and the contact information verification result is the risk detection result of the operator on the user's contact information; a first extraction unit, configured to extract features from the first data through a target model to obtain a first feature of the target account, wherein the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, where X is a positive integer; a scoring determination unit, configured to determine a target risk score of the target account based on the first feature and a second feature of the target account, wherein the second feature is used to characterize the user profile of the user and the user's operational behavior sequence on the target account; and an account unfreezing unit, configured to unfreeze the target account if the target risk score is less than a preset risk threshold.

[0016] According to another aspect of this application, a computer program product is also provided, which stores a computer program, wherein, when the computer program is running, it controls the computer program product to execute any of the above-mentioned financial account processing methods based on large models.

[0017] According to another aspect of this application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the large-model-based financial account processing method described above.

[0018] In this application, a data collection strategy is first determined based on the historical risk score of the target account, where the target account is a frozen financial account. The data collection strategy is used to determine the types of user data to be collected. After receiving the authorization information from the user of the target account, first data is collected based on the data collection strategy. The first data includes at least one of the user's verification code, contact information verification result, facial image, and transaction verification data. The contact information verification result is the risk detection result of the operator on the user's contact information. The first data is then used to extract features from the first data to obtain the first feature of the target account. The target model is a large language model trained on unstructured user data corresponding to X financial accounts, where X is a positive integer. Based on the first and second features of the target account, a target risk score for the target account is determined. The second feature is used to characterize the user profile and the user's operational behavior sequence on the target account. If the target risk score is less than a preset risk threshold, the target account is unfrozen.

[0019] As described above, this application intelligently adjusts the type and depth of user data collection based on the historical risk scores of frozen accounts. It collects unstructured user data (i.e., the first data) based on a data collection strategy. Subsequently, this application uses a large language model to extract features from the collected unstructured user data and combines this with user profiles and operational behavior sequences extracted from structured data to form a more comprehensive account feature set. Following this, the frozen accounts are re-risk-scored based on the features in the account feature set, thereby improving the accuracy of risk detection for frozen accounts. When the new risk score is lower than a preset risk threshold, the financial account is automatically unfrozen, thus solving the problem of erroneous freezing of financial accounts due to misjudgment of risk.

[0020] Therefore, this application adopts a data collection strategy based on the historical risk score of frozen accounts. By extracting features from the collected unstructured user data and the original structured data and re-scoring the risk, it achieves the goal of accurately identifying the user's financial risk level. This improves the accuracy of risk identification of user financial accounts and solves the technical problems of low accuracy in risk detection of financial accounts and false freezing of financial accounts in existing technologies, which leads to poor user experience. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0022] Figure 1This is a hardware structure block diagram of an optional computer terminal (or mobile device) for implementing a large-model-based financial account processing method according to an embodiment of this application.

[0023] Figure 2 This is a flowchart of an optional large-model-based financial account processing method according to an embodiment of this application;

[0024] Figure 3 This is a flowchart of an optional data acquisition method according to an embodiment of this application;

[0025] Figure 4 This is a schematic diagram of an optional large-model-based financial account processing device according to an embodiment of this application;

[0026] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0028] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] It should also be noted that all relevant information (including but not limited to information used for display and analysis) and data (including but not limited to collected structured and unstructured user data) involved in this application are information and data authorized by the user or fully authorized by all parties. For example, this system has an interface with the relevant user or organization. Before obtaining relevant information, it needs to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving consent from the aforementioned user or organization.

[0030] Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant information and data involved in this application all comply with the relevant laws, regulations, and standards of the relevant regions, and necessary confidentiality measures have been taken. This application does not violate public order and good morals. In addition, this application provides a corresponding operation entry point for users to choose to agree to or refuse authorization. If the user chooses to refuse authorization, the corresponding expert decision-making process will be initiated.

[0031] According to an embodiment of this application, an embodiment of a financial account processing method based on a large model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0032] The methods and embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 This is a hardware structure block diagram of an optional computer terminal (or mobile device) for implementing a large-model-based financial account processing method, according to an embodiment of this application. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0033] Optionally, it should be noted that the aforementioned one or more processors 102 and / or other data processing circuits may generally be referred to herein as "data processing circuitry". This data processing circuitry may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be wholly or partially integrated into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, this data processing circuitry serves as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0034] Optionally, the memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the large-model-based financial account processing method in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned large-model-based financial account processing method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0035] Optionally, the transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0036] Alternatively, the display can be configured as a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0037] Optionally, under the above operating environment, this application provides a large-model-based financial account processing system (hereinafter referred to as the processing system) for executing the large-model-based financial account processing method of this application. Figure 2 This is a flowchart of an optional large-model-based financial account processing method according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0038] Step S201: Determine the data collection strategy based on the historical risk score of the target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected.

[0039] Optionally, historical risk score refers to the quantitative representation of the risk level of an account determined by a financial institution at a historical point in time based on relevant account data of the frozen account.

[0040] Optionally, the data collection strategy specifies the data type and range of unstructured user data that need to be collected in this processing. For example, if user A has a low historical risk score, then only the user's SMS verification code needs to be collected.

[0041] Optionally, the scope and types of subsequent unstructured user data are determined based on historical risk scores, which improves the targeting and efficiency of user data collection by the subsequent processing system. This avoids excessive and unnecessary data collection from low-risk financial accounts, reducing the frequency of user data collection, improving user experience, and reducing the amount of data that the subsequent processing system needs to process.

[0042] Step S202: After receiving the authorization information of the user of the target account, first data is collected based on the data collection strategy. The first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data. The contact information verification result is the risk detection result of the operator on the user's contact information.

[0043] Optionally, the processing system only collects unstructured user data after receiving authorization information from the user of the target account. User authorization ensures the compliance of the data collection process of the processing system. Furthermore, the collection of the first data provides a more comprehensive and multi-dimensional data foundation for the subsequent processing system to re-detect account risks using the large control model, thereby improving the accuracy of the account risks obtained in the subsequent re-detection.

[0044] Step S203: Extract features from the first data using the target model to obtain the first feature of the target account. The target model is a large language model trained on unstructured user data corresponding to X financial accounts, where X is a positive integer.

[0045] Optionally, the processing system controls the large model to extract features from the unstructured first data, which can capture complex patterns and correlations in the first data, improve the accuracy of the extracted first features, and at the same time improve the accuracy and coverage of subsequent risk assessment based on the first features, so that the processing system can better adapt to long-tail scenarios in the risk detection process.

[0046] Step S204: Based on the first and second features of the target account, determine the target risk score of the target account, wherein the second feature is used to characterize the user profile of the user and the user's operational behavior sequence on the target account.

[0047] Optionally, the second feature refers to the feature determined by the financial institution through feature extraction of relevant structured user data of the target account during the process of pre-determining historical risk scores.

[0048] Optionally, the processing system performs a weighted summation of the risk scores corresponding to the first feature and the risk scores corresponding to the second feature to determine the target risk score.

[0049] Optionally, the processing system improves the accuracy of the final target risk score by re-detecting the target account based on the first and second features. By integrating structured and unstructured user data, the processing system obtains a dataset that more comprehensively and accurately reflects the true risk status of the account, avoiding the limitations of existing technologies that rely solely on expert rules to analyze structured user data, and reducing the risk of erroneous freezing of user financial accounts.

[0050] Step S205: If the target risk score is less than the preset risk threshold, the target account is unfrozen.

[0051] Optionally, the unfreezing process is used to update the account status of the target account to normal after the processing system confirms that the account risk of the target account is within a controllable range (i.e., less than the preset risk threshold). This step realizes the automatic unfreezing of the account, reduces the workload of financial institution staff, avoids the situation of freezing low-risk accounts for a long time, and improves the user experience.

[0052] As described above, this application intelligently adjusts the type and depth of user data collection based on the historical risk scores of frozen accounts. It collects unstructured user data (i.e., the first data) based on a data collection strategy. Subsequently, this application uses a large language model to extract features from the collected unstructured user data and combines this with user profiles and operational behavior sequences extracted from structured data to form a more comprehensive account feature set. Following this, the frozen accounts are re-risk-scored based on the features in the account feature set, thereby improving the accuracy of risk detection for frozen accounts. When the new risk score is lower than a preset risk threshold, the financial account is automatically unfrozen, thus solving the problem of erroneous freezing of financial accounts due to misjudgment of risk.

[0053] Therefore, this application adopts a data collection strategy based on the historical risk score of frozen accounts. By extracting features from the collected unstructured user data and the original structured data and re-scoring the risk, it achieves the goal of accurately identifying the user's financial risk level. This improves the accuracy of risk identification of user financial accounts and solves the technical problems of low accuracy in risk detection of financial accounts and false freezing of financial accounts in existing technologies, which leads to poor user experience.

[0054] In one optional embodiment, before determining the data collection strategy based on the target account's historical risk score and historical reasons for freezing, the processing system first scans the account status of financial accounts in a preset database based on a preset time period. Then, if the processing system detects that the financial account's account status is frozen, the processing system uses the financial account as the target account.

[0055] Optionally, the preset time period refers to the period during which the processing system periodically scans and detects the preset database; the preset database is used to store relevant information about all financial accounts in the financial institution, including but not limited to account status, transaction records, and associated user information.

[0056] Optionally, a frozen financial account status means that the financial account has been temporarily disabled due to excessive risk. In this case, the user is temporarily unable to perform operations such as withdrawals and transfers.

[0057] Optionally, the processing system can detect changes in the status of financial accounts in a timely manner through regular scanning. In the event of an account being mistakenly frozen, the system can take corresponding measures promptly after the account is frozen, thus preventing users' financial accounts from being mistakenly frozen for a long time, avoiding impact on users' normal trading activities, and improving user experience.

[0058] Optionally, the processing system identifies frozen financial accounts as target accounts, allowing subsequent data collection and risk assessment to focus on these risky accounts. This reduces the amount of data the processing system needs to collect and improves its efficiency. Furthermore, for non-risk accounts (i.e., financial accounts that are not frozen), the processing system can reduce unnecessary data collection and processing, avoiding frequent disruptions to users of normal accounts and further optimizing the user experience.

[0059] In one optional embodiment, during the process of the processing system determining the data collection strategy based on the historical risk score of the target account, if the historical risk score is less than or equal to a first preset score, the processing system uses the first preset strategy as the data collection strategy, wherein the first preset strategy is used to collect the user's facial image; if the historical risk score is greater than the first preset score and less than or equal to a second preset score, the processing system uses the second preset strategy as the data collection strategy, wherein the second preset strategy is used to collect the user's verification code and contact information verification results, and the second preset score is greater than the first preset score; if the historical risk score is greater than the second preset score, the processing system uses a third preset strategy as the data collection strategy, wherein the third preset strategy is used to collect the user's verification code, contact information verification results, facial image, and transaction verification data.

[0060] Optionally, the first preset score can be set to 0.5, and the second preset score can be set to 0.8.

[0061] Optionally, if the processing system detects that the historical risk score is less than or equal to the first preset score, the processing system determines that the target account is of low risk level. At this time, the processing system only prompts the user to input a facial image for face verification, without needing too much other user data, thus avoiding excessive scrutiny of low-risk accounts.

[0062] Optionally, if the processing system detects that the historical risk score is greater than the first preset score and less than or equal to the second preset score, the processing system determines that the target account is at a medium risk level. At this time, the processing system prompts the user to enter the user's verification code (e.g., SMS verification code) and contact information. Then, the processing system, together with the operator corresponding to the contact information, performs collaborative risk verification on the contact information entered by the user to obtain the contact information verification result.

[0063] Optionally, by collecting verification codes and contact information, the processing system increases the types of data for subsequent risk detection of target accounts by the large language model, thereby improving the processing system's risk control capabilities for medium-risk accounts.

[0064] Optionally, if the processing system detects that the historical risk score is greater than the second preset score, the processing system determines that the target account is of a high-risk level. At this time, the processing system prompts the user to enter the user's contact information verification result, facial image, and transaction verification data. The transaction verification data includes at least the relationship between the user and the counterparty and the purpose of the transaction funds.

[0065] Optionally, the processing system employs the most verification methods and collects the widest coverage of unstructured user data for high-risk accounts. Among these methods, the collection of transaction verification data ensures that a high accuracy rate in risk identification can be maintained even when the target account is in a complex trading environment.

[0066] Optionally, by adopting the above-mentioned hierarchical data collection strategy, the processing system can flexibly adjust the verification intensity according to the risk scores of different target accounts, which can effectively control risks while taking into account user experience.

[0067] Optionally, Figure 3 This is a flowchart of an optional data acquisition method according to an embodiment of this application, such as... Figure 3 As shown, the method for collecting the first data in this scheme includes the following steps:

[0068] (1) Input basic information: After logging into the corresponding mini-program or software application, the user enters basic user information such as user name and mobile phone number;

[0069] (2) Query the transaction to be verified: Based on the basic user information entered by the user, query whether the user's account status is frozen. If the user's financial account has been frozen, combine the various verification measures in steps (3) to (6) below according to the historical risk score.

[0070] For example, the following is a pseudocode for an optional collection of unstructured user data, where risk_score represents historical risk scores:

[0071] def select_verify_sources(risk_score):

[0072] if risk_score > 0.8:

[0073] return ["SMS verification code","Contact information","Facial image","Transaction verification data"]#Quadruple strong verification;

[0074] elif risk_score > 0.5:

[0075] return ["SMS verification code","Contact information"]#Two-factor authentication;

[0076] else:

[0077] return ["facial image"] # Basic validation;

[0078] (3) SMS verification: Send a verification code via SMS;

[0079] (4) Mobile phone number verification; call the mobile phone number operator to perform real-name verification;

[0080] (5) Perform facial recognition;

[0081] (6) Fill in transaction verification data: Based on background information such as the relationship between the user and the counterparty and the purpose of funds, check whether the user's corresponding transaction has risks;

[0082] (7) Persistence of verification results: The verification results are inserted into the preset database;

[0083] (8) The verification results will be used by the subsequent risk decision engine (i.e., the big model).

[0084] In one optional embodiment, after determining the data collection strategy based on the target account's historical risk score and historical freezing reasons, the processing system first filters the target account's account identifier in a preset database to obtain the user identifier of the user to which the target account belongs. Then, the processing system filters the user's contact information in the preset database based on the user identifier. Finally, the processing system generates a prompt message through the data collection strategy and sends the prompt message to the user through the contact information. The prompt message is used to remind the user to provide the user data corresponding to the data collection strategy again.

[0085] Optionally, the account identifier of the target account is a unique identifier for the target account.

[0086] Optionally, the user identifier is a unique identifier of the user holding the target account.

[0087] Optionally, the standardized user data in the default database is stored in the form of tables, and the processing system can quickly filter out the required information based on the table header fields.

[0088] Optionally, by associating account identifiers with user identifiers, the processing system can accurately locate the user who needs to provide the first data again, avoiding interference with other users and thus improving the overall user experience.

[0089] Optionally, quickly obtaining contact information through user identification and sending notification messages to users using that contact information can achieve the following:

[0090] (1) Sending notification messages through the contact information provided by the user avoids the need for the user to actively check the account status, which improves the user experience while complying with data protection and privacy regulations.

[0091] (2) The automated prompting mechanism speeds up the process of re-collecting user data, thereby shortening the entire risk assessment and de-control processing cycle and improving the overall operating efficiency of the processing system.

[0092] In one optional embodiment, during the feature extraction process of the first data using the target model, the processing system first performs data cleaning and data desensitization on the text data in the first data using the target model to obtain the first text data. Data cleaning is used to remove noise data from the text data, and data desensitization is used to desensitize sensitive data in the text data. Then, the processing system performs image cropping and grayscale processing on the image data in the first data to obtain the first image data. Next, the processing system extracts features from the first text data and the first image data to obtain first text features and first image features. The first text features are used to characterize the semantic information of keywords in the first text data, and the first image features are used to characterize the pixel information of key points in the first image data. Finally, the processing system performs feature fusion on the first text features and the first image features to obtain the first feature of the target account.

[0093] Optionally, by accurately identifying and removing outliers, irrelevant information, or erroneous data from text data before data cleaning for data analysis, the processing system improves data quality and, consequently, the accuracy and efficiency of subsequent analysis. Through data desensitization, the processing system can process sensitive information in the original text data, such as by replacement, masking, or encryption, thereby protecting user privacy and security.

[0094] Optionally, the processing system can crop the image data in the first data to extract regions containing key information, such as facial features, from the original image, thereby reducing the interference of irrelevant data on the image recognition model. Then, by converting the color image to a grayscale image, the processing system can simplify the color information of the image, which helps the model to focus its attention on the texture and shape features in the image, thereby improving the image processing speed and the accuracy of model recognition.

[0095] Optionally, the processing system extracts features from the first text data and the first image data, thereby transforming unstructured text data and image data into structured data features. This facilitates subsequent models in reprocessing data during risk detection, improving the efficiency and accuracy of risk decision-making in subsequent models, while reducing data storage and computation costs.

[0096] Optionally, the processing system performs feature fusion on the first text features and the first image features, that is, combines different types of unstructured data features to form a more comprehensive feature representation containing multi-dimensional information, so as to improve the accuracy and robustness of the decision model; by analyzing the first feature that integrates text and image, the model can assess account risk from multiple perspectives, thereby improving the accuracy of risk identification.

[0097] In summary, the steps of the above embodiments, by processing the collected first data, including cleaning, desensitizing, cropping, grayscale conversion, and feature extraction and fusion of text and image data, not only improve the quality and processing efficiency of the data, but also enhance the accuracy and reliability of the final target account risk decision.

[0098] In an optional embodiment, before determining the target risk score of the target account based on the first and second features of the target account, the processing system first generates a target statement based on the account identifier of the target account and the user identifier of the user. The target statement is a database operation statement. Then, after receiving the user's authorization information, the processing system filters from a preset database based on the target statement to obtain the user's historical operation records of the target account within a preset historical time period, as well as the user's user data and asset data. Then, the processing system performs feature extraction on the user data and asset data to obtain user profile features, and performs feature extraction on the historical operation records to obtain device fingerprint features and user temporal behavior features. Finally, the processing system uses the user profile features, device fingerprint features, and user temporal behavior features as the second feature.

[0099] Optionally, a preset historical time period refers to a time window set for analyzing user behavior and account risk, which can help financial institutions understand users' recent behavioral patterns.

[0100] Optionally, historical operation records refer to records of user operations such as transactions, logins, and password changes for the target account, which are used to analyze user behavior and potential risks.

[0101] Optionally, device fingerprint features refer to a unique combination of features extracted from a user's operating device, used to identify and track a user's behavior on different devices, including but not limited to the user's login device IP address, the browser type used to access service pages provided by the financial system, and the operating system of the device.

[0102] Optionally, user time-series behavioral characteristics refer to the characteristics of user behavior over time analyzed from historical operation records, such as transaction frequency, amount distribution, login time, etc., which help to reveal the patterns and anomalies of user transaction behavior.

[0103] Optionally, after receiving the user's authorization information, the processing system searches the target account's historical operation records, as well as the user's user data and asset data, using database operation statements to achieve the following:

[0104] (1) Precise data location: The target statement can accurately locate all records related to the target account and user from the huge database, thereby improving the efficiency and accuracy of data retrieval.

[0105] (2) Operation security: Database operation statements generated by account identifier and user identifier ensure the targeting of database operations and avoid the risk of operation errors or data leakage.

[0106] (3) Automated data retrieval: The authorization information activates the data filtering process based on the target statement, realizing automated and efficient data retrieval and reducing the workload of manual operation.

[0107] Optionally, the processing system extracts features from user data and asset data to obtain user profile features, and extracts features from historical operation records to obtain device fingerprint features and user temporal behavior features, thereby achieving the following functions:

[0108] (1) Deep user understanding: Feature extraction transforms user data into structured features, which deepens the understanding of user behavior and preferences and provides the possibility for personalized risk assessment.

[0109] (2) Nonlinear relationship mining: By analyzing time-series behavior and device fingerprints, the processing system can capture nonlinear relationships between data, thereby improving the accuracy of risk prediction and the adaptability to complex scenarios.

[0110] In one optional embodiment, in the process of determining the target risk score of the target account based on the first and second features of the target account, the processing system first determines the first risk score corresponding to the first feature and the second risk score corresponding to the second feature based on a preset scoring rule. The risk score is used to characterize the similarity between the features of the target account and the features of historical financial accounts with transaction risks. Then, the processing system performs a weighted summation of the first risk score and the second risk score to obtain the target risk score.

[0111] Optionally, determining the first risk score corresponding to the first feature and the second risk score corresponding to the second feature based on preset scoring rules has the following effects:

[0112] (1) Risk quantification: Transforming abstract features into specific risk scores makes risk assessment more objective and quantifiable, facilitating subsequent comparison and judgment.

[0113] (2) Multi-angle assessment: By scoring the first and second characteristics respectively, account risk can be assessed from different levels, thereby enhancing the comprehensiveness and accuracy of the assessment.

[0114] Optionally, the processing system obtains the target risk score by weighted summation of the first risk score and the second risk score, which has the following functions:

[0115] (1) Comprehensive risk assessment: The weighted summation takes into account the risk scores of two dimensions, basic data and advanced features, thus forming a more comprehensive target risk score that can reflect the full picture of account risk.

[0116] (2) Flexibility and customization: The selection of weights can be adjusted according to specific business scenarios and risk preferences, so that the processing system can adapt to the ever-changing market environment and business needs.

[0117] (3) Enhanced decision-making basis: As a highly generalized risk indicator, the target risk score provides strong data support for subsequent de-control decisions and helps to make more accurate decisions.

[0118] Optionally, the above embodiments introduce preset scoring rules to transform the first feature and the second feature into risk scores, and then integrate the first risk score and the second risk score into a target risk score through weighted summation. This not only achieves quantitative assessment of risk, but also comprehensively considers the potential risks of financial accounts from different perspectives, thereby improving the accuracy of subsequent risk decisions (e.g., unfreezing frozen accounts).

[0119] As described above, this application intelligently adjusts the type and depth of user data collection based on the historical risk scores of frozen accounts. It collects unstructured user data (i.e., the first data) based on a data collection strategy. Subsequently, this application uses a large language model to extract features from the collected unstructured user data and combines this with user profiles and operational behavior sequences extracted from structured data to form a more comprehensive account feature set. Following this, the frozen accounts are re-risk-scored based on the features in the account feature set, thereby improving the accuracy of risk detection for frozen accounts. When the new risk score is lower than a preset risk threshold, the financial account is automatically unfrozen, thus solving the problem of erroneous freezing of financial accounts due to misjudgment of risk.

[0120] Therefore, this application adopts a data collection strategy based on the historical risk score of frozen accounts. By extracting features from the collected unstructured user data and the original structured data and re-scoring the risk, it achieves the goal of accurately identifying the user's financial risk level. This improves the accuracy of risk identification of user financial accounts and solves the technical problems of low accuracy in risk detection of financial accounts and false freezing of financial accounts in existing technologies, which leads to poor user experience.

[0121] In summary, existing implementations of traditional risk deregulation decisions rely on human experience and rule engines, resulting in problems such as slow response, high misjudgment rate, and lack of dynamic adaptability. This application aims to propose a dynamic risk deregulation system for financial accounts based on multi-party collaborative verification and large-scale intelligent decision-making. By using multiple verification methods to convert verification results into structured data for use by the intelligent decision engine, this system achieves integrated, accurate, efficient, and automated account risk identification and handling. The core advantages of this solution are reflected in the following aspects:

[0122] (1) Interact directly with users, reach users directly through SMS linkage, message push and other means, and prompt users to log in to this processing system through mini-program or software application for collaborative verification.

[0123] (2) Fast processing capability of unstructured data: This processing system centrally stores the unstructured verification results such as face recognition, mobile phone number verification, and device fingerprint verification in a feature format for use by the large model intelligent rule engine, ensuring a response time of 10ms.

[0124] (3) End-to-end automation: The entire process from data input to risk decision-making is automated, reducing human intervention.

[0125] (4) Improve the timeliness of account release and reduce customer complaints: Ensure that the account can be released within 1 hour after customer verification to reduce customer complaints.

[0126] (5) Using multi-source verification dynamic routing improves the processing efficiency of the processing system.

[0127] This application embodiment can also provide a financial account processing device based on a large model. It should be noted that the financial account processing device based on a large model in this application embodiment can be used to execute the financial account processing method based on a large model provided in this application embodiment. The following is a description of the financial account processing device based on a large model provided in this application embodiment.

[0128] According to embodiments of this application, an apparatus for implementing the above-described large-model-based financial account processing method is also provided. Figure 4This is a schematic diagram of an optional large-model-based financial account processing device according to an embodiment of this application, such as... Figure 4 As shown, the device includes: a strategy determination unit 401, a data acquisition unit 402, a first extraction unit 403, a scoring determination unit 404, and an account unfreezing unit 405.

[0129] Optionally, the strategy determination unit 401 is used to determine a data collection strategy based on the historical risk score of the target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected; the data collection unit 402 is used to collect first data based on the data collection strategy after receiving the authorization information of the user of the target account, wherein the first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data, and the contact information verification result is the risk detection result of the operator on the user's contact information; the first extraction unit 403 is used to extract features from the first data through the target model to obtain the first feature of the target account, wherein the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, where X is a positive integer; the scoring determination unit 404 is used to determine the target risk score of the target account based on the first feature and the second feature of the target account, wherein the second feature is used to characterize the user profile of the user and the user's operation behavior sequence on the target account; the account unfreezing unit 405 is used to unfreeze the target account if the target risk score is less than a preset risk threshold.

[0130] In one alternative embodiment, the large-model-based financial account processing apparatus further includes a scanning unit and an account determination unit.

[0131] Optionally, the scanning unit is used to scan the account status of financial accounts in a preset database based on a preset time period; the account determination unit is used to select a financial account as a target account when the account status of the financial account is frozen.

[0132] In one optional embodiment, the strategy determination unit 401 includes: a first determination subunit, a second determination subunit, and a third determination subunit.

[0133] Optionally, the first determining subunit is used to use a first preset strategy as the data collection strategy when the historical risk score is less than or equal to a first preset score, wherein the first preset strategy is used to collect the user's facial image; the second determining subunit is used to use a second preset strategy as the data collection strategy when the historical risk score is greater than the first preset score and less than or equal to a second preset score, wherein the second preset strategy is used to collect the user's verification code and contact information verification results, and the second preset score is greater than the first preset score; the third determining subunit is used to use a third preset strategy as the data collection strategy when the historical risk score is greater than the second preset score, wherein the third preset strategy is used to collect the user's verification code, contact information verification results, facial image, and transaction verification data.

[0134] In one alternative embodiment, the financial account processing device based on a large model further includes: a first screening unit, a second screening unit, and an information generation unit.

[0135] Optionally, the first filtering unit is used to filter the target account in a preset database based on the account identifier of the target account to obtain the user identifier of the user to which the target account belongs; the second filtering unit is used to filter the user in the preset database based on the user identifier to obtain the user's contact information; the information generation unit is used to generate a prompt message through the data collection strategy and send the prompt message to the user through the contact information, wherein the prompt message is used to prompt the user to re-provide the user data corresponding to the data collection strategy.

[0136] In one optional embodiment, the first extraction unit 403 includes: a first data processing subunit, a second data processing subunit, a feature extraction subunit, and a feature fusion subunit.

[0137] Optionally, the first data processing subunit is used to perform data cleaning and data desensitization on the text data in the first data using the target model to obtain the first text data, wherein data cleaning is used to remove noise data from the text data, and data desensitization is used to desensitize sensitive data in the text data; the second data processing subunit is used to perform image cropping and grayscale processing on the image data in the first data to obtain the first image data; the feature extraction subunit is used to extract features from the first text data and the first image data to obtain the first text features and the first image features, wherein the first text features are used to characterize the semantic information of keywords in the first text data, and the first image features are used to characterize the pixel information of key points in the first image data; the feature fusion subunit is used to fuse the first text features and the first image features to obtain the first feature of the target account.

[0138] In one optional embodiment, the financial account processing device based on a large model further includes: a target statement generation unit, a third filtering unit, a second extraction unit, and a feature determination unit.

[0139] Optionally, the target statement generation unit is used to generate a target statement based on the account identifier of the target account and the user identifier of the user, wherein the target statement is a database operation statement; the third filtering unit is used to filter from a preset database based on the target statement after receiving the user's authorization information to obtain the user's historical operation records of the target account within a preset historical time period, as well as the user's user data and asset data; the second extraction unit is used to extract features from the user data and asset data to obtain user profile features, and to extract features from the historical operation records to obtain device fingerprint features and user temporal behavior features; the feature determination unit is used to use the user profile features, device fingerprint features, and user temporal behavior features as the second feature.

[0140] In an optional embodiment, the scoring determination unit 404 includes a scoring subunit and a weighted summation subunit.

[0141] Optionally, the scoring subunit is used to determine a first risk score corresponding to a first feature and a second risk score corresponding to a second feature based on a preset scoring rule, wherein the risk score is used to characterize the similarity between the features of the target account and the features of a historical financial account with transaction risk; the weighted summation subunit is used to perform a weighted summation of the first risk score and the second risk score to obtain the target risk score.

[0142] As described above, this device intelligently adjusts the type and depth of user data collection based on the historical risk scores of frozen accounts. It collects unstructured user data (i.e., the first data) based on a data collection strategy. Subsequently, the device uses a large language model to extract features from the collected unstructured user data and combines this with user profiles and operational behavior sequences extracted from structured data to form a more comprehensive account feature set. Following this, the frozen account is re-evaluated for risk based on the features in the account feature set, thereby improving the accuracy of risk detection for frozen accounts. When the new risk score is lower than a preset risk threshold, the financial account is automatically unfrozen, thus solving the problem of erroneous freezing of financial accounts due to misjudgment of risk.

[0143] Therefore, this device adopts a data collection strategy based on the historical risk score of frozen accounts. By extracting features from the collected unstructured user data and the original structured data, and then re-scoring the risk, it achieves the goal of accurately identifying the user's financial risk level. This improves the accuracy of risk identification of user financial accounts and solves the technical problems of low accuracy in risk detection of financial accounts and the occurrence of false freezing of financial accounts in existing technologies, which leads to a poor user experience.

[0144] It should be noted that the strategy determination unit 401, data acquisition unit 402, first extraction unit 403, scoring determination unit 404, and account unfreezing unit 405 mentioned above correspond to steps S201 to S205 in the method embodiment. The five units and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules can also be part of the device and run in the computer terminal 10 provided in the embodiment.

[0145] Embodiments of this application can also provide an electronic device. Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application, such as... Figure 5 As shown, the electronic device includes: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0146] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0147] The processor can access information and applications stored in memory via a transmission device to execute the following steps: Determine a data collection strategy based on the historical risk score of the target account, where the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected; After receiving authorization information from the user of the target account, collect first data based on the data collection strategy, where the first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data, and the contact information verification result is the operator's risk detection result for the user's contact information; Extract features from the first data using a target model to obtain the first feature of the target account, where the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, where X is a positive integer; Determine the target risk score of the target account based on the first and second features of the target account, where the second feature is used to characterize the user profile and the user's operational behavior sequence on the target account; If the target risk score is less than a preset risk threshold, unfreeze the target account.

[0148] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: scanning the account status of financial accounts in a preset database based on a preset time period; and, if the account status of a financial account is frozen, using the financial account as the target account.

[0149] The processor can access information and applications stored in the memory via a transmission device to execute the following steps: If the historical risk score is less than or equal to a first preset score, a first preset strategy is used as the data acquisition strategy, wherein the first preset strategy is used to acquire the user's facial image; if the historical risk score is greater than the first preset score and less than or equal to a second preset score, a second preset strategy is used as the data acquisition strategy, wherein the second preset strategy is used to acquire the user's verification code and contact information verification results, and the second preset score is greater than the first preset score; if the historical risk score is greater than the second preset score, a third preset strategy is used as the data acquisition strategy, wherein the third preset strategy is used to acquire the user's verification code, contact information verification results, facial image, and transaction verification data.

[0150] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: filtering the target account's account identifier in a preset database to obtain the user identifier of the user to which the target account belongs; filtering the user identifier in the preset database to obtain the user's contact information; generating a prompt message through a data acquisition strategy and sending the prompt message to the user through the contact information, wherein the prompt message is used to prompt the user to provide the user data corresponding to the data acquisition strategy again.

[0151] The processor can invoke information and applications stored in the memory via a transmission device to perform the following steps: Data cleaning and data desensitization are performed on the text data in the first data using the target model to obtain first text data, wherein data cleaning is used to remove noise data from the text data, and data desensitization is used to desensitize sensitive data in the text data; Image cropping and grayscale processing are performed on the image data in the first data to obtain first image data; Feature extraction is performed on the first text data and the first image data to obtain first text features and first image features, wherein the first text features are used to characterize the semantic information of keywords in the first text data, and the first image features are used to characterize the pixel information of key points in the first image data; Feature fusion is performed on the first text features and the first image features to obtain the first feature of the target account.

[0152] The processor can invoke information and applications stored in the memory via a transmission device to execute the following steps: Based on the account identifier of the target account and the user identifier of the user, generate a target statement, wherein the target statement is a database operation statement; after receiving the user's authorization information, filter from a preset database based on the target statement to obtain the user's historical operation records on the target account within a preset historical time period, as well as the user's user data and asset data; extract features from the user data and asset data to obtain user profile features, and extract features from the historical operation records to obtain device fingerprint features and user temporal behavior features; use the user profile features, device fingerprint features, and user temporal behavior features as the second feature.

[0153] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: determine the first risk score corresponding to the first feature and the second risk score corresponding to the second feature based on the preset scoring rules, wherein the risk score is used to characterize the similarity between the features of the target account and the features of historical financial accounts with transaction risks; and perform a weighted summation of the first risk score and the second risk score to obtain the target risk score.

[0154] This application provides a financial account processing solution based on a large-scale model. Based on the historical risk scores of frozen accounts, this application intelligently adjusts the type and depth of user data collection. Unstructured user data (i.e., the first data) is collected based on a data collection strategy. Subsequently, this application uses a large-scale language model to extract features from the collected unstructured user data and combines this with user profiles and operational behavior sequences extracted from structured data beforehand to form a more comprehensive account feature set. Subsequently, the frozen account is re-risk-scored based on the features in the account feature set, thereby improving the accuracy of risk detection for frozen accounts. When the new risk score is lower than a preset risk threshold, the financial account is automatically unfrozen, thus solving the problem of erroneous freezing of financial accounts due to misjudgment of risk.

[0155] Therefore, this application adopts a data collection strategy based on the historical risk score of frozen accounts. By extracting features from the collected unstructured user data and the original structured data and re-scoring the risk, it achieves the goal of accurately identifying the user's financial risk level. This improves the accuracy of risk identification of user financial accounts and solves the technical problems of low accuracy in risk detection of financial accounts and false freezing of financial accounts in existing technologies, which leads to poor user experience.

[0156] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, PDAs, mobile internet devices, PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0157] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0158] Embodiments of this application may also provide a storage medium.

[0159] Optionally, in this embodiment of the application, the storage medium can be used to store the program code executed by the financial account processing method based on a large model provided in the above method embodiments.

[0160] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0161] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing steps of a financial account processing method based on a large model.

[0162] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0163] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0164] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0165] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0166] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0167] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0168] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A financial account processing method based on a large model, characterized in that, include: The data collection strategy is determined based on the historical risk score of the target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected. After receiving the authorization information from the user of the target account, the first data is collected based on the data collection strategy. The first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data. The contact information verification result is the risk detection result of the operator on the user's contact information. The first feature of the target account is obtained by extracting features from the first data using the target model, wherein the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, and X is a positive integer; Based on the first and second features of the target account, a target risk score for the target account is determined, wherein the second feature is used to characterize the user profile of the user and the user's sequence of operational behaviors on the target account; If the target risk score is less than a preset risk threshold, the target account will be unfrozen.

2. The financial account processing method based on a large model according to claim 1, characterized in that, Before determining the data collection strategy based on the historical risk score of the target account, the large-model-based financial account processing method also includes: The account status of financial accounts in a preset database is scanned based on a preset time period. If the financial account is in a frozen state, the financial account will be used as the target account.

3. The financial account processing method based on a large model according to claim 1, characterized in that, The data collection strategy is determined based on the historical risk score of the target account, including: If the historical risk score is less than or equal to the first preset score, the first preset strategy will be used as the data collection strategy, wherein the first preset strategy is used to collect the user's facial image. If the historical risk score is greater than the first preset score and less than or equal to the second preset score, the second preset strategy will be used as the data collection strategy. The second preset strategy is used to collect the user's verification code and contact information verification results. The second preset score is greater than the first preset score. If the historical risk score is greater than the second preset score, the third preset strategy will be used as the data collection strategy. The third preset strategy is used to collect the user's verification code, contact information verification results, facial image, and transaction verification data.

4. The financial account processing method based on a large model according to claim 1, characterized in that, After determining the data collection strategy based on the historical risk score of the target account, the large-model-based financial account processing method further includes: Based on the account identifier of the target account, the user identifier of the user to which the target account belongs is obtained by filtering in a preset database; Based on the user identifier, the user's contact information is obtained by filtering the preset database. The data collection strategy generates a prompt message, which is then sent to the user via the contact method. The prompt message prompts the user to provide the user data corresponding to the data collection strategy again.

5. The financial account processing method based on a large model according to claim 1, characterized in that, The first feature of the target account is obtained by extracting features from the first data using the target model, including: The target model is used to perform data cleaning and data desensitization on the text data in the first data to obtain the first text data. The data cleaning is used to remove noise data from the text data, and the data desensitization is used to desensitize sensitive data in the text data. The image data in the first data is cropped and converted to grayscale to obtain the first image data. Feature extraction is performed on the first text data and the first image data to obtain first text features and first image features, wherein the first text features are used to characterize the semantic information of keywords in the first text data, and the first image features are used to characterize the pixel information of key points in the first image data. The first text feature and the first image feature are fused to obtain the first feature of the target account.

6. The financial account processing method based on a large model according to claim 1, characterized in that, Before determining the target risk score of the target account based on the first and second characteristics of the target account, the large-model-based financial account processing method further includes: Based on the account identifier of the target account and the user identifier of the user, a target statement is generated, wherein the target statement is a database operation statement; After receiving the user's authorization information, the system filters from a preset database based on the target statement to obtain the user's historical operation records of the target account within a preset historical time period, as well as the user's user data and asset data. Feature extraction is performed on the user data and the asset data to obtain user profile features, and feature extraction is performed on the historical operation records to obtain device fingerprint features and user temporal behavior features; The user profile features, the device fingerprint features, and the user temporal behavior features are used as the second feature.

7. The financial account processing method based on a large model according to claim 1, characterized in that, Based on the first and second characteristics of the target account, a target risk score for the target account is determined, including: A first risk score corresponding to the first feature and a second risk score corresponding to the second feature are determined based on a preset scoring rule. The risk score is used to characterize the similarity between the features of the target account and the features of a historical financial account with transaction risk. The target risk score is obtained by weighted summation of the first risk score and the second risk score.

8. A financial account processing device based on a large model, characterized in that, include: The strategy determination unit is used to determine a data collection strategy based on the historical risk score of the target account, wherein the target account is a frozen financial account, and the data collection strategy is used to determine the types of user data to be collected. The data acquisition unit is used to collect first data based on the data acquisition strategy after receiving the authorization information of the user of the target account. The first data is at least one of the user's verification code, contact information verification result, facial image, and transaction verification data. The contact information verification result is the risk detection result of the operator on the user's contact information. The first extraction unit is used to extract features from the first data through the target model to obtain the first feature of the target account, wherein the target model is a large language model trained based on unstructured user data corresponding to X financial accounts, and X is a positive integer; The scoring determination unit is used to determine the target risk score of the target account based on the first feature and the second feature of the target account, wherein the second feature is used to characterize the user profile of the user and the user's sequence of operational behaviors on the target account; The account unfreezing unit is used to unfreeze the target account when the target risk score is less than a preset risk threshold.

9. A computer program product, characterized in that, The computer program product includes a computer program, wherein, when the computer program is executed, it controls the computer program product to perform the financial account processing method based on any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the large-model-based financial account processing method according to any one of claims 1 to 7.

Images