Information Judgment Methods and Systems Based on Big Data Risk Control

By analyzing the interactive behavior of risk control users and the representational data of related users to generate user description vectors, and filtering related risk control users for risk control judgment, the problem of poor reliability of risk control judgment in existing technologies is solved, and more reliable risk control judgment results are achieved.

CN117349836BActive Publication Date: 2026-11-10HANGYIN CONSUMER FINANCE CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202311468360.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-07
Publication Date
2026-11-10
Estimated Expiration
2043-11-07

AI Technical Summary

Technical Problem

Existing technologies have poor reliability in risk control judgments, making it difficult to accurately assess the degree of user risk.

Method used

By analyzing the preset user interaction behaviors of the risk control users to be analyzed and the user representation data and behavioral content representation data of related risk control users, user description vectors are generated. Based on the vector matching parameters, related risk control users are filtered, and risk control dimension representation data of related risk control users are used for risk control judgment.

Benefits of technology

This improves the reliability of risk control judgments and ensures the accuracy and credibility of risk control judgment results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117349836B_ABST
    Figure CN117349836B_ABST
Patent Text Reader

Abstract

This invention provides an information discrimination method and system based on big data risk control, relating to the field of artificial intelligence technology. In this invention, preset user interaction behaviors and interacting risk control users are analyzed to identify the user to be analyzed and the interacting risk control users. Based on the user representation data of the user to be analyzed and the behavioral content representation data of the preset user interaction behaviors, a user description vector is analyzed. Based on the interacting user representation data corresponding to the interacting risk control users and the behavioral content representation data corresponding to the specified user interaction behaviors, an interacting user description vector is analyzed. Based on the vector matching parameters between the user description vector to be analyzed and the interacting user description vector, associated risk control users are determined. Based on the risk control dimension representation data corresponding to the associated risk control users and the risk control dimension representation data corresponding to the user to be analyzed, the target risk control discrimination result is output. Based on the above, the reliability of risk control discrimination can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to an information discrimination method and system based on big data risk control. Background Technology

[0002] Big data refers to data sets that cannot be captured, managed, and processed within a certain timeframe using conventional software tools. It represents massive, rapidly growing, and diverse information assets that require new processing models to achieve stronger decision-making, insightful discovery, and process optimization capabilities. With continuously improving data processing capabilities, effective information mining can be performed on big data. For example, in the field of risk control, user data (such as identity data and credit data) can be extracted from risk control users. This user data can then be analyzed and mined to determine whether there is risk or the degree of risk. However, current technologies suffer from issues with the reliability of risk control judgments. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide an information discrimination method and system based on big data risk control, so as to improve the reliability of risk control discrimination to a certain extent.

[0004] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0005] A big data-based risk control information discrimination method includes:

[0006] For the risk control user to be analyzed, analyze the preset type of user interaction behavior performed by the risk control user to be analyzed within a preset time interval, and analyze at least one interactive risk control user other than the risk control user to be analyzed that is related to the preset type of user interaction behavior.

[0007] Based on the user characterization data of the risk control user to be analyzed and the behavioral content characterization data corresponding to the preset type of user interaction behavior, the user description vector corresponding to the risk control user to be analyzed is analyzed.

[0008] Based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval, an interactive user description vector corresponding to each of the at least one interactive risk control users is analyzed. The interactive user representation data, the user representation data to be analyzed, and the behavioral content representation data each have data forms including text data, audio data, and / or image data.

[0009] Based on the vector matching parameters between the user description vector to be analyzed and the user description vector corresponding to each of the at least one interactive risk control user, a first number of interactive risk control users whose vector matching parameters meet the pre-configured user screening rules are determined, and the first number of interactive risk control users are marked as associated risk control users of the user to be analyzed, so as to obtain a first number of associated risk control users.

[0010] Based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, a risk control discrimination operation is performed on the risk control user to be analyzed to output the corresponding target risk control discrimination result. The target risk control discrimination result is used to reflect the risk level of the risk control user to be analyzed.

[0011] In some preferred embodiments, in the above-mentioned information discrimination method based on big data risk control, the step of analyzing the user to be analyzed for risk control, identifying the user's preset type of user interaction behavior within a preset time interval, and identifying at least one interactive risk control user other than the user to be analyzed related to the preset type of user interaction behavior, includes:

[0012] The system analyzes the various user interaction behaviors of the risk control user under analysis when they are active and passive interaction users within a preset time interval, and marks these user interaction behaviors as preset types of user interaction behaviors performed by the risk control user under analysis. The active interaction user refers to the risk control user under analysis who actively interacts with other risk control users, and the passive interaction user refers to the risk control user under analysis who passively interacts with other risk control users.

[0013] Each user other than the user to be analyzed is marked as an interactive risk control user, corresponding to each preset type of user interaction behavior performed by the user to be analyzed.

[0014] In some preferred embodiments, in the above-mentioned information discrimination method based on big data risk control, the user representation data to be analyzed includes user attribute representation data of various dimensions, and the behavioral content representation data corresponding to the preset type of user interaction behavior includes the interaction content corresponding to the interaction behavior performed by the risk control user to be analyzed.

[0015] The step of analyzing the user description vector corresponding to the user to be analyzed based on the user representation data of the user to be analyzed and the behavioral content representation data corresponding to the preset type of user interaction behavior includes:

[0016] Based on the user attribute representation data of each dimension included in the user representation data to be analyzed, the corresponding key user attribute data is analyzed.

[0017] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the key behavioral record data corresponding to the preset type of user interaction behavior is analyzed.

[0018] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, analyze the behavioral specific key data corresponding to the preset type of user interaction behavior.

[0019] Based on the key user attribute data, the key behavioral record data, and the key behavioral specific data corresponding to the risk control user to be analyzed, the user description vector corresponding to the risk control user to be analyzed is obtained.

[0020] In some preferred embodiments, in the above-described information discrimination method based on big data risk control, the step of analyzing the behavioral record key data corresponding to the preset type of user interaction behavior based on the behavioral content representation data corresponding to the preset type of user interaction behavior includes:

[0021] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the first statistical value of the number of behaviors corresponding to each user interaction behavior mainly performed by the risk control user to be analyzed is obtained, and the duration of each user interaction behavior is analyzed. Each user interaction behavior mainly performed by the risk control user to be analyzed means that the proportion of the user interaction data of the risk control user to be analyzed in the user interaction behavior is not less than the pre-configured reference proportion.

[0022] The second statistical value of the number of user interaction behaviors corresponding to each user interaction behavior whose duration belongs to a pre-configured first reference duration interval is determined, and the attentionable behavior representation parameters corresponding to the preset type of user interaction behavior are analyzed based on the first statistical value of the number of behaviors and the second statistical value of the number of behaviors.

[0023] Based on the first statistical value of the number of behaviors and the characteristic parameters of the attentionable behaviors, the key behavioral record data corresponding to the preset type of user interaction behavior is analyzed.

[0024] In some preferred embodiments, in the above-described information discrimination method based on big data risk control, the step of analyzing the behavior-specific key data corresponding to the preset type of user interaction behavior based on the behavior content representation data corresponding to the preset type of user interaction behavior includes:

[0025] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the main user interaction behaviors of the risk control user to be analyzed are identified, and the target behavioral content representation data in each user interaction behavior is extracted. The main user interaction behaviors of the risk control user to be analyzed refer to the user interaction data volume of the risk control user to be analyzed in the user interaction behavior, which accounts for a proportion of the user interaction data volume of the risk control user to be analyzed that is not less than a pre-configured reference proportion.

[0026] The target behavior content representation data is segmented to form corresponding important data segments and occurrence representation parameters for each important data segment. Furthermore, a cluster of interactive content data segments corresponding to each pre-configured data dimension is extracted. The data dimension refers to the data format of the corresponding behavior content representation data. The occurrence representation parameters are used to reflect the number of important data segments that are the same as the corresponding important data segments.

[0027] Analyze the target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance, and mark the important data fragments corresponding to the target interactive content data fragment clusters as behavioral specific key data corresponding to the preset type of user interaction behavior.

[0028] In some preferred embodiments, in the above-described information discrimination method based on big data risk control, the step of analyzing the cluster of target interactive content data segments whose semantic relevance to each of the important data segments is greater than or equal to a predetermined reference semantic relevance includes:

[0029] Select the target importance data segment with the maximum value of the occurrence representation parameter among the various importance data segments, and analyze the target interactive content data segments whose semantic relevance to the target importance data segments is greater than or equal to a predetermined reference semantic relevance. Also, mark the interactive content data segment clusters including the target interactive content data segments as target interactive content data segment clusters whose semantic relevance to the various importance data segments is greater than or equal to a predetermined reference semantic relevance; or

[0030] The clusters of pending interactive content data fragments to which each important data fragment belongs are analyzed. Based on the occurrence representation parameters corresponding to each important data fragment, the number of important data fragments included in each cluster of pending interactive content data fragments is analyzed. The clusters of pending interactive content data fragments with the maximum number of important data fragments are marked as target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance.

[0031] In some preferred embodiments, in the above-mentioned information discrimination method based on big data risk control, the interactive user representation data includes user attribute representation data of various dimensions, and the behavioral content representation data corresponding to the interactive behavior of the specified type of user includes the interactive content corresponding to the interactive behavior performed by the interactive risk control user.

[0032] The step of analyzing the interactive user description vector corresponding to each of the at least one interactive risk control users based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval includes:

[0033] For the at least one interactive risk control user, determine the user attribute representation data of each dimension included in the corresponding interactive user representation data, so as to output the key user attribute data corresponding to each interactive risk control user in the at least one interactive risk control user.

[0034] Based on the behavioral content representation data corresponding to the specified type of user interaction behavior, determine the behavioral record key data and behavioral specific key data corresponding to each of the at least one interactive risk control users;

[0035] Based on the user attribute key data, the behavior record key data, and the behavior specific key data corresponding to each of the at least one interactive risk control users, an interactive user description vector corresponding to each of the at least one interactive risk control users is determined.

[0036] In some preferred embodiments, in the above-described information discrimination method based on big data risk control, the step of performing risk control discrimination on the risk control user to be analyzed based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result, includes:

[0037] Extract the risk control dimension representation data corresponding to the risk control user to be analyzed to obtain the first risk control dimension representation data. The first risk control dimension representation data refers to the data used to make risk control judgments on the risk control user to be analyzed.

[0038] Extract the risk control dimension representation data corresponding to each of the first number of associated risk control users to obtain a first number of second risk control dimension representation data. The second risk control dimension representation data refers to the data used to make risk control judgments on the associated risk control users.

[0039] Feature mining is performed on the first risk control dimension representation data to form a first risk control data description vector corresponding to the risk control user to be analyzed; and feature mining is performed on each of the second risk control dimension representation parameters to form a first number of second risk control data description vectors corresponding to the risk control user to be analyzed.

[0040] Based on the first number of second risk control data description vectors, a deep association mining operation is performed on the first risk control data description vectors to form a target depth data description vector.

[0041] Based on the target depth data description vector, risk control discrimination operation is performed on the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result.

[0042] In some preferred embodiments, in the above-described information discrimination method based on big data risk control, the step of performing deep correlation mining on the first risk control data description vector based on the first number of second risk control data description vectors to form a target depth data description vector includes:

[0043] Based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector corresponding to each of the at least one interactive risk control user, the importance parameters corresponding to the first number of second risk control data description vectors are determined respectively.

[0044] For each of the first number of second risk control data description vectors, multiple cascaded deep association mining units are used to perform deep association mining operations on the first risk control data description vector based on the second risk control data description vector, so as to form an association deep data description vector corresponding to the second risk control data description vector.

[0045] Based on the importance parameters corresponding to each of the first number of second risk control data description vectors, a first aggregation operation is performed on the association depth data description vector corresponding to each second risk control data description vector to form a corresponding aggregated data description vector.

[0046] The aggregated data description vector and the first risk control data description vector are subjected to a second aggregation operation to form a corresponding target depth data description vector.

[0047] This invention also provides an information discrimination system based on big data risk control, including a processor and a memory. The memory is used to store a computer program, and the processor is used to execute the computer program to implement the above-mentioned information discrimination method based on big data risk control.

[0048] The information discrimination method and system based on big data risk control provided in this invention can first analyze the preset user interaction behaviors and interacting risk control users of the user to be analyzed; based on the user representation data of the user to be analyzed and the behavioral content representation data of the preset user interaction behaviors, analyze the user description vector; based on the interacting user representation data corresponding to the interacting risk control users and the behavioral content representation data corresponding to the specified user interaction behaviors, analyze the interacting user description vector; based on the vector matching parameters between the user description vector to be analyzed and the interacting user description vector, determine the associated risk control users; based on the risk control dimension representation data corresponding to the associated risk control users and the risk control dimension representation data corresponding to the user to be analyzed, output the target risk control discrimination result. Based on the foregoing, since a reliable screening operation for associated risk control users is performed first, the basis for the risk control discrimination operation on the user to be analyzed based on the risk control dimension representation data corresponding to the associated risk control users is more reliable. Therefore, it can improve the reliability of risk control discrimination to a certain extent, thereby improving the problem of poor reliability in the prior art.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0050] Figure 1 The structural block diagram of the information discrimination system based on big data risk control provided in the embodiments of the present invention is shown.

[0051] Figure 2 The flowchart illustrates the steps of the information discrimination method based on big data risk control provided in this embodiment of the invention.

[0052] Figure 3 This is a schematic diagram of the modules included in the information discrimination device based on big data risk control provided in an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0055] like Figure 1 As shown in the figure, this embodiment of the invention provides an information discrimination system based on big data risk control. The information discrimination system based on big data risk control may include a memory and a processor.

[0056] In detail, the memory and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, they can be electrically connected via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that exists in the form of software or firmware. The processor can be used to execute the executable computer program stored in the memory, thereby implementing the information discrimination method based on big data risk control provided in this embodiment of the invention.

[0057] It is understood that, in one possible implementation, the memory may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0058] It is understood that, in one possible implementation, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0059] It is understood that, in one possible implementation, the information discrimination system based on big data risk control can be a server with data processing capabilities.

[0060] Combination Figure 2 This invention also provides an information discrimination method based on big data risk control, which can be applied to the aforementioned information discrimination system based on big data risk control. The method steps defined in the relevant process of the information discrimination method based on big data risk control can be implemented by the information discrimination system based on big data risk control. The following will describe... Figure 2 The specific process shown will be explained in detail.

[0061] Step S110: For the risk control user to be analyzed, analyze the preset type of user interaction behavior performed by the risk control user to be analyzed within a preset time interval, and analyze at least one interactive risk control user other than the risk control user to be analyzed that is related to the preset type of user interaction behavior.

[0062] In this embodiment of the invention, the information discrimination system based on big data risk control can analyze the user to be analyzed for risk control, identify the preset type of user interaction behavior performed by the user to be analyzed within a preset time interval, and identify at least one other user interacting with the user to be analyzed who is related to the preset type of user interaction behavior. The preset time interval can refer to the most recent time interval, such as the most recent month or quarter.

[0063] Step S120: Based on the user characterization data of the risk control user to be analyzed and the behavioral content characterization data corresponding to the preset type of user interaction behavior, analyze the user description vector corresponding to the risk control user to be analyzed.

[0064] In this embodiment of the invention, the information discrimination system based on big data risk control can analyze the user description vector corresponding to the user to be analyzed based on the user characterization data of the user to be analyzed and the behavioral content characterization data corresponding to the preset type of user interaction behavior, and represent the characterization data in the form of a vector.

[0065] Step S130: Based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval, analyze the interactive user description vector corresponding to each of the at least one interactive risk control users.

[0066] In this embodiment of the invention, the information discrimination system based on big data risk control can analyze the interactive user description vector corresponding to each of the at least one interactive risk control users based on the interactive user representation data corresponding to each interactive risk control user and the behavioral content representation data corresponding to the specified types of user interaction behaviors performed by the at least one interactive risk control user within the preset time interval. The interactive user representation data, the user representation data to be analyzed, and the behavioral content representation data each have data formats including text data, audio data, and / or image data, i.e., at least one of the three data formats. For example, the interaction content of an interaction behavior may include only text data, or it may simultaneously include text data, audio data, and / or image data.

[0067] Step S140: Based on the vector matching parameters between the user description vector to be analyzed and the user description vector corresponding to each of the at least one interactive risk control users, determine a first number of interactive risk control users whose vector matching parameters conform to the pre-configured user screening rules, and mark the first number of interactive risk control users as associated risk control users of the user to be analyzed, so as to obtain a first number of associated risk control users.

[0068] In this embodiment of the invention, the information discrimination system based on big data risk control can determine a first number of interactive risk control users whose vector matching parameters conform to pre-configured user screening rules based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector corresponding to each of the at least one interactive risk control user. These first number of interactive risk control users are then marked as associated risk control users of the user to be analyzed, thus obtaining a first number of associated risk control users. The vector matching parameters corresponding to the selected associated risk control users should be greater than or equal to the vector matching parameters corresponding to the interactive risk control users who are not considered associated risk control users. For example, the system can select the first number of interactive risk control users with the largest vector matching parameters, or select each interactive risk control user whose vector matching parameters are greater than or equal to a pre-determined reference vector matching parameter. The vector matching parameters can be positively correlated with the cosine similarity or other similarities between the corresponding vectors.

[0069] Step S150: Based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, perform a risk control discrimination operation on the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result.

[0070] In this embodiment of the invention, the information discrimination system based on big data risk control can perform risk control discrimination operations on the user to be analyzed based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the user to be analyzed, so as to output the corresponding target risk control discrimination result. The target risk control discrimination result is used to reflect the risk level of the user to be analyzed, or in other words, the probability or likelihood of having risk, which can be specifically implemented through a corresponding neural network.

[0071] Based on the foregoing, since a reliable screening operation for associated risk control users is performed first, the basis for risk control judgment operations on the risk control users to be analyzed is more reliable when using the risk control dimension representation data corresponding to the associated risk control users. Therefore, the reliability of risk control judgment can be improved to a certain extent, thereby improving the problem of poor reliability in the existing technology.

[0072] It is understood that, in one possible implementation, step S110 described above, namely, analyzing the preset type of user interaction behavior performed by the user to be analyzed within a preset time interval, and analyzing at least one interactive risk control user other than the user to be analyzed related to the preset type of user interaction behavior, may further include the following specific implementation details:

[0073] The system analyzes the various user interaction behaviors of the risk control user under analysis during a preset time interval, when the user is an active or passive interaction user. It then marks these user interaction behaviors as preset types of user interaction behaviors performed by the risk control user under analysis. An active interaction user refers to the risk control user under analysis actively interacting with other risk control users; that is, in this user interaction behavior, the sending of the first message is performed by the risk control user under analysis. A passive interaction user refers to the risk control user under analysis passively interacting with other risk control users; that is, in this user interaction behavior, the sending of the first message is performed by a risk control user other than the risk control user under analysis.

[0074] Each user other than the user to be analyzed is marked as an interactive risk control user, corresponding to each preset type of user interaction behavior performed by the user to be analyzed. In other words, all other risk control users are regarded as the corresponding interactive risk control users.

[0075] It is understood that, in one possible implementation, the user representation data to be analyzed includes user attribute representation data of each dimension (i.e., each data format), and the behavioral content representation data corresponding to the preset type of user interaction behavior includes the interaction content corresponding to the interaction behavior performed by the risk control user to be analyzed. Based on this, step S120 in the above description, namely, the step of analyzing the user description vector corresponding to the risk control user to be analyzed based on the user representation data to be analyzed and the behavioral content representation data corresponding to the preset type of user interaction behavior, can further include the following specific implementation content:

[0076] Based on the user attribute representation data of each dimension included in the user representation data to be analyzed, the corresponding key user attribute data is analyzed. In other words, the user attribute data of each dimension can be filtered to remove some invalid data, thereby obtaining the key user attribute data. The user attribute data of each dimension may include user identity data recorded in text form, user facial information recorded in image form, etc.

[0077] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the key behavioral record data corresponding to the preset type of user interaction behavior is analyzed.

[0078] Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, analyze the behavioral specific key data corresponding to the preset type of user interaction behavior.

[0079] Based on the key user attribute data, the key behavioral record data, and the key behavioral specific data corresponding to the user to be analyzed, the user description vector corresponding to the user to be analyzed is obtained. For example, the key user attribute data, the key behavioral record data, and the key behavioral specific data can be mapped to the feature space to obtain the corresponding mapping vectors. Then, the mapping vectors can be aggregated, such as by performing a cascade combination operation, to form the corresponding user description vector.

[0080] It is understood that, in one possible implementation, the step of analyzing the behavioral record key data corresponding to the preset type of user interaction behavior based on the behavioral content representation data corresponding to the preset type of user interaction behavior may further include the following specific implementation details:

[0081] Based on the behavioral content representation data corresponding to the preset user interaction behaviors, the first statistical value of the number of behaviors corresponding to each user interaction behavior mainly performed by the risk control user to be analyzed is obtained, and the duration of each user interaction behavior is analyzed, that is, the duration from the start to the end of the interaction behavior. Each user interaction behavior mainly performed by the risk control user to be analyzed means that the proportion of the user interaction data of the risk control user to be analyzed (the proportion in the total data of the interaction between the two parties) is not less than a pre-configured reference proportion, which can be configured according to actual needs.

[0082] The system determines the second statistical value of the number of user interaction behaviors corresponding to each user interaction behavior whose duration falls within a pre-configured first reference duration interval. Based on the first and second statistical values ​​of the number of behaviors, it analyzes the attentionable behavior representation parameters corresponding to the preset type of user interaction behavior. The first reference duration interval can be selected according to actual needs. For example, the lower limit of the first reference duration interval can be a target value, which can be configured according to the duration of historical interaction behaviors. The upper limit of the first reference duration interval can be infinity, etc. For example, the attentionable behavior representation parameter can refer to the ratio of the first statistical value of the number of behaviors to the second statistical value of the number of behaviors.

[0083] Based on the first statistical value of the number of behaviors and the characteristic parameters of the attentionable behaviors, the key behavioral data corresponding to the preset type of user interaction behavior is analyzed. For example, the first statistical value of the number of behaviors and the characteristic parameters of the attentionable behaviors are used as the key behavioral data, that is, the key behavioral data includes the first statistical value of the number of behaviors and the characteristic parameters of the attentionable behaviors.

[0084] It is understood that, in one possible implementation, the step of analyzing the behavior-specific key data corresponding to the preset type of user interaction behavior based on the behavior content representation data corresponding to the preset type of user interaction behavior may further include the following specific implementation details:

[0085] Based on the behavioral content representation data corresponding to the preset user interaction behaviors, the main user interaction behaviors of the risk control user to be analyzed are identified, and the target behavioral content representation data in each user interaction behavior is extracted. The main user interaction behaviors of the risk control user to be analyzed refer to the user interaction data volume of the risk control user to be analyzed in the user interaction behavior, which accounts for a proportion of the user interaction data volume of the risk control user to be analyzed that is not less than a pre-configured reference proportion, as described above.

[0086] The target behavior content representation data is segmented into segments, for example, text data is segmented into units (the unit granularity can be sentences or words), or image or audio data is segmented into frames, to form corresponding important data segments and occurrence representation parameters for each important data segment. For example, data segments with occurrence representation parameters greater than or equal to a certain value can be considered important data segments, or some invalid information can be filtered out to obtain important data segments, such as retaining only some data segments from repetitive data segments as important data segments. Furthermore, a cluster of interactive content data segments corresponding to each pre-configured data dimension is extracted (which can be some data segments related to risk control data, specifically some data segments from the financial or economic fields, etc.). The data dimension refers to the data format of the corresponding behavior content representation data, and the occurrence representation parameter reflects the number of important data segments that are the same as the corresponding important data segments.

[0087] The system analyzes and identifies target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance. It then marks the important data fragments corresponding to these target interactive content data fragment clusters as behavioral-specific key data corresponding to the preset type of user interaction behavior. The specific value of the reference semantic relevance is not limited and can be configured according to actual needs. Furthermore, considering different data formats, the semantic relevance calculation can involve mapping the data fragments to feature spaces and then calculating the similarity of the resulting mapping vectors to obtain the semantic relevance.

[0088] It is understood that, in one possible implementation, the step of analyzing the cluster of target interactive content data fragments whose semantic relevance to each of the important data fragments is greater than or equal to a predetermined reference semantic relevance further includes the following specific implementation details:

[0089] Select the target important data segment with the maximum value of the occurrence characterization parameter among the various important data segments, and analyze the target interactive content data segments whose semantic relevance to the target important data segments is greater than or equal to a predetermined reference semantic relevance. Then, mark the interactive content data segment cluster including the target interactive content data segments as a target interactive content data segment cluster whose semantic relevance to the various important data segments is greater than or equal to a predetermined reference semantic relevance.

[0090] It is understood that, in one possible implementation, the step of analyzing the cluster of target interactive content data fragments whose semantic relevance to each of the important data fragments is greater than or equal to a predetermined reference semantic relevance further includes the following specific implementation details:

[0091] The clusters of pending interactive content data fragments to which each important data fragment belongs are analyzed. Based on the occurrence representation parameters corresponding to each important data fragment, the number of important data fragments included in each cluster of pending interactive content data fragments is analyzed. The clusters of pending interactive content data fragments with the maximum number of important data fragments are marked as target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance.

[0092] It is understood that, in one possible implementation, the interactive user representation data includes user attribute representation data of various dimensions, and the behavioral content representation data corresponding to the specified type of user interaction behavior includes the interaction content corresponding to the interactive behavior (including active and passive interactive behavior, etc.) performed by the interactive risk control user. Based on this, step S130 in the above description, namely, the step of analyzing the interactive user description vector corresponding to each of the at least one interactive risk control user based on the interactive user representation data corresponding to each of the at least one interactive risk control user and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval, can further include the following specific implementation content:

[0093] For the at least one interactive risk control user, determine the user attribute representation data of each dimension included in the corresponding interactive user representation data, so as to output the key user attribute data corresponding to each interactive risk control user among the at least one interactive risk control users, as described above.

[0094] Based on the behavioral content representation data corresponding to the specified type of user interaction behavior, determine the behavioral record key data and behavioral specific key data corresponding to each of the at least one interactive risk control users, as described above.

[0095] Based on the user attribute key data, the behavior record key data, and the behavior specific key data corresponding to each of the at least one interactive risk control users, an interactive user description vector corresponding to each of the at least one interactive risk control users is determined, as described above.

[0096] It is understood that, in one possible implementation, step S150 described above, namely, the step of performing a risk control discrimination operation on the risk control user to be analyzed based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result, may further include the following specific implementation content:

[0097] Extract the risk control dimension representation data (such as identity data, credit data, etc., i.e., the application data that needs to be submitted by the risk control user to be analyzed in the credit business. Since risk control involves issues such as creditworthiness, the risk control dimension representation data may also include corresponding behavioral content representation data) to obtain the first risk control dimension representation data. The first risk control dimension representation data refers to the data used to make risk control judgments on the risk control user to be analyzed.

[0098] Extract the risk control dimension representation data corresponding to each of the first number of associated risk control users to obtain a first number of second risk control dimension representation data. The second risk control dimension representation data refers to the data used to make risk control judgments on the associated risk control users.

[0099] The first risk control dimension representation data is subjected to feature mining operations, such as feature space mapping operations, to form a first risk control data description vector corresponding to the risk control user to be analyzed. In addition, feature mining operations are performed on each second risk control dimension representation parameter to form a first number of second risk control data description vectors corresponding to the risk control user to be analyzed.

[0100] Based on the first number of second risk control data description vectors, a deep association mining operation is performed on the first risk control data description vectors to form a target depth data description vector.

[0101] Based on the target depth data description vector, risk control discrimination is performed on the user to be analyzed to output the corresponding target risk control discrimination result. For example, the target depth data description vector can be fully processed by the fully connected unit included in the neural network, and then the activation function included in the neural network, such as the softmax function, can be used to activate the result of the fully connected processing to output the corresponding target risk control discrimination result.

[0102] It is understood that, in one possible implementation, the step of performing deep correlation mining on the first risk control data description vector based on the first number of second risk control data description vectors to form a target depth data description vector may further include the following specific implementation details:

[0103] Based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector corresponding to each of the at least one interactive risk control user, the importance parameters corresponding to the first number of second risk control data description vectors are determined respectively. For example, the vector matching parameters can be used as the corresponding importance parameters.

[0104] For each of the first number of second risk control data description vectors, multiple cascaded deep association mining units (which may be part of a corresponding neural network) are used to perform deep association mining on the first risk control data description vector based on the second risk control data description vector to form an associated deep data description vector corresponding to the second risk control data description vector. In the multiple cascaded deep association mining units, the input data of the first deep association mining unit is the second risk control data description vector and the first risk control data description vector, and the input data of each other deep association mining unit is the output data of the previous deep association mining unit and the first risk control data description vector.

[0105] Based on the importance parameters corresponding to each of the first number of second risk control data description vectors, a first aggregation operation is performed on the associated depth data description vector corresponding to each second risk control data description vector to form a corresponding aggregated data description vector. For example, the importance parameters can be used as weighting coefficients to perform weighted superposition operations on the associated depth data description vectors to output the corresponding aggregated data description vector.

[0106] The aggregated data description vector and the first risk control data description vector are subjected to a second aggregation operation to form a corresponding target depth data description vector. For example, the aggregated data description vector and the first risk control data description vector can be cascaded together.

[0107] It is understood that, in one possible implementation, the step of performing deep association mining on the first risk control data description vector based on the second risk control data description vector using multiple cascaded deep association mining units for each of the first number of second risk control data description vectors to form an associated deep data description vector corresponding to the second risk control data description vector may further include the following specific implementation details:

[0108] For the first deep correlation mining unit among the multiple cascaded deep correlation mining units, the deep correlation mining unit is used to perform focused feature analysis on the first risk control data description vector based on the second risk control data description vector to form the focused analysis data description vector corresponding to the deep correlation mining unit.

[0109] For each of the other deep association mining units in the cascaded deep association mining units other than the first deep association mining unit, the first risk control data description vector is subjected to focused feature analysis operation based on the focused analysis data description vector corresponding to the previous deep association mining unit. This can be done by referring to relevant existing technologies to form the focused analysis data description vector corresponding to the deep association mining unit.

[0110] Based on the focused analysis data description vector corresponding to the last deep correlation mining unit, the correlation deep data description vector corresponding to the second risk control data description vector is determined. For example, the focused analysis data description vector corresponding to the last deep correlation mining unit can be directly used as the correlation deep data description vector corresponding to the second risk control data description vector.

[0111] It is understood that, in one possible implementation, the step of performing deep association mining on the first risk control data description vector based on the second risk control data description vector using multiple cascaded deep association mining units for each of the first number of second risk control data description vectors to form an associated deep data description vector corresponding to the second risk control data description vector may further include the following specific implementation details:

[0112] For the first deep association mining unit in the cascaded deep association mining units, the deep association mining unit is used to perform deep association mining operation on the first risk control data description vector based on the second risk control data description vector to form the deep mining data description vector corresponding to the deep association mining unit.

[0113] For each of the other deep association mining units in the cascaded deep association mining units other than the first deep association mining unit, the first risk control data description vector is subjected to deep association mining operation based on the deep mining data description vector corresponding to the previous deep association mining unit, so as to form the deep mining data description vector corresponding to the deep association mining unit.

[0114] Based on the deep mining data description vector corresponding to the last deep correlation mining unit, the correlation deep data description vector corresponding to the second risk control data description vector is determined. For example, the deep mining data description vector corresponding to the last deep correlation mining unit can be directly used as the correlation deep data description vector corresponding to the second risk control data description vector.

[0115] It is understood that, in one possible implementation, the step of performing a deep correlation mining operation on the first deep correlation mining unit among the cascaded deep correlation mining units, based on the second risk control data description vector, to form the deep mining data description vector corresponding to the deep correlation mining unit, may further include the following specific implementation details:

[0116] Using the deep association mining unit, the second risk control data description vector and the first risk control data description vector are concatenated and combined to form a corresponding concatenated vector to be processed;

[0117] Using the first weight parameter carried by the deep association mining unit, a weighting operation is performed on the concatenated vector to be processed, and based on the first shift parameter carried by the deep association mining unit, the result of the weighting operation is added together, and the result of the addition operation is subjected to a first activation mapping operation to output the corresponding first activation mapping result. Each weight parameter and shift parameter can be formed during the network optimization process of the deep association mining unit, and the activation mapping function corresponding to each activation mapping operation can be the same or different.

[0118] Using the second weight parameter carried by the deep association mining unit, a weighting operation is performed on the concatenated vector to be processed, and based on the second shift parameter carried by the deep association mining unit, the result of the weighting operation is added together, and the result of the addition operation is subjected to a second excitation mapping operation to output the corresponding second excitation mapping result.

[0119] Using the third weight parameter carried by the deep association mining unit, a weighting operation is performed on the concatenated vector to be processed, and based on the third shift parameter carried by the deep association mining unit, the result of the weighting operation is added together, and the result of the addition operation is subjected to a third excitation mapping operation to output the corresponding third excitation mapping result.

[0120] Using the fourth weight parameter carried by the deep association mining unit, a weighting operation is performed on the concatenated vector to be processed, and based on the fourth shift parameter carried by the deep association mining unit, the result of the weighting operation is added together, and the result of the addition operation is subjected to a fourth excitation mapping operation to output the corresponding fourth excitation mapping result.

[0121] The second and third incentive mapping results are multiplied together, and the result of the multiplication is added to the first incentive mapping result. The result of the addition is then subjected to a fifth incentive mapping operation to output the corresponding fifth incentive mapping result. It is understood that the incentive mapping functions corresponding to the first, second, and fourth incentive mapping operations can be the same, and the incentive mapping functions corresponding to the third and fifth incentive mapping operations can be the same.

[0122] The fourth and fifth incentive mapping results are multiplied together to output the deep mining data description vector corresponding to the deep association mining unit.

[0123] Combination Figure 3 This invention also provides an information discrimination device based on big data risk control, which can be applied to the aforementioned information discrimination system based on big data risk control. The information discrimination device based on big data risk control may include the following software functional modules:

[0124] The risk control behavior user analysis module is used to analyze the risk control user to be analyzed, identify the preset type of user interaction behavior performed by the risk control user to be analyzed within a preset time interval, and identify at least one other risk control user who is related to the preset type of user interaction behavior.

[0125] The first description vector analysis module is used to analyze the description vector of the user to be analyzed based on the user characterization data of the user to be analyzed and the behavioral content characterization data corresponding to the preset type of user interaction behavior.

[0126] The second description vector analysis module is used to analyze the interactive user description vector corresponding to each of the at least one interactive risk control users based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control users in the preset time interval. The interactive user representation data, the user representation data to be analyzed and the behavioral content representation data each have data forms including text data, audio data and / or image data.

[0127] The associated risk control user determination module is used to determine a first number of interactive risk control users whose vector matching parameters conform to a pre-configured user screening rule based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector corresponding to each of the at least one interactive risk control users, and to mark the first number of interactive risk control users as associated risk control users of the user to be analyzed, so as to obtain the first number of associated risk control users.

[0128] The risk control discrimination module is used to perform risk control discrimination operations on the risk control user to be analyzed based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result. The target risk control discrimination result is used to reflect the risk level of the risk control user to be analyzed.

[0129] In summary, the information discrimination method and system based on big data risk control provided by this invention can first analyze the preset user interaction behaviors and interactive risk control users of the user to be analyzed; based on the user representation data of the user to be analyzed and the behavioral content representation data of the preset user interaction behaviors, analyze the user description vector; based on the interactive user representation data corresponding to the interactive risk control user and the behavioral content representation data corresponding to the specified user interaction behaviors, analyze the interactive user description vector; based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector, determine the associated risk control users; based on the risk control dimension representation data corresponding to the associated risk control users and the risk control dimension representation data corresponding to the user to be analyzed, output the target risk control discrimination result. Based on the foregoing, because a reliable screening operation for associated risk control users is performed first, the basis for the risk control discrimination operation on the user to be analyzed based on the risk control dimension representation data corresponding to the associated risk control users is more reliable. Therefore, it can improve the reliability of risk control discrimination to a certain extent, thereby improving the problem of poor reliability in the prior art.

[0130] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An information discrimination method based on big data risk control, characterized in that, include: For the risk control user to be analyzed, analyze the preset type of user interaction behavior performed by the risk control user to be analyzed within a preset time interval, and analyze at least one interactive risk control user other than the risk control user to be analyzed that is related to the preset type of user interaction behavior. Based on the user characterization data of the risk control user to be analyzed and the behavioral content characterization data corresponding to the preset type of user interaction behavior, the user description vector corresponding to the risk control user to be analyzed is analyzed. Based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval, an interactive user description vector corresponding to each of the at least one interactive risk control users is analyzed. The interactive user representation data, the user representation data to be analyzed, and the behavioral content representation data each have data forms including text data, audio data, and / or image data. Based on the vector matching parameters between the user description vector to be analyzed and the user description vector corresponding to each of the at least one interactive risk control user, a first number of interactive risk control users whose vector matching parameters meet the pre-configured user screening rules are determined, and the first number of interactive risk control users are marked as associated risk control users of the user to be analyzed, so as to obtain a first number of associated risk control users. Based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, a risk control discrimination operation is performed on the risk control user to be analyzed to output the corresponding target risk control discrimination result. The target risk control discrimination result is used to reflect the risk level of the risk control user to be analyzed. The step of performing risk control discrimination on the risk control user to be analyzed based on the risk control dimension representation data corresponding to each of the first number of associated risk control users and the risk control dimension representation data corresponding to the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result, includes: Extract the risk control dimension representation data corresponding to the risk control user to be analyzed to obtain the first risk control dimension representation data. The first risk control dimension representation data refers to the data used to make risk control judgments on the risk control user to be analyzed. Extract the risk control dimension representation data corresponding to each of the first number of associated risk control users to obtain a first number of second risk control dimension representation data. The second risk control dimension representation data refers to the data used to make risk control judgments on the associated risk control users. Feature mining is performed on the first risk control dimension representation data to form a first risk control data description vector corresponding to the risk control user to be analyzed; and feature mining is performed on each of the second risk control dimension representation parameters to form a first number of second risk control data description vectors corresponding to the risk control user to be analyzed. Based on the first number of second risk control data description vectors, a deep association mining operation is performed on the first risk control data description vectors to form a target depth data description vector. Based on the target depth data description vector, risk control discrimination operation is performed on the risk control user to be analyzed, so as to output the corresponding target risk control discrimination result; The step of performing deep association mining on the first risk control data description vector based on the first number of second risk control data description vectors to form a target depth data description vector includes: Based on the vector matching parameters between the user description vector to be analyzed and the interactive user description vector corresponding to each of the at least one interactive risk control user, the importance parameters corresponding to the first number of second risk control data description vectors are determined respectively. For each of the first number of second risk control data description vectors, multiple cascaded deep association mining units are used to perform deep association mining operations on the first risk control data description vector based on the second risk control data description vector, so as to form an association deep data description vector corresponding to the second risk control data description vector. Based on the importance parameters corresponding to each of the first number of second risk control data description vectors, a first aggregation operation is performed on the association depth data description vector corresponding to each second risk control data description vector to form a corresponding aggregated data description vector. The aggregated data description vector and the first risk control data description vector are subjected to a second aggregation operation to form a corresponding target depth data description vector.

2. The information discrimination method based on big data risk control as described in claim 1, characterized in that, The steps of analyzing the user to be analyzed for risk control, identifying the user's preset type of user interaction behavior within a preset time interval, and identifying at least one other user interacting with the user to be analyzed that is related to the preset type of user interaction behavior, include: The system analyzes the various user interaction behaviors of the risk control user under analysis when they are active and passive interaction users within a preset time interval, and marks these user interaction behaviors as preset types of user interaction behaviors performed by the risk control user under analysis. The active interaction user refers to the risk control user under analysis who actively interacts with other risk control users, and the passive interaction user refers to the risk control user under analysis who passively interacts with other risk control users. Each user other than the user to be analyzed is marked as an interactive risk control user, corresponding to each preset type of user interaction behavior performed by the user to be analyzed.

3. The information discrimination method based on big data risk control as described in claim 1, characterized in that, The user representation data to be analyzed includes user attribute representation data of various dimensions, and the behavioral content representation data corresponding to the preset type of user interaction behavior includes the interaction content corresponding to the interaction behavior performed by the risk control user to be analyzed. The step of analyzing the user description vector corresponding to the user to be analyzed based on the user representation data of the user to be analyzed and the behavioral content representation data corresponding to the preset type of user interaction behavior includes: Based on the user attribute representation data of each dimension included in the user representation data to be analyzed, the corresponding key user attribute data is analyzed. Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the key behavioral record data corresponding to the preset type of user interaction behavior is analyzed. Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, analyze the behavioral specific key data corresponding to the preset type of user interaction behavior. Based on the key user attribute data, the key behavioral record data, and the key behavioral specific data corresponding to the risk control user to be analyzed, the user description vector corresponding to the risk control user to be analyzed is obtained.

4. The information discrimination method based on big data risk control as described in claim 3, characterized in that, The step of analyzing the behavioral record key data corresponding to the preset type of user interaction behavior based on the behavioral content representation data of the preset type of user interaction behavior includes: Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the first statistical value of the number of behaviors corresponding to each user interaction behavior mainly performed by the risk control user to be analyzed is obtained, and the duration of each user interaction behavior is analyzed. Each user interaction behavior mainly performed by the risk control user to be analyzed means that the proportion of the user interaction data of the risk control user to be analyzed in the user interaction behavior is not less than the pre-configured reference proportion. The second statistical value of the number of user interaction behaviors corresponding to each user interaction behavior whose duration belongs to a pre-configured first reference duration interval is determined, and the attentionable behavior representation parameters corresponding to the preset type of user interaction behavior are analyzed based on the first statistical value of the number of behaviors and the second statistical value of the number of behaviors. Based on the first statistical value of the number of behaviors and the characteristic parameters of the attentionable behaviors, the key behavioral record data corresponding to the preset type of user interaction behavior is analyzed.

5. The information discrimination method based on big data risk control as described in claim 3, characterized in that, The step of analyzing the behavior-specific key data corresponding to the preset type of user interaction behavior based on the behavior content representation data corresponding to the preset type of user interaction behavior includes: Based on the behavioral content representation data corresponding to the preset type of user interaction behavior, the main user interaction behaviors of the risk control user to be analyzed are identified, and the target behavioral content representation data in each user interaction behavior is extracted. The main user interaction behaviors of the risk control user to be analyzed refer to the user interaction data volume of the risk control user to be analyzed in the user interaction behavior, which accounts for a proportion of the user interaction data volume of the risk control user to be analyzed that is not less than a pre-configured reference proportion. The target behavior content representation data is segmented to form corresponding important data segments and occurrence representation parameters for each important data segment. Furthermore, a cluster of interactive content data segments corresponding to each pre-configured data dimension is extracted. The data dimension refers to the data format of the corresponding behavior content representation data. The occurrence representation parameters are used to reflect the number of important data segments that are the same as the corresponding important data segments. Analyze the target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance, and mark the important data fragments corresponding to the target interactive content data fragment clusters as behavioral specific key data corresponding to the preset type of user interaction behavior.

6. The information discrimination method based on big data risk control as described in claim 5, characterized in that, The step of analyzing the cluster of target interactive content data fragments whose semantic relevance to each of the important data fragments is greater than or equal to a predetermined reference semantic relevance includes: Select the target importance data segment with the maximum value of the occurrence representation parameter among the various importance data segments, and analyze the target interactive content data segments whose semantic relevance to the target importance data segments is greater than or equal to a predetermined reference semantic relevance. Also, mark the interactive content data segment clusters including the target interactive content data segments as target interactive content data segment clusters whose semantic relevance to the various importance data segments is greater than or equal to a predetermined reference semantic relevance; or The clusters of pending interactive content data fragments to which each important data fragment belongs are analyzed. Based on the occurrence representation parameters corresponding to each important data fragment, the number of important data fragments included in each cluster of pending interactive content data fragments is analyzed. The clusters of pending interactive content data fragments with the maximum number of important data fragments are marked as target interactive content data fragment clusters whose semantic relevance to each important data fragment is greater than or equal to a predetermined reference semantic relevance.

7. The information discrimination method based on big data risk control as described in claim 1, characterized in that, The interactive user representation data includes user attribute representation data of various dimensions, and the behavioral content representation data corresponding to the interactive behavior of the specified type of user includes the interactive content corresponding to the interactive behavior of the interactive risk control user. The step of analyzing the interactive user description vector corresponding to each of the at least one interactive risk control users based on the interactive user representation data corresponding to each of the at least one interactive risk control users and the behavioral content representation data corresponding to the specified type of user interaction behavior performed by the at least one interactive risk control user in the preset time interval includes: For the at least one interactive risk control user, determine the user attribute representation data of each dimension included in the corresponding interactive user representation data, so as to output the key user attribute data corresponding to each interactive risk control user in the at least one interactive risk control user. Based on the behavioral content representation data corresponding to the specified type of user interaction behavior, determine the behavioral record key data and behavioral specific key data corresponding to each of the at least one interactive risk control users; Based on the user attribute key data, the behavior record key data, and the behavior specific key data corresponding to each of the at least one interactive risk control users, an interactive user description vector corresponding to each of the at least one interactive risk control users is determined.

8. An information discrimination system based on big data risk control, characterized in that, It includes a processor and a memory, the memory being used to store a computer program, and the processor being used to execute the computer program to implement the method of any one of claims 1-7.

Citation Information

Patent Citations

  • Data processing method, computer equipment and readable storage medium

    CN115511501A

  • Grade risk prevention and control method and device thereof, equipment, medium and product

    CN116128287A