Risk Control Optimization Method and System Based on User Stability

Abstract

This application relates to the fields of Internet finance and artificial intelligence technologies, and provides a risk control optimization method and system based on user stability. Through behavior intention mining and processing, it can identify downstream behavior events in user session data and assign accurate session scenario labels and behavior status categories to them, laying a foundation for the in-depth understanding and prediction of user behavior. Through the extraction of behavior trend vectors at the level of target stability jump, it can capture the subtle changes in user behavior, thereby more accurately assessing its potential risks. By comparing the behavior trend consistency weights with template behavior events, it can timely detect abnormal fluctuations in user behavior and output targeted risk control optimization suggestions. In this way, it can solve the deficiencies of traditional risk control systems in the dynamic analysis of user behavior, and provide a more accurate and efficient risk control solution for the Internet finance industry by introducing innovative data mining and risk assessment methods.

Description

Technical Field

[0001] This application relates to the fields of Internet finance and artificial intelligence technologies. Specifically, it relates to a risk control optimization method and system based on user stability. Background Art

[0002] In the context of the rapid development of Internet finance, how to accurately capture users' behavioral intentions and effectively evaluate potential risks has become an important challenge faced by the industry. Traditional risk control systems often focus on the review of users' static information while ignoring the analysis of users' dynamic behaviors, resulting in difficulties in ensuring the accuracy and timeliness of risk assessment. Summary of the Invention

[0003] To address the above problems, this application provides a risk control optimization method and system based on user stability.

[0004] In the first aspect of the embodiments of this application, a risk control optimization method based on user stability is provided, which is applied to a risk control optimization system. The method includes:

[0005] Obtain the Internet finance business session data corresponding to the users of the target platform; the Internet finance business session data includes the to-be-processed business interaction behavior information monitored for the users of the target platform based on the authorized risk control threads;

[0006] Perform behavior intention mining processing on the to-be-processed business interaction behavior information in the Internet finance business session data to obtain a behavior intention mining result; the behavior intention mining result represents the session scene labels corresponding to at least one downstream behavior event in the Internet finance business session data and the behavior status categories of each downstream behavior event;

[0007] Extract the target behavior trend vectors of each downstream behavior event at the target stability jump level according to the behavior intention mining result, and obtain the target behavior trend vectors of each downstream behavior event at the target stability jump level;

[0008] For each downstream behavior event, obtain the template behavior trend vector of the template behavior event at the target stability jump level, and determine the behavior trend consistency weight between the target behavior trend vector of the downstream behavior event and the template behavior trend vector; the template behavior event is the behavior event matching the downstream behavior event in the template business interaction behavior information corresponding to the users of the target platform;

[0009] Based on the discriminant view of the behavior trend consistency weight corresponding to the downstream behavior event and the target behavior trend setting weight, output the abnormal risk control optimization suggestion of the downstream behavior event at the target stability jump level; the target behavior trend setting weight is the behavior trend setting weight corresponding to the target stability jump level.

[0010] Preferably, the target stability jump level includes a behavior event heat index, a behavior event distribution index, and a behavior event security index; the extraction of the target behavior trend vector of each downstream behavior event at the target stability jump level according to the behavior intention mining result includes:

[0011] For each downstream behavior event, extract the positioning features of each conversation semantic block in the conversation scene label corresponding to the downstream behavior event;

[0012] According to the positioning features of each conversation semantic block, determine the behavior event heat vector and the behavior event distribution vector of the downstream behavior event;

[0013] According to the behavior state category of the downstream behavior event, determine the behavior event development trajectory of the downstream behavior event, and according to the behavior event development trajectory and the positioning features of each conversation semantic block, determine the behavior event security feature of the downstream behavior event.

[0014] Preferably, the determination of the behavior event heat vector and the behavior event distribution vector of the downstream behavior event according to the positioning features of each conversation semantic block includes:

[0015] Screen the first positioning element, the second positioning element, the first query mapping element, and the second query mapping element from the positioning features of each conversation semantic block;

[0016] According to the comparison result between the first positioning element and the second positioning element, and the comparison result between the first query mapping element and the second query mapping element, determine the behavior event heat vector of the downstream behavior event;

[0017] According to the positioning features of each conversation semantic block, determine the global positioning element and the global query mapping element, and obtain the behavior event distribution vector of the downstream behavior event.

[0018] Preferably, the determination of the behavior event security feature of the downstream behavior event according to the behavior event development trajectory and the positioning features of each conversation semantic block includes:

[0019] Determine the behavior activation attribute positioning feature and the behavior interception attribute positioning feature from the positioning features of each conversation semantic block according to the development track of the behavior event;

[0020] Determine the event relevance of the downstream behavior event according to the key-value pair feature in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature, and determine the behavior risk coefficient of the downstream behavior event according to the behavior risk feature in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature;

[0021] Take the weighted result of the event relevance and the behavior risk coefficient as the security discrimination variable, and determine the behavior event security feature of the downstream behavior event as the decoded result corresponding to the security discrimination variable.

[0022] Preferably, the method further includes:

[0023] Extract the attention vector at the user community attention level from the to-be-processed business interaction behavior information according to the behavior intention mining result, and obtain the user community attention vector of the to-be-processed business interaction behavior information at the user community attention level;

[0024] Obtain the template attention vector of the template business interaction behavior information at the user community attention level;

[0025] Output the community risk control conduction view of the to-be-processed business interaction behavior information at the user community attention level according to the attention vector comparison result between the user community attention vector and the template attention vector.

[0026] Preferably, the user community attention level includes a transaction warning heat index and a transaction warning distribution index; the extracting the attention vector at the user community attention level from the to-be-processed business interaction behavior information according to the behavior intention mining result, and obtaining the user community attention vector of the to-be-processed business interaction behavior information at the user community attention level includes:

[0027] Determine the global scenario risk control confidence of the session scenario label corresponding to each downstream behavior event in the at least one downstream behavior event, and obtain the platform user risk control confidence of the to-be-processed business interaction behavior information; the platform user risk control confidence is used as the user community attention vector of the transaction warning heat index;

[0028] Determine the target key semantic block of the to-be-processed business interaction behavior information according to the positioning feature of the conversation semantic block in the session scenario label corresponding to each downstream behavior event in the at least one downstream behavior event; the target key semantic block is used as the user community attention vector of the transaction warning distribution index.

[0029] Preferably, the output of the community risk control conduction view of the to-be-processed service interaction behavior information at the user community attention level according to the attention vector comparison result between the user community attention vector and the template attention vector includes:

[0030] Determine the feature difference weight by obtaining the feature difference between the target key semantic block and the template key semantic block, and determine the first community risk control conduction view of the to-be-processed service interaction behavior information according to the feature difference weight in the transaction warning distribution index;

[0031] Determine the confidence difference coefficient by obtaining the confidence comparison result between the platform user risk control confidence and the template transaction warning confidence;

[0032] When the confidence difference coefficient is greater than the confidence difference threshold, determine the second community risk control conduction view of the to-be-processed service interaction behavior information in the transaction warning heat index;

[0033] Output the first community risk control conduction view and the second community risk control conduction view as the community risk control conduction view of the to-be-processed service interaction behavior information.

[0034] Preferably, the user community attention level further includes a transaction warning level; the method further includes:

[0035] When the confidence difference coefficient is less than or equal to the confidence difference threshold, determine the abnormal transaction warning feature of the to-be-processed service interaction behavior information according to the positioning feature of the conversation semantic block in the session scene label corresponding to each downstream behavior event in the at least one downstream behavior event; the abnormal transaction warning feature is used as the user community attention vector at the transaction warning level;

[0036] Determine the level error by obtaining the level comparison result between the abnormal transaction warning feature and the template transaction warning feature, and determine the third community risk control conduction view of the to-be-processed service interaction behavior information at the transaction warning level according to the level error;

[0037] Output the first community risk control conduction view and the third community risk control conduction view as the community risk control conduction view of the to-be-processed service interaction behavior information.

[0038] Preferably, the method further includes:

[0039] For the behavior trend consistency weight of each of the downstream behavior events at the target stability jump level, based on the weighted result between the behavior trend consistency weight and the template behavior trend vector of the corresponding template behavior event at the target stability jump level, determine the risk control stability metric value of the downstream behavior event at the target stability jump level;

[0040] Perform an averaging process on the risk control stability metric values of each downstream behavior event in the at least one downstream behavior event at the target stability jump level to obtain the risk control stability metric value of the to-be-processed business interaction behavior information;

[0041] Based on the weighted result between the attention vector comparison result of the to-be-processed business interaction behavior information at the user community attention level and the template attention vector, determine the first transaction warning metric value of the to-be-processed business interaction behavior information at the user community attention level;

[0042] Determine the feature commonality value between the Internet financial business session data and the template business interaction behavior mapping features corresponding to the template business interaction behavior information to obtain the second transaction warning metric value of the to-be-processed business interaction behavior information;

[0043] Based on the first transaction warning metric value, the second transaction warning metric value, and the risk control stability metric value of the to-be-processed business interaction behavior information, output the transaction warning metric value of the to-be-processed business interaction behavior information.

[0044] Preferably, the obtaining of the Internet financial business session data corresponding to the target platform user includes:

[0045] In response to a data sampling request for the current risk control optimization system, perform a security authorization monitoring on the current risk control optimization system to obtain a web spider crawling result; the web spider crawling result includes at least one page session collected in the current risk control optimization system based on an authorized risk control thread, and each collected page session is within a page session flow of the current risk control optimization system;

[0046] Decompose the web spider crawling result according to the page session flow to obtain a plurality of page session data;

[0047] Screen the page session data containing transaction warning tags from the plurality of page session data to obtain at least one target page session data;

[0048] Use any one of the at least one target page session data as the Internet financial business session data.

[0049] In the second aspect of the embodiments of the present application, a risk control optimization system is provided, including: a processor, a memory and a bus connected to the processor; the processor and the memory complete communication with each other through the bus; the processor is configured to call a computer program in the memory to execute the above-mentioned risk control optimization method based on user stability.

[0050] In the third aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a program is stored, and when the program is executed by a processor, the above-mentioned risk control optimization method based on user stability is implemented.

[0051] For the risk control optimization method and system based on user stability provided by the embodiments of the present application, by real-time monitoring and recording the business interaction behaviors of users, the system can accumulate a large amount of user behavior data, providing a rich data source for subsequent behavior intention mining and risk assessment. However, how to extract valuable information from these massive data and accordingly output accurate risk predictions and optimization suggestions has become the key technical problem to be solved in the embodiments of the present application.

[0052] In response to the above problems, the embodiments of the present application introduce a series of innovative technical means. First, through behavior intention mining processing, the system can identify downstream behavior events in user session data and assign accurate session scenario labels and behavior status categories to them. This lays a foundation for in-depth understanding and prediction of user behaviors. Second, through the extraction of behavior trend vectors at the level of target stability jump, the system can capture subtle changes in user behaviors, thereby more accurately assessing their potential risks. Finally, through the comparison of the behavior trend consistency weights with template behavior events, the system can promptly detect abnormal fluctuations in user behaviors and output targeted risk control optimization suggestions. In summary, the embodiments of the present application aim to solve the deficiencies of traditional risk control systems in the dynamic analysis of user behaviors, and provide a more accurate and efficient risk control solution for the Internet finance industry by introducing innovative data mining and risk assessment methods. Description of the Drawings

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 It is a flowchart of a risk control optimization method based on user stability provided by the embodiments of the present application.

[0055] Figure 2Schematic diagram of product modules of a risk control optimization system provided by an embodiment of the present application. Detailed implementation manners

[0056] Exemplary embodiments disclosed in the present application will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0057] To better understand the above technical solutions, the technical solutions of the present application will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical solutions of the present application, rather than limitations on the technical solutions of the present application. Without conflict, the technical features in the embodiments of the present application and the embodiments can be combined with each other.

[0058] Please refer to Figure 1 , which is a flowchart of a risk control optimization method based on user stability provided by an embodiment of the present application. This method is applied to a risk control optimization system, and the specific content included in this method is described as S101 - S105.

[0059] S101. Obtain the Internet financial business session data corresponding to the users of the target platform; the Internet financial business session data includes the to - be - processed business interaction behavior information monitored for the users of the target platform based on the authorized risk control threads.

[0060] S102. Perform behavior intention mining processing on the to - be - processed business interaction behavior information in the Internet financial business session data to obtain a behavior intention mining result; the behavior intention mining result represents the session scene tags corresponding to at least one downstream behavior event in the Internet financial business session data and the behavior status category of each downstream behavior event.

[0061] S103. Extract the behavior trend vectors at the target stability jump level for each downstream behavior event according to the behavior intention mining result, and obtain the target behavior trend vectors of each downstream behavior event at the target stability jump level.

[0062] S104. For each downstream behavior event, obtain the template behavior trend vector of the template behavior event at the target stability jump level, and determine the behavior trend consistency weight between the target behavior trend vector of the downstream behavior event and the template behavior trend vector; the template behavior event is the behavior event that matches the downstream behavior event in the template business interaction behavior information corresponding to the users of the target platform.

[0063] S105. Output the abnormal risk control optimization suggestions of the downstream behavior event at the target stability jump level according to the discrimination view of the behavior trend consistency weight corresponding to the downstream behavior event and the weight of the target behavior trend; the weight of the target behavior trend setting is the weight of the behavior trend corresponding to the target stability jump level.

[0064] To facilitate the understanding of the above technical solution, first, a complete application scenario example will be introduced below, and then each of the above steps will be elaborated based on this application scenario to ensure that those skilled in the art can implement the embodiments of the present application completely and clearly.

[0065] In an exemplary application scenario, it involves a risk control optimization system corresponding to an Internet finance company, and this risk control optimization system is used to improve business security and user experience. In this application scenario, the risk control optimization system first obtains the session data of a user (referred to as User A) on the target platform in Internet finance business. These data include the to-be-processed business interaction behavior information generated by User A under the monitoring of the authorized risk control thread, such as detailed data on page browsing records, transaction operations, capital transfers, etc.

[0066] Next, the risk control optimization system conducts in-depth behavior intention mining processing on the to-be-processed business interaction behavior information in these session data. The system uses advanced algorithm models, such as machine learning classifiers, to analyze the behavior patterns of User A in the session, identify and classify different downstream behavior events. For example, browsing specific financial products, making investment operations, cash withdrawals, etc., and assigns session scenario labels (such as "information collection before investment", "investment decision", "fund cash withdrawal", etc.) and behavior status categories (such as "normal browsing", "abnormally frequent queries", "high-risk cash withdrawal", etc.) to these events.

[0067] After obtaining the behavior intention mining results, the risk control optimization system further extracts the behavior trend vectors of each downstream behavior event at the target stability jump level. This means that the system will analyze the stability and change trends of these behaviors in the time series. For example, in the "information collection before investment" scenario, whether the browsing behavior of User A is continuously stable or suddenly undergoes a significant change (i.e., a jump) at a certain time point, so as to obtain the target behavior trend vectors of each downstream behavior event at the target stability jump level.

[0068] Subsequently, the risk control optimization system will obtain the behavior trend vectors of the template behavior events that match the current downstream behavior event from the template business interaction behavior information of User A or similar users. These template behavior events are standard behavior patterns constructed based on a large amount of historical data. The system determines the consistency weight between them, that is, the similarity or deviation degree between the two, by comparing the target behavior trend vector and the template behavior trend vector.

[0069] Finally, based on the comparison result of the behavior trend consistency weight of the downstream behavior event and the weight set for the target behavior trend, the risk control optimization system will output abnormal risk control optimization suggestions. For example, if the behavior trend consistency weight of a certain downstream behavior event is much lower than the weight set for the target behavior trend, the system can suggest adding additional identity verification steps or restricting certain high-risk operations to enhance security. Conversely, if the consistency weight is high, the system may suggest reducing unnecessary verification links to improve the user experience.

[0070] Designed in this way, the risk control optimization system can not only monitor and identify potential risk behaviors in real time, but also provide targeted optimization suggestions to help Internet financial enterprises better balance business security and user experience.

[0071] Based on the above application scenario, the target platform user refers to a specific customer who uses the services or products of an Internet financial company. For example, User A is a specific user who registers and conducts financial transactions (such as investment, lending, payment, etc.) on this Internet financial platform. The risk control optimization system will pay special attention to the behaviors of such users to ensure the security and compliance of their transactions.

[0072] Internet financial business session data refers to all data generated during the interaction between users and the Internet financial platform. These data include but are not limited to user login information, page browsing history, transaction records, fund transfer details, search queries, click behaviors, etc. These data can be organized and stored in units of sessions, and each session represents all interaction behaviors of the user during one login or continuous activity.

[0073] Authorized risk control threads refer to the security threads in the Internet financial platform that are responsible for real-time monitoring and recording of user behaviors. These threads operate on the basis of the user's consent to the privacy policy and service terms to legally collect and analyze user data, aiming to identify and prevent potential risk behaviors.

[0074] Pending business interaction behavior information refers to the original user interaction data captured from authorized risk control threads that has not been deeply analyzed. These data may contain a lot of noise and irrelevant information and need to be further processed and analyzed to extract meaningful user behavior patterns and intentions.

[0075] Specifically, in S101, the primary task of the risk control optimization system is to obtain the session data generated by the target platform users (such as user A) in the Internet finance business. These data are monitored and recorded in real time by authorized risk control threads when users interact with the platform. For example, when user A logs in to the Internet finance platform and starts various operations, such as browsing financial products, conducting transactions, or querying account information, the authorized risk control threads will run silently in the background, capturing and recording every detail of these interaction behaviors. These details include, but are not limited to, which links user A clicks on, how long user A stays on which pages, which keywords user A searches for, which transactions user A conducts, and the specific amounts and times of the transactions. These raw data are then integrated into the form of session data, where each session represents the continuous activities of user A within a certain time period. These session data not only contain all the operation records of user A but also imply important information about their behavior patterns and intentions, which are the basis for subsequent risk control analysis. The risk control optimization system will collect these session data and prepare for more in-depth analysis and processing. The key to this step is to ensure the integrity and accuracy of the data, because any omission or error may lead to deviations or misjudgments in subsequent analysis. At the same time, the system also needs to comply with relevant data protection and privacy regulations to ensure the legality and security of user data. Through this step, the risk control optimization system provides a rich data source for subsequent advanced functions such as behavior intention mining and behavior trend analysis, laying a solid foundation for improving the security of the business and the user experience.

[0076] Furthermore, behavior intention mining processing is used to mine the potential behavior intentions of users from a large amount of user interaction behavior data. In the field of Internet finance, this may involve in-depth analysis of user session data to identify the intentions, preferences, and possible next actions of users in each interaction link. Through this behavior intention mining processing, enterprises can better understand user needs, optimize service processes, improve user experience, and strengthen risk control.

[0077] The behavior intention mining result refers to the output result obtained after behavior intention mining processing. It may contain information about various behavior intentions shown by users in Internet finance business sessions. These information can be further analyzed and interpreted to reveal users' potential needs, behavior patterns, and risk preferences. The behavior intention mining result is an important basis for subsequent risk control optimization and personalized service recommendation.

[0078] During the interaction between users and Internet financial platforms, downstream behavior events refer to the subsequent behaviors that users may trigger after completing a certain operation or reaching a certain state. For example, users may perform investment operations after browsing financial products, or withdraw funds after completing a transaction. These downstream behavior events are the focus of the risk control optimization system because they are often closely related to the user's risk level and business requirements.

[0079] Session scenario tags refer to the tags assigned to the scenarios where users are located based on their specific behaviors in Internet financial business sessions. For example, when a user is browsing different types of financial products, they can be tagged with "information collection before investment"; when a user is performing actual investment operations, they can be tagged with "investment decision-making". These session scenario tags help the risk control optimization system understand the user's current behaviors and intentions more precisely.

[0080] Behavior status categories classify the behavior status of users in the current session scenario. For example, in the "information collection before investment" scenario, a user's behavior may be classified as "normal browsing" or "abnormally frequent queries"; in the "investment decision-making" scenario, a user's behavior may be classified as "prudent investment" or "high-risk investment". These behavior status categories provide important information about the risk and preference of users' behaviors for the risk control optimization system.

[0081] In S102, the risk control optimization system will conduct in-depth mining and processing of the obtained Internet financial business session data. This step is a key link in subsequent risk control analysis, aiming to extract valuable behavior patterns and intention information from a large amount of user interaction data. For example, the system will use advanced algorithm models, such as machine learning classifiers, clustering algorithms, etc., to analyze the user session data in detail. These algorithms can identify various behavior characteristics shown by users in the session, such as click frequency, browsing duration, search keywords, etc., and infer the potential behavior intentions of users based on this. Through the mining and processing of behavior intentions, the system can identify and classify different downstream behavior events. For example, the system may find that users often perform investment operations after browsing a certain type of financial product, which forms a typical downstream behavior event. At the same time, the system will also assign corresponding session scenario tags and behavior status categories to these events to more precisely describe the behaviors and intentions of users in the current scenario. The results of behavior intention mining not only include the downstream behavior events that users may trigger, but also reveal the occurrence probabilities and risk levels of these events in different session scenarios. This information is crucial for subsequent risk control optimization because it helps the system more accurately predict the behavior trends of users, so as to take corresponding risk control measures in a timely manner.

[0082] Furthermore, the target stability jump level refers to that when analyzing user behavior, special attention is paid to the stability and sudden changes (jumps) of user behavior in the time series. In the field of Internet finance, some behaviors of users can change significantly in a short period of time, and such changes may imply potential risks or opportunities. For example, a user who usually rarely makes large transactions suddenly makes frequent large transfers, and this jump in behavior may be a risk signal that needs attention. Therefore, the target stability jump level is an important dimension in the risk control system for identifying and evaluating abnormal or risky user behavior.

[0083] Behavior trend vector extraction refers to extracting a mathematical vector from the user's behavior data that can represent their behavior trend. The behavior trend vector can reflect the behavior characteristics of the user over a period of time, including the frequency, intensity, duration, etc. of the behavior. By extracting the behavior trend vector, the risk control system can more accurately depict the user's behavior pattern and then predict their future behavior trend.

[0084] The target behavior trend vector refers to the mathematical vector corresponding to the specific behavior trend shown by the user in a certain specific time period or scenario. It is a quantitative description of the user's behavior characteristics in that time period or scenario and can be used to compare with the behavior in other time periods or scenarios to discover abnormal or changing trends in the user's behavior. In the risk control system, the target behavior trend vector is an important basis for evaluating the stability and riskiness of user behavior.

[0085] In S103, based on the behavior intention mining results obtained in the previous step S102, the risk control optimization system extracts the behavior trend vectors at the level of target stability jump for each downstream behavior event. The core purpose of this step is to quantify the behavior stability and change trend of users in each downstream behavior event, so as to provide data support for subsequent risk assessment and optimization suggestions. For example, the system will analyze the time series characteristics of each downstream behavior event in the user session data. By statistically calculating key indicators such as the behavior frequency and intensity of users at each time point, the system can construct a mathematical vector representing the behavior trend of this behavior event, that is, the target behavior trend vector. This vector not only reflects the overall behavior characteristics of users in a certain downstream behavior event, but also reveals the stability of users' behavior over time and possible jump points. For example, in the downstream behavior event of "information collection before investment", if the browsing behavior of users remains stable for a period of time, but suddenly there is a significant increase in the number of views at a certain time point, then this jump will be captured by the behavior trend vector. Through the extraction of the behavior trend vector in S103, the risk control optimization system can more accurately grasp the behavior dynamics of users in each downstream behavior event, providing strong data support for subsequent risk identification, prediction and optimization suggestions. The completion of this step also marks that the system's in-depth understanding and analysis of users' behavior has entered a new stage.

[0086] Next, a template behavior event refers to a standard or typical behavior pattern constructed based on a large amount of historical data and user behavior analysis in the Internet finance business. These behavior patterns represent the regular or expected behaviors of users in specific scenarios, such as normal transfer operations, standard investment consultation processes, etc. The template behavior event provides a reference benchmark for the risk control system to compare and identify whether the user's current behavior is abnormal.

[0087] The template behavior trend vector is a quantitative description of the template behavior event at the level of target stability jump. It represents the stability and change trend of the template behavior event in the time series through a series of numerical values, reflecting the regular or standard pattern of this behavior. The template behavior trend vector is obtained by analyzing and refining a large amount of historical behavior data and is used to compare with the actual behavior trend of users.

[0088] The behavior trend consistency weight is an indicator to measure the similarity or consistency between the user's current behavior trend and the template behavior trend vector. The higher this weight value, the closer the user's current behavior is to the standard behavior pattern, and the relatively lower the risk; on the contrary, the lower the weight value, the greater the deviation between the user's behavior and the standard pattern, and there may be potential risks.

[0089] The template business interaction behavior information refers to a set of standard business interaction processes and behavior patterns constructed based on historical data and business rules. It includes the standard steps and behavior norms that users should follow when performing various operations on the Internet financial platform. These template information provide a basis for the risk control system to evaluate the compliance and risk of user behavior.

[0090] In S104, the risk control optimization system will conduct an in-depth analysis of each downstream behavior event. By comparing its behavior trend vector with that of the template behavior event at the target stability jump level, it determines the behavior trend consistency weight between the two. First, for each downstream behavior event, the system will find the matching template behavior event from the template business interaction behavior information to which it belongs. These template behavior events are constructed based on historical data and business rules and represent the standard or expected behavior patterns for this downstream behavior event. Next, the system will obtain the template behavior trend vectors of these template behavior events at the target stability jump level. These vectors are quantitative descriptions of the stability and change trends of the template behavior events in the time series, providing a benchmark for subsequent comparisons. Then, the system will compare the target behavior trend vector of the current downstream behavior event with the corresponding template behavior trend vector. This comparison process aims to measure the similarity or consistency between the two, thereby obtaining the behavior trend consistency weight. This weight value will directly reflect the degree of coincidence between the current behavior and the standard behavior pattern. Finally, based on the comparison results, the system can identify those downstream behavior events that deviate significantly from the standard behavior pattern. These events may hide potential risks or anomalies and require further risk assessment and handling. Through the analysis and comparison in S104, the risk control optimization system can more accurately identify the anomalies and potential risks in user behavior, providing strong data support for subsequent risk control measures and optimization suggestions.

[0091] Finally, the target behavior trend setting weight refers to a standard or expected value set for each downstream behavior event at the target stability jump level, used to measure whether the stability and change trend of this behavior event in the time series conform to the preset safety or normal standards. This weight value is set comprehensively based on historical data, business rules, and risk control requirements and represents the expected level of the system for the stability and change trend of this behavior event. When the actually monitored behavior trend deviates significantly from this set weight, it may indicate the existence of potential risks or anomalies.

[0092] A discrimination perspective refers to a judgment criterion or perspective adopted when comparing the weight of behavioral trend consistency with the set weight of the target behavioral trend. It helps the risk control system determine when the behavioral trend of a downstream behavioral event is regarded as abnormal or risky. The discrimination perspective can include various methods such as absolute threshold judgment (e.g., if the weight of behavioral trend consistency is lower than a certain fixed value, it is regarded as abnormal) and relative change judgment (e.g., a sudden significant decrease in the weight of behavioral trend consistency is regarded as abnormal), depending on business requirements and risk tolerance.

[0093] Abnormal risk control optimization suggestions refer to the targeted improvement opinions or measures proposed by the risk control optimization system when detecting significant differences between the behavioral trend of a downstream behavioral event and the set weight. These suggestions aim to help the system or management adjust the risk control strategy in a timely manner to reduce potential risks and enhance system security and user experience. Abnormal risk control optimization suggestions can include, but are not limited to, adding identity verification steps, restricting high-risk operations, and reminding users to pay attention to account security.

[0094] In S105, the risk control optimization system will output targeted abnormal risk control optimization suggestions based on the comparison result between the weight of behavioral trend consistency of the downstream behavioral event and the set weight of the target behavioral trend. First, the system will review the weight of behavioral trend consistency of each downstream behavioral event at the level of target stability jump. This weight reflects the similarity or deviation between the user's current behavior and the standard behavior pattern and is an important basis for evaluating behavioral risks. Then, the system will compare these weights of behavioral trend consistency with the set weight of the target behavioral trend. The set weight of the target behavioral trend is the expected value of stability and change trend set by the system for each downstream behavioral event according to historical data, business rules, and risk control requirements. During the comparison process, the system will adopt a preset discrimination perspective to judge whether the current behavioral trend is abnormal or risky. For example, if the weight of behavioral trend consistency of a certain downstream behavioral event is much lower than the set weight of the target behavioral trend, then according to the discrimination perspective, this behavioral event may be regarded as abnormal or high-risk. Finally, based on the above comparison and discrimination results, the risk control optimization system will output corresponding abnormal risk control optimization suggestions. These suggestions aim to help the system or management adjust the risk control strategy in a timely manner to cope with potential risks. Specific suggestions can include adding additional identity verification steps to improve account security, restricting or monitoring certain high-risk operations to prevent fraud, or sending security prompt messages to users to enhance their risk awareness. Through the analysis and judgment of S105, the risk control optimization system can timely detect and respond to abnormalities and risks in user behavior, thus ensuring the security of Internet financial services and the legitimate rights and interests of users.

[0095] Embodiments of the present application achieve precise grasp of user behavior intentions and risk assessment by deeply mining Internet finance business session data. Specifically, embodiments of the present application not only capture every interaction of users on the financial platform, but also extract users' behavior intentions and potential trends from these interactions. Such in-depth data analysis enables the system to more comprehensively understand user needs, predict their future behavior trends, and thus provide more considerate and secure financial services for users.

[0096] At the same time, embodiments of the present application introduce the concept of target stability jump in behavior trend analysis. By comparing the consistency between user behavior and template behavior, abnormal fluctuations in user behavior are accurately identified. Such fluctuations may indicate potential risk points, such as fraud behavior, account theft, etc., thus providing strong data support for enterprise risk control.

[0097] In addition, embodiments of the present application also have high flexibility and adaptability. The system can adjust the extraction method of behavior trend vectors and the calculation method of consistency weights according to different business scenarios and user groups to ensure the accuracy and effectiveness of risk assessment. Such a personalized risk assessment strategy not only improves the user experience, but also reduces potential economic losses and legal risks for enterprises.

[0098] In summary, embodiments of the present application significantly improve the risk control ability and user service quality of Internet finance business by comprehensively applying technical means such as big data analysis, behavior intention mining, and target stability jump analysis, providing strong technical support for the digital transformation of the financial industry.

[0099] In some alternative embodiments, the target stability jump level includes a behavior event heat index, a behavior event distribution index, and a behavior event security index; the extraction of the target behavior trend vector of each downstream behavior event at the target stability jump level according to the behavior intention mining result includes: for each downstream behavior event, extracting the positioning features of each conversation semantic block in the session scenario label corresponding to the downstream behavior event; determining the behavior event heat vector and the behavior event distribution vector of the downstream behavior event according to the positioning features of the conversation semantic blocks; determining the behavior event development trajectory of the downstream behavior event according to the behavior state category of the downstream behavior event, and determining the behavior event security feature of the downstream behavior event according to the behavior event development trajectory and the positioning features of the conversation semantic blocks.

[0100] Based on this embodiment, the risk control optimization system further refines the analysis dimensions at the level of target stability jumps, specifically including the heat index of behavior events, the distribution index of behavior events, and the security index of behavior events. These indexes provide a more comprehensive user behavior evaluation framework for the risk control optimization system, which helps to more accurately identify potential risk points.

[0101] When implementing these analyses, the risk control optimization system first carefully processes each downstream behavior event. The system extracts the positioning features of each conversational semantic block in the session scene label corresponding to the downstream behavior event. These positioning features may include the click position of the user on the page, the stay time, etc., which provide basic data for analyzing the user's behavior pattern.

[0102] Next, based on these positioning features, the risk control optimization system determines the behavior event heat vector and the behavior event distribution vector of the downstream behavior event. The behavior event heat vector reflects the frequency of user activities in a specific area, such as the click heat of a certain function area; while the behavior event distribution vector reveals the spatial distribution of user behavior, such as the access order and frequency of the user in different areas of the page.

[0103] In addition, the risk control optimization system also determines the development trajectory of the behavior event according to the behavior status category of the downstream behavior event. This trajectory shows the change process of user behavior over time, which helps the system understand the continuity and change trend of user behavior. Combining the positioning features of each conversational semantic block, the system further determines the behavior event security features of the downstream behavior event, which may include identification marks of potential risk behaviors such as abnormal logins and unconventional transactions.

[0104] Through the above steps, the risk control optimization system can build a multi-dimensional user behavior analysis model. This model not only considers the heat, distribution, and development trajectory of user behavior, but also incorporates the analysis of security features, thus providing a more three-dimensional and comprehensive risk assessment for users. In this way, the risk control optimization system can understand user behavior more deeply, predict potential risks more accurately, and provide more personalized risk control suggestions for users.

[0105] In some other preferred embodiments, determining the behavior event heat vector and the behavior event distribution vector of the downstream behavior event according to the positioning features of the respective conversation semantic blocks includes: screening a first positioning element, a second positioning element, a first query mapping element, and a second query mapping element from the positioning features of the respective conversation semantic blocks; determining the behavior event heat vector of the downstream behavior event according to the comparison result between the first positioning element and the second positioning element and the comparison result between the first query mapping element and the second query mapping element; determining a global positioning element and a global query mapping element according to the positioning features of the respective conversation semantic blocks, and obtaining the behavior event distribution vector of the downstream behavior event.

[0106] In this embodiment, when determining the behavior event heat vector and the behavior event distribution vector of the downstream behavior event, the risk control optimization system adopts a more refined method.

[0107] First, the system will screen out key elements from the positioning features of each conversation semantic block. These elements include a first positioning element, a second positioning element, a first query mapping element, and a second query mapping element. These elements are all extracted through in-depth analysis of the conversation data and can reflect the key behavior characteristics of the user during the conversation.

[0108] Next, the risk control optimization system will determine the behavior event heat vector of the downstream behavior event according to the comparison result between the first positioning element and the second positioning element and the comparison result between the first query mapping element and the second query mapping element. This heat vector is actually a quantitative indicator that can reflect the behavior activity of the user in a specific area. For example, if the click frequency of the user in a certain function area is significantly higher than that in other areas, then the behavior event heat vector of this area will be relatively high.

[0109] To obtain the behavior event distribution vector, the risk control optimization system will further analyze the positioning features of each conversation semantic block, extract the global positioning element and the global query mapping element from them. These global elements can reflect the behavior distribution of the user during the entire conversation. The system will construct the behavior event distribution vector according to these global elements and in combination with the analysis result of the behavior event heat vector. This vector not only reveals the distribution of the user's behavior in the global scope but also helps the system identify the patterns and trends of the user's behavior.

[0110] Through this method, the risk control optimization system can more accurately evaluate the user's behavior intention and potential risks. For example, if the system detects that a certain user frequently accesses a high-risk area within a short period of time, and both the behavior event heat vector and the distribution vector show abnormal patterns, then the system can issue a warning in a timely manner to prompt relevant personnel to pay attention to the subsequent behavior of this user. This makes the risk control optimization system more accurate and efficient when processing Internet financial business session data. By deeply analyzing the positioning characteristics and query mapping elements of session data, the system can more comprehensively understand the user's behavior patterns and intentions, thereby providing more personalized services for users and effectively reducing potential risks.

[0111] In some other alternative embodiments, determining the behavior event security feature of the downstream behavior event based on the behavior event development trajectory and the positioning characteristics of each session semantic block includes: determining the behavior activation attribute positioning feature and the behavior interception attribute positioning feature from the positioning characteristics of each session semantic block according to the behavior event development trajectory; determining the event relevance of the downstream behavior event based on the key-value pair features in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature, and determining the behavior risk coefficient of the downstream behavior event based on the behavior risk features in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature; using the weighted result of the event relevance and the behavior risk coefficient as a security discrimination variable, and determining the feature decoding result corresponding to the security discrimination variable as the behavior event security feature of the downstream behavior event.

[0112] In this embodiment, when determining the behavior event security feature of the downstream behavior event, the risk control optimization system adopts a more refined and scientific analysis method.

[0113] First of all, the risk control optimization system will accurately extract the behavior activation attribute positioning feature and the behavior interception attribute positioning feature from the positioning characteristics of each session semantic block according to the behavior event development trajectory. These positioning features can reflect the activation state of the user when performing a specific behavior and the possible risk interception situations.

[0114] Next, the system will determine the event relevance of the downstream behavior event according to the key-value pair features in these positioning features. The key-value pair feature is a data structure that can identify the connection between different events. By analyzing these features, the system can judge which behavior events are related to each other, and then understand the overall behavior logic of the user.

[0115] At the same time, the risk control optimization system will also use the behavior risk features in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature to determine the behavior risk coefficient of the downstream behavior event. This risk coefficient is a quantitative indicator that can intuitively reflect the risk level of the user's current behavior.

[0116] After obtaining the event relevance and behavior risk coefficient, the risk control optimization system will weight these two factors to obtain a security discrimination variable. This variable comprehensively considers the relevance and risk of user behavior, and thus can more comprehensively evaluate the security of user behavior.

[0117] Finally, the system will decode the features corresponding to the security discrimination variable, and the result obtained is the behavior event security feature of the downstream behavior event. This feature not only includes the security evaluation of user behavior, but also can provide an important reference basis for subsequent risk control and optimization.

[0118] Through this implementation method, the risk control optimization system can more accurately identify potential risks in user behavior, take effective risk control measures in a timely manner, so as to ensure the security and stability of financial transactions. This not only helps to improve the user experience, but also can create greater commercial value for financial institutions. At the same time, this embodiment also enhances the flexibility and adaptability of the system, enabling it to formulate personalized risk control strategies according to the behavior characteristics and risk levels of different users.

[0119] In some alternative embodiments, the method further includes: extracting an attention vector at the user community attention level for the to-be-processed business interaction behavior information according to the behavior intention mining result, to obtain a user community attention vector of the to-be-processed business interaction behavior information at the user community attention level; obtaining a template attention vector of the template business interaction behavior information at the user community attention level; and outputting a community risk control conduction view of the to-be-processed business interaction behavior information at the user community attention level according to the attention vector comparison result between the user community attention vector and the template attention vector.

[0120] Based on this embodiment, the risk control optimization system further introduces the concept of user community attention to more comprehensively evaluate the risk of business interaction behavior. This step not only considers the behavior characteristics of individual users, but also takes the user into consideration in a broader community context.

[0121] For example, the risk control optimization system will first extract an attention vector at the user community attention level for the to-be-processed business interaction behavior information according to the behavior intention mining result. The "attention vector" mentioned here is a quantitative indicator that measures the attention of the user community to a specific business interaction behavior. The system extracts a vector reflecting the user community attention by analyzing data such as the interaction behavior, topic discussion, likes, and forwards of users in the community.

[0122] After extracting the user community attention vector of the business interaction behavior information to be processed, the risk control optimization system will also obtain the template attention vector of the template business interaction behavior information at the same level. The template business interaction behavior information refers to those business interaction behaviors that have been recognized as normal or abnormal by the system, and their attention vectors can be used as a comparison benchmark.

[0123] Next, the system will compare the attention vectors. By calculating the similarity or difference degree between the user community attention vector and the template attention vector, the system can determine whether the attention degree of the business interaction behavior to be processed in the user community is consistent with the template behavior. If there is a significant difference between the two, it may mean that there are abnormalities or risks in the business interaction behavior to be processed.

[0124] Finally, the risk control optimization system will output the community risk control conduction view of the business interaction behavior information to be processed at the user community attention level based on the comparison result of the attention vectors. This view is actually a comprehensive evaluation result, which takes into account the overall attitude and reaction of the user community towards the behavior to be processed. If the community's attention to this behavior is extremely high or extremely low, the system will regard it as a potential risk signal and adjust the risk control strategy accordingly.

[0125] Through these embodiments, the risk control optimization system can more comprehensively evaluate the risks of business interaction behaviors. It not only considers the behavior characteristics of individual users but also introduces the concept of community attention, thereby improving the accuracy and effectiveness of risk assessment. This risk control method based on community attention helps to detect and respond to potential risk behaviors in a timely manner, ensuring the security and stability of financial transactions. At the same time, this method also improves the user experience because the system can more accurately identify the needs and preferences of users and provide more personalized services.

[0126] In the next steps, the user community attention level includes a transaction warning heat index and a transaction warning distribution index; the extraction of the attention vector of the business interaction behavior information to be processed at the user community attention level based on the behavior intention mining result, and obtaining the user community attention vector of the business interaction behavior information to be processed at the user community attention level includes: determining the global scenario risk control confidence of each downstream behavior event corresponding to the session scenario label in the at least one downstream behavior event, and obtaining the platform user risk control confidence of the business interaction behavior information to be processed; the platform user risk control confidence is used as the user community attention vector of the transaction warning heat index; based on the positioning characteristics of the discourse semantic blocks in the session scenario labels corresponding to each downstream behavior event in the at least one downstream behavior event, determining the target key semantic block of the business interaction behavior information to be processed; the target key semantic block is used as the user community attention vector of the transaction warning distribution index.

[0127] In this embodiment, the risk control optimization system further refines the analysis at the level of user community attention, specifically introducing a trading warning heat index and a trading warning distribution index. These two indexes respectively reflect the attention degree and risk distribution of the user community to business interaction behaviors from different perspectives.

[0128] First, in order to obtain the user community attention vector of the trading warning heat index, the risk control optimization system determines the global scenario risk control confidence of the session scenario labels corresponding to each downstream behavior event in at least one downstream behavior event. The global scenario risk control confidence mentioned here is a comprehensive evaluation index, which considers multiple risk factors in the session scenario, such as user behavior patterns, historical transaction records, credit scores, etc., and represents the risk level in this scenario in a quantitative form. By calculating the global scenario risk control confidence of the session scenarios corresponding to each downstream behavior event, the system can obtain a comprehensive risk control confidence of platform users. This confidence reflects the risk perception and warning heat of the entire user community in the current business interaction behavior, and is therefore used as the user community attention vector of the trading warning heat index.

[0129] Second, in order to obtain the user community attention vector of the trading warning distribution index, the risk control optimization system determines the target key semantic block of the business interaction behavior information to be processed according to the positioning characteristics of the discourse semantic blocks in the session scenario labels corresponding to each downstream behavior event. The discourse semantic block is the key information unit in the session data, which contains important semantic information in the user interaction behavior. By analyzing the positioning characteristics of these semantic blocks, such as occurrence frequency, positional relationship, etc., the system can identify the target key semantic blocks that play a key role in the current business interaction behavior. These key semantic blocks reflect the focus of attention and risk distribution of the user community in specific interaction links, and are therefore used as the user community attention vector of the trading warning distribution index.

[0130] To sum up, by introducing the trading warning heat index and the trading warning distribution index, the risk control optimization system can more comprehensively evaluate the attention degree and risk perception of the user community to the current business interaction behavior. This not only helps the system to timely discover potential risk points, but also provides strong data support for subsequent risk control strategies. At the same time, by refining the analysis at the level of user community attention, the system can more accurately grasp the user needs and preferences, and then provide more personalized services to improve the user experience.

[0131] In some other preferred embodiments, the community risk control conduction view of the to-be-processed service interaction behavior information at the user community attention level output according to the attention vector comparison result between the user community attention vector and the template attention vector includes: determining the feature difference weight between the target key semantic block and the template key semantic block, and determining the first community risk control conduction view of the to-be-processed service interaction behavior information in the transaction warning distribution index according to the feature difference weight; determining the confidence comparison result between the platform user risk control confidence and the template transaction warning confidence to obtain a confidence difference coefficient; when the confidence difference coefficient is greater than the confidence difference threshold, determining the second community risk control conduction view of the to-be-processed service interaction behavior information in the transaction warning heat index; and outputting the first community risk control conduction view and the second community risk control conduction view as the community risk control conduction view of the to-be-processed service interaction behavior information.

[0132] Specifically, through refined comparison and analysis, the risk control optimization system outputs the community risk control conduction view of the to-be-processed service interaction behavior information at the user community attention level. This process involves the comparison between the target key semantic block and the template key semantic block, as well as the comparison between the platform user risk control confidence and the template transaction warning confidence.

[0133] First, in order to form the first community risk control conduction view in the transaction warning distribution index, the risk control optimization system determines the feature difference between the target key semantic block and the template key semantic block. The feature difference in the embodiments of the present application may refer to the differences in semantic content, occurrence frequency, correlation degree, etc. between the two. The system calculates these differences to obtain a feature difference weight. This weight reflects the similarity or deviation degree between the target key semantic block and the template key semantic block, and further helps the system judge the risk situation of the to-be-processed service interaction behavior in the transaction warning distribution. Based on this feature difference weight, the system can determine the first community risk control conduction view of the to-be-processed service interaction behavior information in the transaction warning distribution index.

[0134] Second, in order to form the second community risk control conduction view in the transaction warning heat index, the system determines the confidence comparison result between the platform user risk control confidence and the template transaction warning confidence. This comparison is achieved by calculating the difference coefficient between the two, which reflects the deviation degree between the risk perception of the current service interaction behavior of the platform user and the template transaction warning confidence. When this difference coefficient is greater than a preset confidence difference threshold, it means that the risk level of the current service interaction behavior significantly deviates from the normal range. At this time, the system determines the second community risk control conduction view of the to-be-processed service interaction behavior information in the transaction warning heat index.

[0135] Finally, the risk control optimization system will combine the first community risk control conduction view and the second community risk control conduction view, and output them as the community risk control conduction view of the business interaction behavior information to be processed. This comprehensive view not only considers the risk situation of business interaction behavior in the transaction warning distribution, but also incorporates the risk assessment results on the transaction warning heat index, thus providing a comprehensive and multi-dimensional risk judgment basis for users and financial institutions.

[0136] By implementing these preferred embodiments, the risk control optimization system can more accurately identify and evaluate potential risks in business interaction behavior. This not only helps to improve the security of financial transactions, reduce the occurrence of fraud and improper behavior, but also enhances users' trust in financial institutions. At the same time, through refined risk control strategies, the system improves the flexibility and effectiveness of risk response, creating greater commercial value for financial institutions.

[0137] In some examples, the user community attention level further includes the transaction warning level; the method further includes: when the confidence difference coefficient is less than or equal to the confidence difference threshold, determining the abnormal transaction warning feature of the business interaction behavior information to be processed according to the positioning features of the discourse semantic blocks in the session scene labels corresponding to each downstream behavior event in the at least one downstream behavior event; using the abnormal transaction warning feature as the user community attention vector at the transaction warning level; determining the level comparison result between the abnormal transaction warning feature and the template transaction warning feature to obtain a level error, and determining the third community risk control conduction view of the business interaction behavior information to be processed at the transaction warning level according to the level error; outputting the first community risk control conduction view and the third community risk control conduction view as the community risk control conduction view of the business interaction behavior information to be processed.

[0138] Based on this example, the risk control optimization system further considers the transaction warning level in the user community attention level to more carefully evaluate the risks of business interaction behavior. The introduction of this level enables the system to formulate more precise risk control strategies according to different levels of risk warnings.

[0139] When the confidence difference coefficient is less than or equal to the confidence difference threshold, it means that the risk of the business interaction behavior to be processed on the transaction warning heat index does not deviate significantly from the normal range. However, this does not mean that the behavior is completely risk-free. Therefore, the risk control optimization system will further determine the abnormal transaction warning feature of the business interaction behavior information to be processed according to the positioning features of the discourse semantic blocks in the session scene labels corresponding to each downstream behavior event in the at least one downstream behavior event.

[0140] The abnormal transaction warning features in the embodiments of the present application may refer to features such as abnormal occurrence frequency, abnormal semantic content, or abnormal relevance to other session scenario tags in the session scenario. These features may imply potential risks. The system uses these abnormal transaction warning features as the user community attention vectors at the transaction warning level for subsequent risk assessment.

[0141] Next, the risk control optimization system will determine the level comparison result between the abnormal transaction warning features and the template transaction warning features. The template transaction warning features represent the features of business interaction behaviors with known risk levels. By comparing the differences between the two, the system can calculate a level error, which reflects the risk level difference between the business interaction behavior to be processed and the behavior with a known risk level.

[0142] Based on this level error, the system can determine the third community risk control conduction view of the business interaction behavior to be processed at the transaction warning level. This view is a comprehensive assessment of the risk level of the current business interaction behavior. It takes into account the differences between the abnormal transaction warning features and the template features, thus providing a strong basis for subsequent risk control decisions.

[0143] Finally, the risk control optimization system will combine the first community risk control conduction view (based on the transaction warning distribution index) and the third community risk control conduction view (based on the transaction warning level) and output them as the community risk control conduction view of the business interaction behavior information to be processed. This comprehensive view not only considers the risk situation of the business interaction behavior in the transaction warning distribution but also incorporates the risk assessment results at the transaction warning level, providing a more comprehensive and in-depth risk analysis perspective for users and financial institutions.

[0144] In this way, the risk control optimization system can more accurately identify and evaluate the risk levels in business interaction behaviors, thereby formulating more effective risk control strategies. This not only helps to improve the security of financial transactions but also enhances users' satisfaction and trust in the services of financial institutions. At the same time, through multi-dimensional risk assessment methods, the system enhances the flexibility and accuracy of risk response, bringing greater commercial value to financial institutions.

[0145] Under a possible technical approach, the method further includes: for the behavior trend consistency weight of each downstream behavior event at the target stability jump level, determining the risk control stability metric value of the downstream behavior event at the target stability jump level according to the weighted result between the behavior trend consistency weight and the template behavior trend vector of the corresponding template behavior event at the target stability jump level; performing an averaging process on the risk control stability metric values of each downstream behavior event in the at least one downstream behavior event at the target stability jump level to obtain the risk control stability metric value of the to-be-processed business interaction behavior information; determining the first transaction warning metric value of the to-be-processed business interaction behavior information at the user community attention level according to the weighted result between the attention vector comparison result of the to-be-processed business interaction behavior information at the user community attention level and the template attention vector; determining the feature commonality value between the Internet financial business session data and the template business interaction behavior mapping features corresponding to the template business interaction behavior information to obtain the second transaction warning metric value of the to-be-processed business interaction behavior information; and outputting the transaction warning metric value of the to-be-processed business interaction behavior information according to the first transaction warning metric value, the second transaction warning metric value, and the risk control stability metric value of the to-be-processed business interaction behavior information.

[0146] Based on this technical approach, the risk control optimization system further synthesizes information from multiple levels to comprehensively evaluate the risk of the to-be-processed business interaction behavior. This approach not only considers the user community attention level but also incorporates the characteristics of the target stability jump level and the Internet financial business session data, thereby more accurately measuring the level of transaction warning.

[0147] First, for the behavior of each downstream behavior event at the target stability jump level, the system calculates its behavior trend consistency weight. This weight reflects the similarity between the behavior and the behavior trend vector of the corresponding template behavior event at the target stability jump level. Through weighted calculation, the risk control stability metric value of each downstream behavior event at the target stability jump level can be determined. The higher this value, the more consistent the behavior is with the template behavior in terms of stability, and the relatively lower the risk.

[0148] Next, the system performs an averaging process on the risk control stability metric values of all downstream behavior events at the target stability jump level. This is done to obtain a unified risk control stability metric value representing the entire to-be-processed business interaction behavior information. This value can comprehensively reflect the overall performance of all downstream behavior events in terms of stability.

[0149] Meanwhile, the system also determines the first transaction warning metric value of the to-be-processed business interaction behavior information at the user community attention level based on the weighted result between the attention vector and the template attention vector of the to-be-processed business interaction behavior information at the user community attention level. This value reflects the attention degree and risk perception of the user community towards the current business interaction behavior.

[0150] In addition, the system also determines the feature commonality value between the Internet finance business session data and the template business interaction behavior mapping features corresponding to the template business interaction behavior information. This commonality value represents the similarity between the current business interaction behavior and the known risk patterns, thereby obtaining the second transaction warning metric value of the to-be-processed business interaction behavior information. The higher this value, the more similar the current business interaction behavior is to the known high-risk patterns, so a higher risk warning needs to be given.

[0151] Finally, the risk control optimization system comprehensively outputs the transaction warning metric value of the to-be-processed business interaction behavior information based on the first transaction warning metric value, the second transaction warning metric value, and the risk control stability metric value of the to-be-processed business interaction behavior information. This comprehensive metric value not only considers the attention of the user community and the stability of the business interaction behavior, but also compares with the known risk patterns, so it can more comprehensively and accurately evaluate the risk level of the current business interaction behavior.

[0152] In this way, the risk control optimization system can comprehensively utilize information at multiple levels to evaluate the risk of business interaction behaviors, thereby providing more accurate and personalized risk control services for users. This not only helps to improve the security of financial transactions, reduce the occurrence of fraud and improper behaviors, but also enhances users' satisfaction and trust in the services of financial institutions. At the same time, this comprehensive evaluation method also enhances the system's ability to handle complex risk scenarios and creates greater commercial value for financial institutions.

[0153] In some optional embodiments, the obtaining of the Internet finance business session data corresponding to the target platform user includes: in response to a data sampling request for the current risk control optimization system, performing security authorization monitoring on the current risk control optimization system to obtain the web spider crawling result; the web spider crawling result includes at least one page session collected in the current risk control optimization system based on the authorized risk control thread, and each collected page session is within a page session flow of the current risk control optimization system; disassembling the web spider crawling result according to the page session flow to obtain a plurality of page session data; screening the page session data containing the transaction warning label from the plurality of page session data to obtain at least one target page session data; and using any one of the at least one target page session data as the Internet finance business session data.

[0154] Based on this embodiment, when the risk control optimization system obtains the Internet financial business session data corresponding to the users of the target platform, it follows a rigorous and efficient data collection and processing process.

[0155] First of all, this process begins with the system responding to a data sampling request. When the risk control optimization system receives such a request, it will start a security authorization monitoring mechanism, which can ensure compliance with relevant privacy and security regulations during the data collection process. Through this monitoring mechanism, the system obtains a web spider crawling result. Here, the "web spider" is an automated program that can crawl information on the network and help the system collect and analyze data.

[0156] The web spider crawling result contains at least one page session collected by the authorized risk control thread in the current risk control optimization system. These page sessions are generated when users interact with the system, and each session is within a page session flow of the system. The page session flow can be understood as the data flow formed by a series of consecutive operations of the user within the system.

[0157] Next, the risk control optimization system will disassemble the web spider crawling result. This step is carried out according to the page session flow, aiming to split the mixed data into multiple independent page session data. This helps with subsequent data analysis and processing.

[0158] After the disassembly is completed, the system will screen out the page session data that contains transaction warning tags from the obtained multiple page session data. The transaction warning tags are marks set by the system for user behaviors according to certain rules, used to indicate which behaviors may pose risks. Through screening, the system can obtain at least one target page session data, which is the focus of subsequent risk control analysis.

[0159] Finally, the risk control optimization system will select any one of these target page session data as the Internet financial business session data. This data will be used in subsequent risk assessment, warning, and processing, etc.

[0160] Through this set of processes, the risk control optimization system can efficiently obtain the Internet financial business session data corresponding to the users of the target platform, providing a solid data foundation for subsequent risk control operations. At the same time, this process also ensures the accuracy and compliance of the data, reducing the risks caused by data problems. In this way, through the refined data collection and processing process, the ability of the risk control optimization system in identifying and managing financial risks is improved.

[0161] It is worth mentioning that, based on the discriminant view of the behavior trend consistency weight corresponding to the downstream behavior event and the target behavior trend setting weight, abnormal risk control optimization suggestions for the downstream behavior event at the target stability jump level are output, including: First, calculate the weight difference value between the behavior trend consistency weight corresponding to each downstream behavior event and the target behavior trend setting weight; then, according to the weight difference value, use a preset discriminant model to determine the risk deviation index of each downstream behavior event at the target stability jump level, and the risk deviation index is used to quantitatively represent the deviation degree between the downstream behavior event and the target stable state; then, for each downstream behavior event, compare its risk deviation index with a preset risk threshold: if the risk deviation index exceeds the risk threshold, it is determined that the downstream behavior event is abnormal at the target stability jump level, and corresponding abnormal risk control optimization suggestions are generated, and the abnormal risk control optimization suggestions include strengthening the monitoring frequency of the downstream behavior event, adjusting the risk assessment model parameters of the downstream behavior event, or taking additional risk verification measures for the downstream behavior event; if the risk deviation index does not exceed the risk threshold, it is determined that the downstream behavior event is normal at the target stability jump level, and no abnormal risk control optimization suggestions need to be output; finally, sort the generated abnormal risk control optimization suggestions according to the size of the risk deviation index, and give priority to processing the downstream behavior events with higher risk deviation indexes to improve the efficiency and effect of risk control optimization.

[0162] Specifically, in the operation process of the risk control optimization system, a key link is to output abnormal risk control optimization suggestions for the downstream behavior event at the target stability jump level based on the discriminant view of the behavior trend consistency weight corresponding to the downstream behavior event and the target behavior trend setting weight.

[0163] First, the system calculates the weight difference value between the behavior trend consistency weight corresponding to each downstream behavior event and the target behavior trend setting weight. This difference value reflects the deviation between the actual behavior trend and the expected behavior trend, and is an important indicator for evaluating risks.

[0164] Next, using this weight difference value, the system determines the risk deviation index of each downstream behavior event at the target stability jump level through a preset discriminant model. This risk deviation index is a quantitative value used to specifically represent the deviation degree between the downstream behavior event and the target stable state. The higher the deviation index, the greater the difference between the behavior event and the stable state, and the higher the possible risk.

[0165] Then, the system compares the risk deviation index of each downstream behavior event with a preset risk threshold one by one. This risk threshold is set by the system based on historical data and risk tolerance, and is used as a standard for judging whether the behavior event is normal.

[0166] If the risk deviation index of a downstream behavior event exceeds the risk threshold, then the system determines that this behavior event is abnormal at the target stability jump level. For such an abnormal situation, the system generates corresponding abnormal risk control optimization suggestions. These suggestions may include strengthening the monitoring frequency of this downstream behavior event to detect and handle potential risks more promptly; adjusting the risk assessment model parameters of this downstream behavior event to make it more accurately reflect the actual risk situation; or performing additional risk verification measures on this downstream behavior event to ensure its security.

[0167] On the contrary, if the risk deviation index of a downstream behavior event does not exceed the risk threshold, then the system determines that this behavior event is normal at the target stability jump level and there is no need to output abnormal risk control optimization suggestions.

[0168] Finally, to improve the efficiency and effectiveness of risk control optimization, the system sorts the generated abnormal risk control optimization suggestions according to the magnitude of the risk deviation index. In this way, downstream behavior events with higher risk deviation indices will be processed preferentially, thus ensuring that the system can concentrate resources on solving the most important risk problems.

[0169] Through this entire set of processes, the risk control optimization system can accurately identify abnormal downstream behavior events and give targeted optimization suggestions. This not only helps to enhance the system's risk prevention and control capabilities but also provides a safer and more stable operating environment for financial institutions. At the same time, by preferentially processing high-risk events, the system can use resources more effectively and improve the efficiency and effectiveness of risk control work.

[0170] Please refer to Figure 2 For the embodiments of the present application, a risk control optimization system 100 is further provided, including a processor 111, as well as a memory 112 and a bus 113 connected to the processor 111. Among them, the processor 111 and the memory 112 communicate with each other through the bus 113. The processor 111 is used to call program instructions in the memory 112 to execute the above-mentioned risk control optimization method based on user stability.

[0171] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, commodity or system including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, commodity or system. Without further limitations, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, commodity or system including the element.

[0172] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] The above are only the embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A risk control optimization method based on user stability, characterized in that: Applied to the risk control optimization system, the method includes: Acquire Internet financial service session data corresponding to the target platform user; the Internet financial service session data includes to-be-processed business interaction behavior information obtained by monitoring the target platform user based on the authorized risk control thread; Performing behavior intention mining on the to-be-processed business interaction behavior information in the Internet financial business session data to obtain a behavior intention mining result; the behavior intention mining result represents a session scenario label corresponding to at least one downstream behavior event in the Internet financial business session data and a behavior state category of each downstream behavior event; Extracting the behavior trend vector at the target stability jump level for each downstream behavior event according to the behavior intention mining result, and obtaining the target behavior trend vector at the target stability jump level for each downstream behavior event; For each of the downstream behavior events, obtain the template behavior trend vector of the template behavior event at the target stability jump level, and determine the behavior trend consistency weight between the target behavior trend vector of the downstream behavior event and the template behavior trend vector; the template behavior event is a behavior event that matches the downstream behavior event in the template business interaction behavior information corresponding to the target platform user; Output the abnormal risk control optimization suggestions for the downstream behavior event at the target stability jump level based on the judgment viewpoints of the behavior trend consistency weight corresponding to the downstream behavior event and the target behavior trend setting weight; the target behavior trend setting weight is the behavior trend setting weight corresponding to the target stability jump level; The target stability jump level includes a behavior event thermal index, a behavior event distribution index, and a behavior event safety index; the behavior trend vector of the target stability jump level is extracted for each downstream behavior event based on the behavior intention mining result, and the target behavior trend vector of each downstream behavior event at the target stability jump level is obtained, including: For each of the downstream behavior events, extracting positioning features of each conversation semantic block in the conversation scene label corresponding to the downstream behavior event; According to the positioning features of each session semantic block, the behavior event heat vector and the behavior event distribution vector of the downstream behavior event are determined; wherein the behavior event heat vector reflects the frequency of user activities in a specific area, including the click heat of a functional area; the behavior event distribution vector reveals the spatial distribution of user behaviors, including the order and frequency of user visits to different areas of the page; Based on the behavior state category of the downstream behavior event, the behavior event development trajectory of the downstream behavior event is determined, and based on the behavior event development trajectory and the positioning features of each conversation semantic block, the behavior event security features of the downstream behavior event are determined.

2. The method according to claim 1, characterized in that Determining the behavior event heat vector and the behavior event distribution vector of the downstream behavior event based on the positioning features of each conversation semantic block includes: Filtering the first positioning element, the second positioning element, the first query mapping element and the second query mapping element from the positioning features of each session semantic block; Determining a behavior event heat vector of the downstream behavior event according to a comparison result between the first positioning element and the second positioning element, and a comparison result between the first query mapping element and the second query mapping element; According to the positioning features of each session semantic block, a global positioning element and a global query mapping element are determined to obtain a behavior event distribution vector of the downstream behavior event.

3. The method according to claim 1, characterized in that Determining the behavior event security features of the downstream behavior event based on the behavior event development trajectory and the positioning features of each conversation semantic block includes: According to the development trajectory of the behavior event, determining the behavior activation attribute positioning feature and the behavior interception attribute positioning feature from the positioning features of each conversation semantic block; Determine the event correlation of the downstream behavior event based on the key-value pair features in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature, and determine the behavior risk coefficient of the downstream behavior event based on the behavior risk features in the behavior activation attribute positioning feature and the behavior interception attribute positioning feature; A weighted result of the event correlation and the behavior risk coefficient is used as a safety discrimination variable, and a feature decoding result corresponding to the safety discrimination variable is determined as a behavior event safety feature of the downstream behavior event.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: Extracting the attention vector of the to-be-processed business interaction behavior information at the user community attention level according to the behavior intention mining result, and obtaining the user community attention vector of the to-be-processed business interaction behavior information at the user community attention level; Obtaining a template attention vector of the template business interaction behavior information at the user community attention level; Based on the attention vector comparison result between the user community attention vector and the template attention vector, the community risk control transmission opinion of the to-be-processed business interaction behavior information at the user community attention level is output.

5. The method according to claim 4, characterized in that The user community attention level includes a transaction warning heat index and a transaction warning distribution index; the attention vector of the user community attention level is extracted from the to-be-processed business interaction behavior information according to the behavior intention mining result, and the user community attention vector of the to-be-processed business interaction behavior information at the user community attention level is obtained, which includes: Determine the global scenario risk control confidence of the conversation scenario label corresponding to each downstream behavior event in the at least one downstream behavior event, and obtain the platform user risk control confidence of the to-be-processed business interaction behavior information; the platform user risk control confidence is used as the user community attention vector of the transaction warning heat index; Determine the target key semantic block of the to-be-processed business interaction behavior information according to the positioning features of the conversation semantic block in the conversation scenario tag corresponding to each downstream behavior event in the at least one downstream behavior event; the target key semantic block is used as the user community attention vector of the transaction warning distribution indicator; The method of outputting the community risk control transmission viewpoint of the pending business interaction behavior information at the user community attention level based on the attention vector comparison result between the user community attention vector and the template attention vector includes: determining the feature difference between the target key semantic block and the template key semantic block to obtain the feature difference weight, and determining the first community risk control transmission viewpoint of the pending business interaction behavior information in the transaction warning distribution index based on the feature difference weight; determining the confidence comparison result between the platform user risk control confidence and the template transaction warning confidence to obtain the confidence difference coefficient; when the confidence difference coefficient is greater than the confidence difference threshold, determining the second community risk control transmission viewpoint of the pending business interaction behavior information in the transaction warning thermal index; and outputting the first community risk control transmission viewpoint and the second community risk control transmission viewpoint as the community risk control transmission viewpoints of the pending business interaction behavior information; The user community attention level also includes a transaction warning level level; the method also includes: when the confidence difference coefficient is less than or equal to the confidence difference threshold, determining the abnormal transaction warning feature of the pending business interaction behavior information based on the positioning features of the conversation semantic block in the conversation scenario label corresponding to each downstream behavior event in the at least one downstream behavior event; the abnormal transaction warning feature is used as the user community attention vector at the transaction warning level level; determining the level comparison result between the abnormal transaction warning feature and the template transaction warning feature to obtain a level error, and determining the third community risk control conduction opinion of the pending business interaction behavior information at the transaction warning level based on the level error; and outputting the first community risk control conduction opinion and the third community risk control conduction opinion as the community risk control conduction opinion of the pending business interaction behavior information.

6. The method according to claim 4, characterized in that The method further comprises: For each of the downstream behavior events at the target stability jump level, the behavior trend consistency weight is used to determine the risk control stability metric value of the downstream behavior event at the target stability jump level according to the weighted result between the behavior trend consistency weight and the template behavior trend vector of the corresponding template behavior event at the target stability jump level; Performing averaging processing on the risk control stability metric value of each downstream behavior event in the at least one downstream behavior event at the target stability jump level to obtain the risk control stability metric value of the to-be-processed business interaction behavior information; Determine a first transaction warning metric value of the to-be-processed business interaction behavior information at the user community attention level according to a weighted result between an attention vector comparison result of the to-be-processed business interaction behavior information at the user community attention level and the template attention vector; Determine a feature commonality value between the Internet financial service session data and the template business interaction behavior mapping feature corresponding to the template business interaction behavior information, and obtain a second transaction warning metric value of the to-be-processed business interaction behavior information; Outputting the transaction warning metric value of the business interaction behavior information to be processed according to the first transaction warning metric value, the second transaction warning metric value and the risk control stability metric value of the business interaction behavior information to be processed.

7. The method according to claim 1, characterized in that The obtaining of Internet financial service session data corresponding to the target platform user includes: In response to a data sampling request for the current risk control optimization system, a security authorization monitoring is performed on the current risk control optimization system to obtain a web spider crawling result; the web spider crawling result includes at least one page session collected in the current risk control optimization system based on the authorized risk control thread, and each of the collected page sessions is in a page session flow of the current risk control optimization system; Decomposing the web spider crawling result according to the page session flow to obtain a plurality of page session data; Filtering the page session data containing the transaction warning tag from the plurality of page session data to obtain at least one target page session data; Any target page session data among the at least one target page session data is used as the Internet financial service session data.

8. A risk control optimization system, characterized in that: It includes a processor, a memory and a bus connected to the processor; the processor and the memory communicate with each other through the bus; the processor is used to call the computer program in the memory to execute the risk control optimization method based on user stability as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the risk control optimization method based on user stability described in any one of claims 1-7 is implemented.

Images