Risk early warning method and device based on multi-channel dynamic routing, equipment and medium

Abstract

The invention relates to the field of signal transmission, and particularly discloses a risk early warning method and device based on multi-channel dynamic routing, equipment and a medium, and the method comprises the steps: obtaining risk data, and generating a risk early warning signal based on the risk data; obtaining channel information of multiple selectable push channels, and determining signal channel comprehensive scores of the multiple selectable push channels based on the channel information; the signal channel comprehensive score is determined by at least one of the availability estimated value, the load estimated value and the channel response speed of the selectable push channel; updating a pushing rule base based on the signal channel comprehensive score and the early warning signal type and the early warning level of the risk early warning signal; and pushing the risk early warning signal based on the pushing rule base. The selectable pushing channels are evaluated through multiple dimensions, it can be ensured that the risk early warning signals rapidly and accurately reach the customer manager, and the accuracy and efficiency of the whole intelligent risk early warning system are improved.

Description

Technical Field

[0001] This application relates to the field of signal transmission, specifically to a risk warning method, apparatus, device, and medium based on multi-channel dynamic routing. Background Technology

[0002] Risk warning notification methods in the field of risk management aim to deliver potential risk warning signals for various businesses (such as credit business) to account managers in real time and accurately, so as to help account managers complete risk identification, decision-making intervention and resource allocation tasks, and ultimately achieve the goal of ensuring the safety of bank assets and improving risk management efficiency.

[0003] However, in the rapidly changing financial market, traditional notification methods are unable to keep up with the rapid increase in the types and number of warning rules and signals when faced with warning signals that have high timeliness requirements. They also cannot meet the needs of credit management personnel such as account managers for timely review and handling of warning signals.

[0004] Specifically, traditional commercial bank risk warning notification methods are limited, typically relying on SMS or a single internal system. This results in lengthy approval times and delays when the channels are congested, failing to meet the high timeliness requirements of risk warning notifications. When warning signals rely on a single internal system, bank risk management personnel, such as account managers, must use a fixed internal terminal to process them. If the person in charge does not have a terminal readily available, they cannot review and process pending warnings in a timely manner, easily leading to delays in risk handling.

[0005] At the same time, the existing methods also lack a tracking mechanism for the early warning status. If only the account manager receives the early warning notification, the early warning processing process may be blocked due to busy work or negligence, and a closed loop of early warning signal sending, reaching, handling and feedback cannot be formed, which will ultimately lead to delays in risk handling. Summary of the Invention

[0006] To address the aforementioned problems, this application proposes a risk warning method, apparatus, device, and medium based on multi-channel dynamic routing, wherein the method includes: Obtain risk data and generate a risk warning signal based on the risk data; obtain channel information for multiple optional push channels and determine a comprehensive signal channel score for the multiple optional push channels based on the channel information; the comprehensive signal channel score is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels; update the push rule base based on the comprehensive signal channel score, the warning signal type, and the warning level of the risk warning signal; and push the risk warning signal based on the push rule base.

[0007] In one example, acquiring risk data and generating a risk warning signal based on the risk data specifically includes: removing duplicate risk data by comparing a first key field of the initial risk data to obtain first intermediate risk data; the first key field includes at least one of customer name, customer number, unified social credit code, risk category, and timestamp; normalizing the first intermediate risk data to obtain second intermediate risk data; the normalization process includes at least one of field alignment, data type conversion, and numerical normalization; detecting outliers in the second intermediate risk data and correcting the outliers to obtain third intermediate risk data; extracting a second key field from the third intermediate risk data to construct a structured risk warning signal; the second key field includes at least one of risk category, warning level, correlation, warning time, and warning details.

[0008] In one example, the step of pushing the risk warning signal based on the push rule base specifically includes: determining the warning signal type and warning level corresponding to the risk warning signal to be pushed; determining the target push rule corresponding to the risk warning signal to be pushed in the push rule base based on the warning signal type and the warning level; determining the target push channel corresponding to the target push rule, and using the target push channel to push the risk warning signal according to the warning level.

[0009] In one example, the optional push channels include at least one of internal communication system channels, email channels, SMS channels, and telephone channels; the step of using the target push channel to push the risk warning signal according to the warning level specifically includes: obtaining push response data of the target push channel, the push response data including at least one of user online status in the system, email response time, and SMS channel load status; in response to the push response data not meeting the channel usage conditions, switching the target push channel to other optional push channels according to a preset channel switching order; in response to all optional push channels not meeting the channel usage conditions, adding the current risk warning signal to a waiting queue; and pushing the current risk warning signal again after waiting for any optional push channel to meet the channel usage conditions; after pushing the risk warning signal through the target push channel, obtaining the account manager's handling time for the risk warning signal; in response to the handling time exceeding a preset time threshold, re-pushing the risk warning signal using other optional push channels.

[0010] In one example, determining the warning signal type and warning level corresponding to the risk warning signal to be pushed specifically includes: determining the dynamic score of the risk warning signal based on the real-time warning score corresponding to the current risk warning signal, the historical warning score corresponding to the historical risk warning signal, and the anomaly detection score of the pushed abnormal signal; determining the dynamic assessment volatility based on the current dynamic score and the dynamic score within the historical time period; and determining the warning level threshold corresponding to each warning level based on the dynamic assessment volatility and the business demand adjustment factor.

[0011] In one example, after pushing the risk warning signal based on the push rule base, the method further includes: recording the push status of each warning signal push process, the push status including at least one of push time, push channel, warning handling time, and push handling result; determining the warning signal handling success rate based on the push status, the warning signal handling success rate being the ratio of successful warning handling results to all warning handling results; determining a push effect score based on the warning signal handling success rate, warning signal delivery rate, and warning signal click rate; the warning signal delivery rate being the ratio of warning signals successfully pushed to the correct account manager device to all warning signals; the warning signal click rate being the ratio of account managers clicking on warning signals to all successfully pushed warning signals; and storing the push effect score in the push rule base.

[0012] In one example, obtaining channel information for multiple optional push channels and determining a comprehensive signal channel score for the multiple optional push channels based on the channel information specifically includes: determining the availability estimate of each optional push channel based on online rate and fault recovery rate; the online rate is the percentage of time a channel is online within a statistical period; the fault recovery rate is the ratio of the number of times a channel successfully recovers after a fault to the total number of faults; determining the load assessment value of each optional push channel based on signal queue length, channel queue capacity, number of unprocessed warning signals, maximum personnel load, and resource utilization rate; the signal queue length is the length of the queue of signals to be pushed in the optional push channel for the current warning signal; the channel queue capacity represents the maximum capacity of the current warning signals to be pushed in the signal queue; the number of unprocessed warning signals is the number of current type warning signals that have not been processed after being successfully pushed through the signal queue; The maximum processing load per person represents the maximum load a single risk management personnel can handle for the current type of early warning signal; resource utilization rate refers to the ratio of system resources occupied by the channel to the total resources; the response speed of each optional push channel is determined based on channel response time, average response time, signal processing time, and average processing time of similar signals; channel response time is the time from receiving the current early warning signal to completing the push from the current push channel; average response time is the average time from receiving all early warning signals to completing the push from the current push channel; signal processing time is the time from completing the push of the current early warning signal to completing processing from the current push channel; average processing time of similar signals is the average time from completing the push of early warning signals of the same type as the current signal to completing processing from the current push channel; the comprehensive score of the signal channel is determined based on at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels.

[0013] This application also provides a risk warning device based on multi-channel dynamic routing, comprising: a risk warning signal generation module for acquiring risk data and generating a risk warning signal based on the risk data; an optional push channel scoring module for acquiring channel information of multiple optional push channels and determining a comprehensive signal channel score for the multiple optional push channels based on the channel information; the comprehensive signal channel score is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels; a push rule base update module for updating the push rule base based on the comprehensive signal channel score, the warning signal type, and the warning level of the risk warning signal; and a risk warning signal push module for pushing the risk warning signal based on the push rule base.

[0014] This application also provides a risk warning device based on multi-channel dynamic routing, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the method described in any of the above examples.

[0015] This application also provides a non-volatile computer storage medium storing computer-executable instructions configured to perform the steps of the method described in any of the above examples.

[0016] The method proposed in this application can bring the following beneficial effects: 1. It solves the problems of untimely warnings and lack of tracking mechanisms for warning signals in traditional risk warning notification methods.

[0017] 2. Based on real-time channel scoring and dynamic routing rules, the system matches the optimal channel for different types of early warning signals. It also collects key indicators such as the delivery rate, handling rate, and handling time of early warning signals through real-time stream processing technology. Based on these indicators, the system calculates the push effect score and uses the push effect score to optimize the dynamic routing rules, thereby improving system efficiency and the accuracy of notifications.

[0018] 3. The entire process of risk warning signal generation, processing, push notification, and feedback is tracked, and dynamic routing optimization is performed using push notification effectiveness scoring, improving the accuracy and resource utilization of warning signal push notifications. Furthermore, the selection and priority of notification channels for different types of warning signals can be dynamically adjusted based on push notification effectiveness scores, providing a new solution for risk prevention and control. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating a risk warning method based on multi-channel dynamic routing in an embodiment of this application; Figure 2 This is a schematic diagram of a risk warning system based on multi-channel dynamic routing in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a risk warning device based on multi-channel dynamic routing in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a risk warning device based on multi-channel dynamic routing in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This document presents a flowchart illustrating a risk warning method based on multi-channel dynamic routing, provided for one or more embodiments. This method can be applied to various business areas, such as bank risk alerts, securities trading warnings, and insurance claim warnings. The following explanation uses a bank risk alert scenario as an example. Some input parameters or intermediate results (such as various preset thresholds) in the process can be manually adjusted to help improve accuracy.

[0023] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0024] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations in this regard.

[0025] like Figure 1 As shown, this application provides a risk warning method based on multi-channel dynamic routing. The following description uses bank risk warning information push as an example, including: S101: Obtain risk data and generate a risk warning signal based on the risk data.

[0026] The server periodically (e.g., at midnight each day) connects to data sources from multiple platforms in batches to initiate a data aggregation process. These platforms and their corresponding data sources can include: a comprehensive risk monitoring and early warning platform, containing batch and real-time risk information for corporate clients across pre-loan, mid-loan, and post-loan dimensions; a central risk measurement engine, which provides batch and real-time risk assessment and measurement information for corporate clients based on quantitative models; and an external data management system, containing batch and real-time risk information for corporate clients across external data dimensions such as corporate credit reporting, business registration, administrative penalties, legal proceedings, and negative public opinion. Aggregating risk information from these data sources provides a comprehensive risk perspective for corporate clients, ensuring the capture of potential risks in different scenarios and activating them as risk warning signals. After acquiring the risk data, risk warning signals can be generated based on it.

[0027] In one embodiment, when generating a risk warning signal, the risk signal must first be preprocessed. This preprocessing includes deduplication, field alignment, data type conversion, and numerical normalization. Specifically, during deduplication, the collected risk data needs to be identified and deduplicated to ensure the uniqueness and accuracy of the data. The deduplication algorithm compares key fields of the risk data—customer name, customer number, unified social credit code, risk category, timestamp, etc.—and eliminates completely duplicate signals. The deduplication formula is as follows:

[0028] in, This represents a warning signal. After deduplicating the initial risk data, the first intermediate risk data is obtained. When aligning fields, warning signals from different data sources need to be mapped to fields predefined in this module, such as mapping "Customer Number" to "Unified Social Credit Code".

[0029] When performing data type conversion, it is necessary to unify the format of specific fields. For example, convert all date fields to a unified date format "yyyymmdd", such as "20250220".

[0030] When performing numerical normalization, non-monetary numerical fields need to be normalized to ensure that the data within the field are within the same unit of measurement, which facilitates subsequent analysis and processing. The normalization formula is as follows:

[0031] in, These are the data values ​​obtained after normalization. For the original data, min( ) represents the minimum value in the dataset, and max( The first intermediate risk data represents the maximum value in the dataset. After performing field alignment, data type conversion, and numerical normalization on the first intermediate risk data, the second intermediate risk data is obtained.

[0032] After obtaining the second intermediate data, outlier detection is required. This involves detecting and processing outliers in the second intermediate risk data, identifying and correcting or deleting obviously erroneous or unreasonable data. For example, unreasonable negative values ​​in the second intermediate data are corrected, and abnormally high and low values ​​are marked to ensure the rationality of the warning signal. The outlier detection formula is as follows:

[0033] in, Refers to outliers, The 25th percentile of the dataset. It is the 75th percentile. The interquartile range is denoted as . After outlier detection on the second intermediate risk data, the third intermediate risk data is obtained.

[0034] After obtaining the third intermediate risk data, risk warning signals can be generated based on it. Specifically, a second key field can be extracted from the third intermediate risk data to generate a structured list of warning signals to be notified. This second key field includes: risk category, warning level, correlation, warning time, and warning details. The risk category clarifies the risk category to which the warning signal belongs, such as credit risk, operational risk, or public opinion risk. Risk level: Based on the urgency of the risks associated with the warning signal, the warning signals are divided into three levels: red, orange, and blue. The risk level calculation formula is as follows:

[0035] The initial threshold for each type of risk warning is provided by the business, and the risk score is calculated based on a weighted sum of multiple risk indicators:

[0036] in, These are the weighting coefficients. Let i be the value of the i-th risk indicator. This represents the total number of risk indicator types. The relationship section uses the customer-employee relationship table to query the ownership of account management and business rights for corporate clients involved in the warning signal, ensuring that notifications accurately reach responsible risk management personnel. The warning time records the generation time of the warning signal, such as 20250220, facilitating subsequent time series analysis. The warning details briefly describe the specific content of the warning signal, such as "Red Warning, Abnormal Enterprise Credit Information: There is overdue loan from this bank, specifically: the overdue loan amount from this bank is {1,000,000} million yuan," allowing risk management personnel to quickly understand the risk situation. The structured list of warning signal categories and the customer-employee relationship table are stored in the system's database and displayed on the system's front-end page for business personnel to maintain and update, facilitating subsequent dynamic routing configuration and notification push.

[0037] In one embodiment, quality monitoring of data aggregation and preprocessing can be performed to detect the completeness of risky data and the accuracy of early warning signals in real time. By setting data quality indicators, such as thresholds for missing data rate, duplication rate, and outlier ratio, data quality problems can be detected promptly and alerts can be triggered. The formulas for calculating data quality indicators are as follows:

[0038]

[0039]

[0040] Furthermore, in one embodiment, detailed logs of data aggregation and preprocessing can be recorded, including data access time, data volume, normalization, and quality indicator calculation formulas. These log entries provide a basis for subsequent auditing and problem investigation, ensuring the traceability of risk data related to warning signals. Based on these steps, risk data can be collected and preprocessed efficiently and accurately, warning signals can be activated, and a structured, high-quality list of risk warning signals can be generated.

[0041] S102: Obtain channel information for multiple optional push channels, and determine the comprehensive signal channel score for the multiple optional push channels based on the channel information.

[0042] By evaluating various optional push channels, it is possible to select the channel that is more suitable for pushing the current risk warning signal from among the multiple optional push channels.

[0043] In one embodiment, when determining the overall score of signal channels across multiple optional push channels, the availability estimate of each optional push channel can be determined based on online rate and fault recovery rate. Specifically, the availability estimate of the channel is calculated by monitoring indicators such as the channel's online status, fault rate, and recovery time. The availability estimate formula is as follows:

[0044] in, and Here, are weighting coefficients, representing the relative importance of online rate and fault recovery rate in the evaluation, and satisfying the following conditions: Online rate refers to the percentage of time a channel is online within a statistical period; fault recovery rate refers to the ratio of the number of times a channel successfully recovers after a fault to the total number of faults.

[0045] Then, based on the signal queue length, channel queue capacity, number of unprocessed warning signals, maximum personnel load, and resource utilization, the load assessment value for each optional push channel can be determined. That is, the load and handling status of each level and category of warning signal, such as the {credit risk, red} warning signal, in each channel is quantitatively assessed, taking into account the current signal queue length, number of unprocessed signals, and system resource utilization. The load assessment value can be calculated using the following formula:

[0046] in, This is the load assessment value. , , These are the weighting coefficients. The signal queue length represents the length of the queue of signals to be pushed in this channel for the current warning signal. The channel queue capacity represents the maximum capacity of the current queue to be pushed risk warning signals. The number of unprocessed risk warning signals represents the number of current type warning signals pushed through this queue that have not been processed. The maximum number of personnel that can handle the current type of risk warning signals represents the maximum load that a single risk management personnel can handle. The resource utilization rate refers to the proportion of system resources occupied by the channel, such as CPU and memory, to the total resources.

[0047] The response speed of each selectable push channel can be determined based on channel response time, average response time, signal processing time, and the average processing time of similar signals. This involves quantitatively evaluating the response time of each channel and the processing speed of risk warning signals notified through that channel, and measuring the channel's response performance by calculating the average response time and timeout rate. The response speed evaluation value can be calculated using the following formula:

[0048] in, and The weighting coefficients represent the relative importance of channel response efficiency and risk warning signal processing efficiency in the evaluation. Channel response time refers to the time it takes for the current risk warning signal to be received and pushed from the channel. Average response time refers to the average time it takes for all risk warning signals to be received and pushed from the channel. Signal processing time refers to the time it takes for the current risk warning signal to be pushed from the channel and then processed. Average processing time for similar risk warning signals refers to the average time it takes for similar risk warning signals to be pushed from the channel and then processed from the channel.

[0049] After obtaining the availability estimate, load assessment value, and channel response speed, the comprehensive score of the signal channel can be determined based on at least one of the availability estimate, load assessment value, and channel response speed of the selectable push channel. That is, based on the above assessment indicators, the comprehensive score of different categories and levels of warning signals in each notification channel is calculated. The formula for calculating the comprehensive score estimate is as follows:

[0050] in, , , These are the weighting coefficients for availability estimation, load assessment, and response speed estimation, respectively, and they satisfy... The weighting coefficients are adjusted based on actual business needs and channel characteristics to ensure the rationality and effectiveness of the scoring.

[0051] S103: Update the push rule base based on the comprehensive score of the signal channel, the warning signal type and warning level of the risk warning signal.

[0052] In one embodiment, after receiving a risk warning signal, dynamic routing rules can be generated based on a comprehensive score of the warning signal's category, level, and communication channel. These dynamic routing rules are represented as {warning signal category, warning level, optimal channel}. Once obtained, the dynamic routing rules can be stored in a push rule base, enabling periodic updates to the push rule base.

[0053] In one embodiment, when updating the push rule base, a rule validity assessment is first performed, that is, the validity of existing rules is evaluated based on the channel's real-time performance data. If the overall score of a channel remains below a threshold, the rule is considered invalid.

[0054] In addition, dynamic routing rules can be regenerated and the rule base updated based on the latest channel performance data and early warning signal characteristics. The update formula is as follows:

[0055] Meanwhile, the system records the time, content, and reason for each rule update, which facilitates subsequent auditing and optimization.

[0056] S104: Based on the push rule base, push the risk warning signal.

[0057] When pushing risk warning signals, the signals can be pushed based on the push rules in the push rule base.

[0058] Specifically, firstly, the type and level of the risk warning signal to be pushed must be determined. Based on the warning signal type and level, a target push rule corresponding to the risk warning signal is determined from the push rule base. Then, a target push channel is determined for the target push rule, and the risk warning signal is pushed using the target push channel according to the warning level. For example, for a red credit risk warning signal, if the SMS channel has the highest overall score, the generated rule is {credit risk, red, SMS channel}. Thus, the optimal channel is dynamically selected based on dynamic routing rules.

[0059] In one embodiment, when pushing risk warning signals, the system allocates the warning signals to the corresponding notification channels according to dynamic routing rules. The allocation logic is: allocation task = {risk warning signal, optimal channel}. When a large number of warning signals need to be notified, such as when updating second-quarter financial reports in batches on the same day and exposing a large number of financial risks, this module will automatically trigger a channel switching mechanism if it detects during the task allocation process that the currently selected channel cannot meet the needs of the notification task (e.g., excessive load, excessive response delay).

[0060] Specifically, the first step is to acquire push response data from the target push channel. This data includes at least one of the following: user online status within the system, email response time, and SMS channel load. If the push response data does not meet the channel usage conditions, the target push channel is switched to other available push channels according to a preset channel switching order. If all available push channels fail to meet the usage conditions, the current risk warning signal is added to a waiting queue. The signal is then pushed again only after any available push channel meets the usage conditions. After the risk warning signal is pushed through the target push channel, the account manager's processing time for the risk warning signal is acquired. If the processing time exceeds a preset time threshold, the risk warning signal is re-pushed using other available push channels.

[0061] For example, the switching logic can be set as follows: When using the internal communication system channel for push notifications, if the desktop / mobile application of the internal communication system associated with the current warning signal is offline, switch to the email channel to push the warning signal to the account manager's email address; if the email channel response delay is too large (average response time exceeds a preset time threshold), switch to the SMS channel to push the warning signal to the account manager's mobile phone via SMS; if the SMS channel is overloaded (average response time exceeds a preset time threshold), switch back to the internal communication system channel; if none of the above three channels can send the warning signal, the current warning signal enters a waiting queue, waiting for any channel to meet the usage conditions before being pushed; if the current warning signal remains unresolved for a long time (account manager's handling time exceeds a preset time threshold) and the warning level is red, notify system administrators using the remaining two channels and contact the account manager associated with the current warning signal via telephone.

[0062] In one embodiment, load balancing management can be implemented for each notification channel to ensure that the load on each channel is within a reasonable range. The load balancing strategy includes dynamically adjusting the task allocation ratio, as well as task queuing and priority adjustment. Dynamically adjusting the task allocation ratio means adjusting the task allocation ratio based on the load of each channel. If the load on a certain channel is too high, the proportion of warning signal push tasks allocated to that channel is reduced; conversely, the proportion is increased. Task queuing and priority adjustment means that for channels with high load, warning signal push tasks are placed in a waiting queue, and the priority of tasks is adjusted according to the warning signal level. Higher warning levels are processed first, while lower warning levels can be appropriately delayed. The load balancing formula is as follows: By applying load balancing strategies, push delays or failures caused by excessive load on a single channel are effectively avoided, improving overall push efficiency and stability. Simultaneously, the system supports real-time monitoring and dynamic adjustment of push tasks. When an anomaly or performance degradation is detected in a channel, the system can automatically adjust task allocation, transferring tasks originally assigned to that channel to other normally functioning channels, ensuring the smooth completion of push tasks.

[0063] In one embodiment, channel usage can be analyzed periodically to optimize dynamic routing rules. Specifically, performance metric analysis can be performed, identifying poorly performing channels within the current time window by analyzing performance metrics such as failure rate, response time, and load balancing. Rule optimization can also be performed, adjusting dynamic routing rules based on performance analysis results. If the failure rate of a channel is too high within the current time window, its priority is reduced; if the response time of a channel is too long, the allocation ratio of warning signal push tasks is adjusted to reduce the amount of tasks assigned to that channel.

[0064] The optimized formula is: For example, if the system detects that the failure rate of the email channel is consistently higher than a threshold, it generates optimization rules to switch some email channel tasks to the SMS channel or internal communication system channel and updates the rule base.

[0065] Through the above steps, the system can dynamically select the optimal channel based on the level of the risk warning signal and the real-time status of the notification channel resources, ensuring the efficient and reliable transmission of notification tasks while optimizing the utilization efficiency of channel resources. This process not only improves the accuracy and timeliness of warning signal delivery but also enhances the system's flexibility and adaptability, effectively addressing complex risk prevention scenarios and dynamically changing network environments.

[0066] In one embodiment, the server can integrate Apache Flink as a real-time stream processing framework, leveraging its low latency, high throughput, and powerful state management capabilities to process the entire process of risk warning signals.

[0067] When integrating real-time stream processing technologies, Flink uses an event-driven model to process early warning signal data streams in real time, ensuring that early warning signals are processed the instant they are generated. Flink's distributed architecture supports horizontal scaling, enabling it to handle large-scale data streams, while its fault tolerance mechanism ensures the reliability of data processing. Flink's Complex Event Processing (CEP) function allows the system to define complex early warning signals and identify abnormal patterns in early warning signal sequences, providing robust technical support for end-to-end tracking of risk early warning signals.

[0068] When constructing the data pipeline, the generation, processing, push, and feedback of risk warning signals are linked together. Data from each stage is transmitted and processed through Flink's distributed data stream engine, ensuring the integrity and consistency of the warning signal data. The constructed data pipeline includes: warning signal generation, processing, push, and feedback. Flink's Web UI can display the processing status, task execution, and resource usage of the entire data process in real time. The visual interface supports real-time monitoring and multi-dimensional data analysis by risk system administrators, facilitating rapid problem identification.

[0069] Based on Flink's CEP (Continuous Error Detection) function, the system can detect anomalies in warning signals. When the monitored warning signal data does not conform to preset thresholds or rules, the system automatically reports and notifies system administrators to take appropriate manual action. The preset threshold range is set based on historical data and business requirements to determine whether the warning signal data is abnormal.

[0070] Furthermore, when anomalies in risk warning signals are detected, the signals can be dynamically evaluated based on real-time and historical warning data. Flink's CEP (Continuous Error Prediction) function is used to identify abnormal patterns in the warning signal sequence and calculate a dynamic score for the warning signal. The dynamic scoring formula for risk warning signals is as follows:

[0071] in, , , The weighting coefficients represent the importance ratios of real-time early warning signals, historical early warning signals, and anomaly detection in the dynamic scoring of early warning signals, and satisfy the following conditions: Real-time data scores are calculated based on the level and impact range of the current risk warning signal; historical data scores are calculated based on the processing results and processing time of past risk warning signals of the same category; and anomaly detection scores are calculated by identifying abnormal patterns in the sequence of warning events using Flink's CEP library.

[0072] Then, based on the current dynamic score and the historical dynamic scores over a given period, the dynamic assessment volatility can be determined. Dynamic score volatility refers to the rate of change of the dynamic score over a certain period, reflecting its trend. This indicator helps the system respond promptly to changes in warning signals, dynamically adjust warning thresholds, and ensure the timeliness and accuracy of warnings. The formula for calculating dynamic score volatility is as follows:

[0073] After obtaining the dynamic scoring volatility, the corresponding warning level threshold for each warning level can be determined based on the dynamic assessment volatility and business demand adjustment factors. The risk warning threshold is then dynamically adjusted according to the dynamic scoring of the warning signal and real-time business needs. The risk warning threshold adjustment formula is calculated as follows:

[0074] in, This is a weighting coefficient for the dynamic score volatility, used to adjust the impact of dynamic score changes on the early warning threshold; The weighting coefficients of the business demand adjustment factors are used to adjust the warning thresholds based on real-time business needs. Once the warning thresholds are obtained, the corresponding warning level thresholds for each warning level can be determined based on these thresholds.

[0075] In one embodiment, after pushing a risk warning signal, the push effect can be evaluated based on the push results. Specifically, the push status of each warning signal push process is recorded, including at least one of the push time, push channel, warning handling time, and push handling result. Based on the push status, the warning signal handling success rate is determined, which is the ratio of successful warning handling results to all warning handling results. Based on the warning signal handling success rate, warning signal delivery rate, and warning signal click rate, a push effect score is determined. The warning signal delivery rate is the ratio of warning signals successfully pushed to the correct account manager's device to all warning signals. The warning signal click rate is the ratio of account managers clicking on push notifications (internal message read, email read, SMS replied) to all successful push notifications. The push effect score is stored in the push rule base.

[0076] Specifically, the server can evaluate the effectiveness of each alert rule's push notifications using Apache Flink real-time stream processing technology. The evaluation metrics include alert delivery rate, click-through rate, response rate, and response time. The evaluation process is implemented through data collection and statistical analysis, aiming to optimize the push notification effect under each alert rule and ensure more accurate and rapid delivery to account managers. This module uses Flink's real-time stream processing capabilities to record the alert delivery rate and click-through rate under the same alert rule, and calculates the alert response success rate by obtaining the alert response time and results.

[0077] Furthermore, Flink's windowing mechanism and state management capabilities can be used to perform real-time statistical analysis on the collected data. Data is grouped using time windows (each window lasting 30 minutes), and the average of four metrics—delivery rate, click-through rate, handling rate, and handling time—is calculated for each window. This average score determines the effectiveness of the push notification and is incorporated into the push rule base. The specific calculation can be performed using the following formula:

[0078] in, Let be the weighting coefficients, representing the importance ratio of the four items in the push effect score, and satisfying the following conditions: .

[0079] In one embodiment, the server can collect feedback from account managers at preset intervals (e.g., monthly) through various methods, including SMS replies, email feedback, and questionnaire completion. Feedback information = {Account Manager ID, Feedback Content, Feedback Time}, where the feedback content includes the accuracy of specific warning signals and the effectiveness of the warning signal notification method. By organizing and analyzing the collected feedback information, the system can further understand account managers' satisfaction with the risk warning signal push service and their actual needs. The system can use natural language processing technology to perform semantic analysis on the feedback content, extracting key information such as account managers' evaluation of warning accuracy and preferences for notification methods. Based on these analysis results, the system can identify problems and shortcomings in the current push service, such as low accuracy of certain types of warning signals or poor performance of certain notification methods in specific scenarios. To address these issues, the system can automatically generate optimization suggestions, such as adjusting the warning signal generation rules to improve accuracy or optimizing the notification channel selection strategy to enhance notification effectiveness.

[0080] like Figure 2 As shown, the risk warning method based on multi-channel dynamic routing provided in this application embodiment can... Figure 2 The system is implemented using the four modules shown. Module 1 is used for risk information collection and risk warning signal generation; Module 2 is used for channel resource assessment, dynamic routing rule generation, channel task allocation, and channel load balancing calculation; Module 3 is used for the full-process tracking mechanism, including dynamic assessment of risk signals and adjustment of warning thresholds for each warning level; and Module 4 is used for warning signal push and evaluation and optimization of push effect.

[0081] The following example illustrates the risk warning signal push process in the above embodiments. The warning system uses the risk information collection and risk warning signal generation module to detect changes in financial indicators on the company's financial statements and generate a financial warning signal: the company's main business revenue has decreased year-on-year. Then, the multi-channel dynamic routing configuration module is used to push the risk warning signal notification. Finally, the warning threshold and push effect are optimized based on the risk warning signal full-process tracking and warning threshold adjustment module and the risk warning signal notification push and feedback optimization module.

[0082] Step 1: Risk Information Collection and Early Warning Signal Generation Module: The server accesses the enterprise's financial data and financial statement data from the comprehensive risk monitoring and early warning platform, and uses COR technology to identify and extract the main business revenue from multiple periods of the enterprise's financial statements and convert it into structured data. If the enterprise's main business revenue (main business revenue in the same period of the previous year - main business revenue in the current period) / main business revenue in the same period of the previous year is ≥10% for two consecutive periods, and the main business revenue in the two consecutive periods is not 0, then the "Financial Early Warning Signal: Main Business Revenue Decreases Year-on-Year" is activated.

[0083] By comparing data based on the enterprise's unified social credit code, risk category, and timestamp, the financial early warning signals are deduplicated. Then, through field alignment, data type conversion, and numerical normalization, the financial early warning signals are standardized into a standard format.

[0084] Extract the risk category, company name, warning time, and warning details from the standard format financial warning signal, and match the associated account manager. Then, calculate the score for "Financial Warning Signal: Year-on-Year Decrease in Main Business Revenue" as 0.2. Based on the initial thresholds set by risk control professionals (0.3 for orange warnings and 0.7 for red warnings), the current risk level is determined to be blue. Finally, the key fields for this warning are: Financial Warning Signal: Year-on-Year Decrease in Main Business Revenue = {"Risk Name": "Financial Risk: Year-on-Year Decrease in Enterprise Main Business Revenue", "Risk Category": "Financial", "Company Name": "Jinghai ABCD Group", "Risk Level": "Blue", "Warning Time": "March 4, 2025", "Account Manager Number": "9539501", "Warning Details":} "The latest financial report shows that the main business revenue is RMB 10,000,000, compared to RMB 12,000,000 in the same period last year." The main business revenue in the previous period was RMB 10,500,000, and the revenue in the same period last year was RMB 12,500,000. Both the current year-on-year and the previous year-on-year revenue decreased by more than 10%, and the current main business revenue needs to be monitored in light of the company's operating conditions.

[0085] Step 2: Multi-channel dynamic routing configuration module: Calculates the real-time availability score of the channel based on the online status, failure rate and recovery time of SMS, email and internal communication systems. It also takes into account factors such as the current queue length, number of unprocessed tasks, system resource utilization and average response time to calculate the load assessment value and response speed score of the three channels respectively, and quantitatively evaluates the load and response speed of each channel.

[0086] Based on the channel real-time availability estimate, load assessment value, and response speed estimate calculated in Step 1, a comprehensive score for each channel is calculated, thereby generating dynamic routing rules: Financial Early Warning Routing Rule = {"Early Warning Signal Type": "Financial Risk", "Early Warning Level": "Blue", "Optimal Channel": "Internal Communication System"}. The Financial Early Warning Routing Rule is then stored in the system rule base.

[0087] According to dynamic routing rules, the "Financial Red Alert Signal: Year-on-Year Decline in Enterprise Main Business Revenue" is assigned to the internal communication system channel. During the push process, if the system detects that the current channel cannot meet the notification task's requirements, it automatically triggers a channel switching mechanism to assign the notification to another channel. Load balancing management is implemented for each notification channel, dynamically adjusting the task allocation ratio and task priority to ensure that the load on each channel is within a reasonable range.

[0088] Step 3, Module for full-process tracking of early warning signals and adjustment of early warning thresholds: Based on the real-time stream processing framework Apache Flink, a data link is constructed: The data link for "Financial early warning signal: Year-on-year decline in the company's main business revenue" = {early warning signal generation -> processing -> push -> feedback}, and the push progress and handling status of the early warning signal are displayed in real time through Flink's Web UI.

[0089] Based on the real-time and historical early warning signals of "financial early warning signal: year-on-year decline in the company's main business revenue", the dynamic score of the early warning signal is calculated, and the early warning threshold is adjusted according to the dynamic score and real-time business needs.

[0090] Step 4, Warning Signal Notification Push and Feedback Optimization Module: At 9:00 AM on weekdays, the "Financial Warning Signal: Year-on-Year Decrease in Enterprise Main Business Revenue" is pushed to the corresponding account manager through the internal communication system, and the push status of this warning is recorded: Push Status of Financial Warning Signal: Year-on-Year Decrease in Enterprise Main Business Revenue = {"Push Time": "March 4, 2025, 9:15", "Notification Channel": "Internal Communication System", "Handling Time": "March 4, 2025, 9:35", "Handling Result": "Success"}. The push effect of this warning is evaluated using Flink real-time stream processing technology = {"Delivery Rate": "98.5%", "Click-Through Rate": "95.3%", "Handling Rate": "92%"}, and a push effect score is calculated. Monthly feedback from account managers on early warning handling is collected. The feedback content is: "It is suggested that the threshold for financial early warning amounts in the E48 construction industry be increased", "Feedback date": "February 25, 2025".

[0091] In summary, this invention relates to the field of intelligent risk early warning in commercial banks, specifically proposing a risk early warning notification method for commercial banks based on multi-channel dynamic routing and real-time stream processing technology. This method includes the following steps: risk information collection and risk early warning signal generation, multi-channel dynamic routing configuration, full-process tracking of risk early warning signals and adjustment of early warning thresholds, and notification push and feedback optimization of risk early warning signals. This invention enables rapid push and full-process tracking of real-time early warning signals, and can also perform dynamic routing optimization based on push effect scores, ensuring that risk early warning signals can quickly and accurately reach account managers, thereby improving the accuracy and efficiency of the overall intelligent risk early warning system.

[0092] like Figure 3 As shown in the figure, this application embodiment also provides a risk warning device based on multi-channel dynamic routing, including: The risk warning signal generation module 301 acquires risk data and generates a risk warning signal based on the risk data.

[0093] The optional push channel scoring module 302 acquires channel information of multiple optional push channels and determines a comprehensive signal channel score for the multiple optional push channels based on the channel information; the comprehensive signal channel score is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels.

[0094] The push rule base update module 303 updates the push rule base based on the comprehensive score of the signal channel, the warning signal type and warning level of the risk warning signal.

[0095] The risk warning signal push module 304 pushes the risk warning signal based on the push rule base.

[0096] like Figure 4 As shown in the illustration, this application embodiment also provides a risk warning device based on multi-channel dynamic routing, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: acquire risk data and generate a risk warning signal based on the risk data; acquire channel information of multiple optional push channels and determine a comprehensive signal channel score for the multiple optional push channels based on the channel information; the comprehensive signal channel score is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels; update the push rule base based on the comprehensive signal channel score, the warning signal type, and the warning level of the risk warning signal; and push the risk warning signal based on the push rule base.

[0097] This application embodiment also provides a non-volatile computer storage medium storing computer-executable instructions, the computer-executable instructions being configured to: acquire risk data and generate a risk warning signal based on the risk data; acquire channel information of multiple optional push channels and determine a comprehensive signal channel score for the multiple optional push channels based on the channel information; the comprehensive signal channel score is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channels; update the push rule base based on the comprehensive signal channel score, the warning signal type, and the warning level of the risk warning signal; and push the risk warning signal based on the push rule base.

[0098] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0099] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0100] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0102] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0103] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0104] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0105] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0106] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0107] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0108] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A risk warning method based on multi-channel dynamic routing, characterized in that, include: Acquire risk data and generate risk warning signals based on the risk data; Obtain channel information for multiple optional push channels, and determine the comprehensive signal channel score for the multiple optional push channels based on the channel information; The overall score of the signal channel is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channel; The push rule base is updated based on the comprehensive score of the signal channel, the warning signal type and warning level of the risk warning signal; The risk warning signal is pushed based on the push rule base.

2. The method according to claim 1, characterized in that, The acquisition of risk data and the generation of risk warning signals based on the risk data specifically include: By comparing the first key field of the initial risk data, duplicate risk data is removed to obtain the first intermediate risk data; the first key field includes at least one of customer name, customer number, unified social credit code, risk category and timestamp. The first intermediate risk data is normalized to obtain the second intermediate risk data; the normalization process includes at least one of field alignment, data type conversion, and numerical normalization. Outlier detection is performed on the second intermediate risk data, and outliers are corrected to obtain the third intermediate risk data; A second key field is extracted from the third intermediate risk data to construct a structured risk warning signal; the second key field includes at least one of risk category, warning level, correlation, warning time, and warning details.

3. The method according to claim 1, characterized in that, The step of pushing the risk warning signal based on the push rule base specifically includes: Determine the type and level of the risk warning signal to be pushed; Based on the warning signal type and the warning level, the target push rule corresponding to the risk warning signal to be pushed is determined in the push rule base; Determine the target push channel corresponding to the target push rule, and use the target push channel to push the risk warning signal according to the warning level.

4. The method according to claim 3, characterized in that, The optional push channel includes at least one of an internal communication system channel, an email channel, an SMS channel, and a telephone channel; the step of using the target push channel to push the risk warning signal according to the warning level specifically includes: Obtain push response data from the target push channel, wherein the push response data includes at least one of the following: user online status within the system, email response time, and SMS channel load. If the push response data does not meet the channel usage conditions, the target push channel is switched to other available push channels according to the preset channel switching order; In response to the fact that none of the available push channels meet the channel usage conditions, the current risk warning signal is added to the waiting queue; and the current risk warning signal is pushed only after any available push channel meets the channel usage conditions. After the risk warning signal is pushed through the target push channel, the processing time of the account manager for the risk warning signal is obtained; If the processing time exceeds a preset time threshold, the risk warning signal is re-pushed using other optional push channels.

5. The method according to claim 3, characterized in that, The determination of the warning signal type and warning level corresponding to the risk warning signal to be pushed specifically includes: The dynamic score of the risk warning signal is determined based on the real-time warning score corresponding to the current risk warning signal, the historical warning score corresponding to the historical risk warning signal, and the anomaly detection score of the pushed abnormal signal. Based on the current dynamic score and the dynamic score over a historical period, determine the dynamic assessment volatility; Based on the dynamic assessment volatility and business demand adjustment factors, the warning level thresholds corresponding to each warning level are determined.

6. The method according to claim 1, characterized in that, After pushing the risk warning signal based on the push rule base, the method further includes: Record the push status of each early warning signal push process, wherein the push status includes at least one of the following: push time, push channel, early warning handling time, and push handling result; Based on the push status, the success rate of handling the early warning signal is determined, wherein the success rate of handling the early warning signal is the ratio of the successful handling result of the early warning to the total number of early warning handling results; The push effect score is determined based on the success rate of the warning signal processing, the warning signal delivery rate, and the warning signal click rate; the warning signal delivery rate is the ratio of warning signals that are successfully pushed to the correct account manager device out of all warning signals. The click-through rate of the warning signal is the percentage of warning signals clicked by account managers out of all successfully pushed warning signals; The push performance score is stored in the push rule base.

7. The method according to claim 1, characterized in that, The process of acquiring channel information for multiple selectable push channels and determining a comprehensive signal channel score for the multiple selectable push channels based on the channel information specifically includes: Based on online rate and fault recovery rate, determine the availability estimate of each optional push channel; The online rate is the percentage of time the channel is online within the statistical period; the fault recovery rate is the ratio of the number of times the channel successfully recovers after a fault to the total number of faults. Based on the signal queue length, channel queue capacity, number of unprocessed warning signals, maximum number of personnel, and resource utilization rate, determine the load assessment value of each optional push channel; The signal queue length is the length of the queue of signals to be pushed in the selectable push channel for the current early warning signal; the channel queue capacity represents the maximum capacity of the current early warning signals to be pushed in the signal queue; the number of unprocessed early warning signals is the number of current type early warning signals that have not been processed after being successfully pushed through the signal queue; the maximum personnel processing load represents the maximum load of a single risk management personnel on the current type of early warning signals; the resource utilization rate refers to the ratio of system resources occupied by the channel to the total resources; The response speed of each selectable push channel is determined based on channel response time, average response time, signal processing time, and average processing time of similar signals. Channel response time is the time it takes for the current warning signal to be received from the current push channel and for the push to be completed; average response time is the average time it takes for all warning signals to be received from the current push channel and for the push to be completed; signal processing time is the time it takes for the current warning signal to be pushed from the current push channel and for the processing to be completed; average processing time for similar signals is the average time it takes for warning signals of the same type as the current signal to be pushed from the current push channel and for the processing to be completed. The overall score of the signal channel is determined based on at least one of the availability estimate, load assessment value, and channel response speed of the optional push channel.

8. A risk warning device based on multi-channel dynamic routing, characterized in that, include: A risk warning signal generation module acquires risk data and generates a risk warning signal based on the risk data; The optional push channel scoring module acquires channel information for multiple optional push channels and determines a comprehensive signal channel score for the multiple optional push channels based on the channel information. The overall score of the signal channel is determined by at least one of the availability estimate, load assessment value, and channel response speed of the optional push channel; The push rule base update module updates the push rule base based on the comprehensive score of the signal channel, the warning signal type and warning level of the risk warning signal; The risk warning signal push module pushes the risk warning signal based on the push rule base.

9. A risk warning device based on multi-channel dynamic routing, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the method as claimed in any one of claims 1-7.

10. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are configured to perform the steps of the method as described in any one of claims 1-7.

Images