Risk prediction information generation methods, devices, equipment and media

Abstract

This application relates to the field of data analysis technology and discloses a method, apparatus, device, and medium for generating risk prediction information. The method includes constructing a multi-level evaluation system; acquiring current state data; calculating the current risk score based on a linear regression engine and a rule-weighted engine; and predicting target risk information based on the current risk score and a risk score threshold. Through this approach, this application achieves dynamic alignment between evaluation dimensions and target business by constructing a multi-level evaluation system. The dual-engine fusion calculation, combining the data objectivity of the linear regression engine with the business logic of the rule-weighted engine, ensures that the risk score possesses both mathematical rigor and business interpretability. The generation of risk prediction information based on real-time comparison of scores and thresholds enables the generation of risk prediction information from static assessment to dynamic early warning, improving the efficiency of risk management systems in processing multi-source heterogeneous business data in the insurance business field.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular to a method, apparatus, device and medium for generating risk prediction information. Background Technology

[0002] With the digital transformation of the insurance industry, cost management and risk control, as important components of property insurance, are increasingly becoming key to the core competitiveness of insurance companies. Accurate assessment of insurance product costs not only affects the profitability of insurance companies but also directly impacts the reasonable pricing of insurance premiums and compliance with market regulations. Traditional insurance cost assessment methods mainly rely on static scorecard models, fixed rule engines, or single linear regression models. These methods involve statistical analysis of historical policy data, claims records, and expense allocation to generate cost scores and risk rankings for the business source systems.

[0003] Existing risk management systems often employ a single algorithm architecture for data processing, such as scoring models based on traditional linear regression or decision engines based on fixed rules. Linear regression models establish a mapping relationship between feature variables and scoring results through statistical analysis of historical business data. However, insurance business data is influenced by multiple non-linear factors, including policy adjustments, regional differences, user behavior, and market environment. A single algorithm struggles to adequately fit the complex data distribution, leading to insufficient feature extraction and significant discrepancies between the scoring results and real-time business data.

[0004] The weight configuration of risk management systems is typically implemented using hard-coding, resulting in deep coupling between evaluation dimensions and calculation logic. When business scenarios change or data characteristics drift, the technical team must modify the source code, recompile, and re-deploy, leading to poor system scalability and long response times. Furthermore, static weights cannot be dynamically adjusted based on real-time data distribution characteristics, causing data processing parameters to mismatch with the current business state and limiting the accuracy of scoring calculations.

[0005] Risk management systems typically only set a single evaluation dimension, lacking a nested, multi-level feature extraction mechanism. Data features are extracted and calculated at a single granularity, failing to fully explore the correlations and hierarchical characteristics between indicators at different levels. This results in insufficient data representation capabilities and poor discrimination and stability of the scoring model. Therefore, in the insurance business, improving the efficiency of risk management systems in fusing and processing multi-source heterogeneous business data has become a pressing technical problem to be solved. Summary of the Invention

[0006] This application provides a method, apparatus, device, and medium for generating risk prediction information to improve the efficiency of risk management systems in fusion processing of multi-source heterogeneous business data.

[0007] Firstly, this application provides a method for generating risk prediction information, the method comprising: A multi-level evaluation system is constructed based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. Obtain the current status data of the preset business objective through a message queue; The current risk score of the preset business objective is calculated based on the dynamic weight configuration rules, linear regression engine, rule weight engine built into the multi-level evaluation system and the current state data. Based on the current risk score and the preset risk score threshold, predict the target risk information of the preset business objective.

[0008] Secondly, this application also provides a risk prediction information generation device, the device comprising: A multi-level evaluation system construction module is used to construct a multi-level evaluation system based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. The current status data acquisition module is used to acquire the current status data of the preset business target through a message queue; The current risk score calculation module is used to calculate the current risk score of the preset business objective based on the dynamic weight configuration rules, linear regression engine, rule weight engine and the current state data built into the multi-level evaluation system. The target risk information prediction module is used to predict the target risk information of the preset business target based on the current risk score and the preset risk score threshold.

[0009] Thirdly, this application also provides a computer device, the computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement the risk prediction information generation method as described above.

[0010] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the risk prediction information generation method described above.

[0011] This application discloses a method, apparatus, device, and medium for generating risk prediction information. The method includes constructing a multi-level evaluation system based on preset business objectives, wherein the multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator; obtaining the current state data of the preset business objectives through a message queue; calculating the current risk score of the preset business objectives based on the dynamic weight configuration rules, linear regression engine, rule weight engine, and the current state data built into the multi-level evaluation system; and predicting the target risk information of the preset business objectives based on the current risk score and a preset risk score threshold. Through the above method, this application achieves dynamic alignment between evaluation dimensions and target business by constructing a multi-level evaluation system, using a message queue to obtain real-time data, and combining the data objectivity of the linear regression engine with the business logic of the rule weight engine for dual-engine fusion calculation, making the risk score both mathematically rigorous and business interpretable. The generation of risk prediction information based on real-time comparison of scores and thresholds enables risk prediction information generation to move from static assessment to dynamic early warning, improving the efficiency of risk management systems in processing multi-source heterogeneous business data in the insurance business field. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic flowchart illustrating a risk prediction information generation method provided in an embodiment of this application; Figure 2 A schematic block diagram of a risk prediction information generation device provided for embodiments of this application; Figure 3 A schematic block diagram of the structure of a computer device provided for an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0016] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0017] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0018] This application provides a method, apparatus, device, and medium for generating risk prediction information. The method can be applied to a risk management system. By constructing a multi-level evaluation system, it achieves dynamic alignment between evaluation dimensions and target business. It utilizes message queues to acquire real-time data and combines the data objectivity of a linear regression engine with the business logic of a rule-weighted engine for dual-engine fusion calculation, ensuring that risk scores possess both mathematical rigor and business interpretability. Based on real-time comparison of scores and thresholds, risk prediction information is generated, enabling a shift from static assessment to dynamic early warning. In the insurance business field, this improves the efficiency of risk management systems in fusing and processing multi-source heterogeneous business data.

[0019] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0020] Please see Figure 1 , Figure 1 This is a schematic flowchart illustrating a risk prediction information generation method provided in an embodiment of this application. This risk prediction information generation method can be applied in a risk management system to improve the efficiency of the risk management system in fusing and processing multi-source heterogeneous business data.

[0021] like Figure 1 As shown, the risk prediction information generation method specifically includes steps S10 to S40.

[0022] Step S10: Construct a multi-level evaluation system based on preset business objectives, wherein the multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. Specifically, in the insurance business data processing scenario, the preset data processing objective is to fuse and score multi-dimensional data from the auto insurance business source system. Based on this objective, a multi-level, dynamically configurable data evaluation system is constructed to perform structured feature extraction and hierarchical weight calculation on the real-time business data stream received from the message queue.

[0023] In the primary dimension indicator configuration area, three primary dimension indicators are set: "Policy Data Dimension," "Structure Data Dimension," and "Deployment Data Dimension" (corresponding to policy volume data, channel structure data, and resource deployment data, respectively). Each primary dimension indicator supports global weight configuration and verifies whether the sum of all primary dimension weights is 100%. If it is not equal to 100%, the error item is highlighted.

[0024] Under the primary dimension of "Policy Data Dimension," add two secondary dimension indicators: "Resource Allocation Ratio" and "Cooperation Factory Data Ratio" (corresponding to the resource allocation quantity field and the cooperation factory repair volume / premium volume ratio field in the business system, respectively). The weights of the secondary dimension indicators adopt a parent-child linkage calculation rule. For example, if the allocation ratio of "Resource Allocation Ratio" within its primary dimension is 60%, and the global weight of that primary dimension is 40%, then the actual weight of "Resource Allocation Ratio" is 40% × 60% = 24%.

[0025] Level 3 dimensional metrics are calculable items specifically bound to business data fields. For example, the "Cooperating Factory Data Ratio" can be bound to the "Cooperating Factory Business Inflow" and "Cooperating Factory Business Outflow" fields in the database, with the calculation formula being Business Outflow / Business Inflow. Simultaneously, linear scoring interval rules are configured: at least two threshold nodes and corresponding score values ​​are set. For instance, 0 points are awarded when the ratio is ≥2%, 100 points are awarded when the ratio is ≤-1%, and points are awarded based on linear interpolation for values ​​between these two thresholds.

[0026] Step S20: Obtain the current status data of the preset business target through the message queue; Specifically, data acquisition agents are deployed in various business systems (such as the core policy system, financial system, and claims system) to capture changes in business data in real time. For example, when a new policy is issued, a claim is filed, or an expense is incurred, the acquisition agent encapsulates the raw data into a standard message format and pushes it to a designated topic in a message queue. The message queue acts as a data buffer, receiving massive amounts of real-time messages from various data sources.

[0027] Messages are continuously pulled from the message queue using streaming computing. Based on the message topic and timestamp, the data is automatically correlated and mapped to three-level dimensional metrics. To support time series trend analysis, a sliding window data can be configured. For example, with a window size of 30 days and a sliding step of 1 day, historical data caches from various business source systems can be updated in real time.

[0028] In one embodiment, the target object requiring data processing (such as a specific business node, agency channel, or partner institution) and its associated data fields are clearly defined to constitute the current state data. For example, monthly policy volume data for a business node, comprehensive ratio data A (such as the ratio of business expenses to policy volume), business amount data B (such as claims amount), and business ratio data of partner institutions, etc.

[0029] After generating new business data, each business source system encapsulates the data into a message according to a predefined JSON format. The message body includes metadata such as data identifier, timestamp, indicator code, indicator value, and data source, and sends it to the designated topic partition of the message queue in real time. The risk management system continuously monitors the message queue through consumer groups. Once a new business data message arrives, it immediately acquires the business data and parses, cleans, and standardizes its format, extracting the business source system identifier and various indicator data. Based on a predefined mapping table, it converts data from different sources and formats into current state data and writes it to the memory cache for subsequent reading by the computing engine.

[0030] In one embodiment, the objects corresponding to the "preset business objectives" that need to be monitored (such as a specific branch, agency channel, or repair partner) and their key business indicators are clearly defined to constitute "current status data," such as: the number of monthly policies, comprehensive expense ratio, compensation amount, and repair ratio of a certain branch.

[0031] After generating new business data, each business source system encapsulates the data into a message according to a predefined format and sends it to a message queue in real time. The risk management system continuously monitors the message queue, and as soon as a new business data message arrives, it immediately retrieves and parses the business source system identifier and various indicator data. Based on predefined rules or mapping tables, it converts data from different sources and formats into current state data. Step S30: Calculate the current risk score of the preset business objective based on the dynamic weight configuration rules, linear regression engine, rule weight engine, and current state data built into the multi-level evaluation system. Specifically, the built-in dynamic weight configuration rules are read from the configuration metadata of the multi-level evaluation system. These rules generate engine fusion parameters based on the deviation between the current state data's data distribution characteristics and the historical training data distribution. A distribution consistency index (such as feature mean offset or covariance matrix difference) is calculated relative to the historical training data. When the distribution consistency index is less than a preset threshold, the first fusion weight of the linear regression engine is determined to be higher than the second fusion weight of the rule-based weight engine. When the distribution consistency index is greater than or equal to the preset threshold, the first fusion weight is determined to be lower than the second fusion weight, and the sum of the first and second fusion weights is a fixed value.

[0032] For each of the three dimensions in the multi-level evaluation system, training samples within a preset time window are extracted from the historical time-series database to construct feature vectors, including policy volume features, regional coding features, average vehicle age features, and time-series features. Based on the training samples, a multiple linear regression equation is trained: Predicted indicator value = α × Feature 1 + β × Feature 2 + γ × Feature 3 + ε, where α, β, and γ are regression coefficients estimated using the least squares method, and ε is the error term. The current state data is input into the trained regression equation to output the basic risk score for each of the three dimensions.

[0033] The system maps the values ​​of each indicator in the current state data to the corresponding score intervals, and calculates the rule scores for each of the three-dimensional indicators using a linear interpolation algorithm. A bottom-up recursive traversal algorithm is employed for layer-by-layer aggregation. Based on the first and second fusion weights, the basic risk score and the rule risk score are weighted and fused: Fusion Risk Score = First Fusion Weight × Basic Risk Score + Second Fusion Weight × Rule Risk Score. According to the hierarchical structure of the multi-level evaluation system, the fusion risk score is normalized to generate the current risk score.

[0034] Step S40: Based on the current risk score and the preset risk score threshold, predict the target risk information of the preset business objective.

[0035] Specifically, the system retrieves current risk scores and historical trend data from memory and time-series databases, and conducts a comprehensive assessment by combining preset early warning rules and multi-level risk thresholds. This includes single-point threshold comparison, continuous periodic monitoring, year-on-year and month-on-month change rate analysis, and trend prediction. This determines the risk level and generates structured target risk information (including risk description, potential impacts, and recommended measures). Based on the risk level, the system triggers early warning notifications through multiple channels and initiates an early warning escalation mechanism to form a closed-loop response.

[0036] In one embodiment, the current risk score is compared with preset risk score thresholds at multiple levels to determine the risk level range corresponding to the current risk score. The preset risk score thresholds include a normal threshold, a warning threshold, and a critical threshold, with the values ​​of the normal threshold, warning threshold, and critical threshold increasing progressively. If the current risk score is less than or equal to the normal threshold, the risk level of the preset business objective is determined to be normal; if the current risk score is greater than the normal threshold but less than or equal to the warning threshold, the risk level is determined to be attentive; if the current risk score is greater than the warning threshold but less than or equal to the critical threshold, the risk level is determined to be low-risk; and if the current risk score is greater than the critical threshold, the risk level is determined to be high-risk.

[0037] Based on a weighted tree structure of a multi-level evaluation system, the contribution of each third-level, second-level, and first-level dimension indicator to the current risk score is traced from bottom to top. The sensitivity coefficient of each indicator is calculated, which is equal to the product of the indicator's weight and its score change rate, divided by the total score change rate. Key causal indicators with sensitivity coefficients exceeding a preset threshold are selected, and their data time-series changes within historical time windows are correlated to pinpoint the specific business processes and time points where data distribution anomalies occur, generating risk causal analysis information.

[0038] When the risk level is set to "Attention" or "High Risk," an intelligent early warning process is triggered. Target risk information is sent to preset receiving nodes via message push channels, including email, SMS, or instant messaging. If the target risk information is not processed within a preset response time limit, the early warning level is automatically escalated and forwarded to the superior management node.

[0039] This embodiment discloses a method, apparatus, device, and medium for generating risk prediction information. The method includes constructing a multi-level evaluation system based on preset business objectives, wherein the multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator; obtaining the current state data of the preset business objectives through a message queue; calculating the current risk score of the preset business objectives based on the dynamic weight configuration rules, linear regression engine, rule weight engine, and the current state data built into the multi-level evaluation system; and predicting the target risk information of the preset business objectives based on the current risk score and a preset risk score threshold. Through the above method, this application achieves dynamic alignment between evaluation dimensions and target business by constructing a multi-level evaluation system, using a message queue to obtain real-time data, and combining the data objectivity of the linear regression engine with the business logic of the rule weight engine for dual-engine fusion calculation, making the risk score both mathematically rigorous and business interpretable. The generation of risk prediction information based on real-time comparison of scores and thresholds enables risk prediction information generation to move from static assessment to dynamic early warning, improving the efficiency of risk management systems in processing multi-source heterogeneous business data in the insurance business field.

[0040] based on Figure 1 In the illustrated embodiment, step S10 includes: At least one primary dimension indicator is determined from the preset indicator library based on the preset business objectives; Obtain the detailed decomposition items corresponding to each of the first-level dimension indicators, and determine at least one of the second-level dimension indicators based on the weights of the second-level indicators and each of the detailed decomposition items; Obtain the computable items corresponding to each of the secondary dimension indicators, and determine at least one of the tertiary dimension indicators based on the weights of the tertiary indicators and each of the computable items. The multi-level evaluation system is constructed based on the first-level indicators, the second-level indicators, and the third-level indicators.

[0041] Specifically, based on preset business objectives, at least one primary dimension indicator is selected from a preset indicator library as the top-level evaluation direction of the evaluation system. For each primary dimension indicator, its corresponding detailed decomposition items are obtained as a candidate set of secondary dimension indicators. Based on preset secondary indicator weights, at least one secondary dimension indicator is determined from the candidate set, achieving a refined decomposition from primary to secondary dimensions. For each secondary dimension indicator, its corresponding computable items are obtained as a candidate set of tertiary dimension indicators. Based on preset tertiary indicator weights, at least one tertiary dimension indicator is determined from the candidate set, achieving the binding of secondary dimensions to specific business data indicators. A hierarchical, multi-level evaluation system is constructed through parent-child relationships.

[0042] In one embodiment, at least one primary dimension indicator (such as "policy cost") is selected from a preset indicator library based on a preset business objective (such as auto insurance cost optimization), and a global weight (such as 40%) is configured for it. The system automatically verifies whether the sum of the weights of all primary dimensions is 100%.

[0043] For each primary dimension indicator, obtain its corresponding detailed decomposition items (such as "resource overflow rate" and "cooperative factory production-to-insurance ratio"). Based on the weights of secondary indicators, assign each detailed item its proportion within the primary dimension (such as 60% and 40%). Automatically calculate the actual weight of the secondary dimension according to the parent-child linkage rules. Obtain the computable items corresponding to each secondary dimension indicator (such as binding "premium income" and "claims payment" data fields to "cooperative factory production-to-insurance ratio"), and set the linear scoring interval rules for the indicator according to the weights of tertiary indicators (such as setting threshold nodes and corresponding scores). Summarize all configured primary, secondary, and tertiary dimension indicators, their weight relationships, and scoring rules to generate a complete weight tree structure.

[0044] based on Figure 1 In the illustrated embodiment, step S20 includes: Streaming data of the preset business objectives is collected through a message queue; The streaming data is used to generate a sliding window rating trend curve based on the time series, and the rating trend curve is used as the current state data.

[0045] Specifically, streaming data of various business metrics are collected in real time through a message queue. An instantaneous risk score is calculated for each business source system, and this score, along with a timestamp, is saved to a time-series database. Based on the historical score data sequence stored in the time-series database, a sliding window algorithm (e.g., using a fixed-length window, such as 12 months, scrolling along the time axis) is used to aggregate and calculate the data points within the window (e.g., calculating a moving average), dynamically generating a continuous score trend curve.

[0046] For example, streaming data is collected from multiple business systems, including core underwriting, claims, and finance, covering key fields such as policy volume, claims payouts, and expense allocation. Sliding windows (e.g., a 12-month rolling window) are divided based on event time. Within each window, historical scores are aggregated and trend models are built, generating score trend curves that can be switched at daily, weekly, and monthly granularity. The year-on-year and month-on-month change rates of the curves are calculated, abnormal fluctuation ranges are marked, and the final result data, including trend characteristics and change rate indicators, is used as the current state data.

[0047] In a specific embodiment, step S40 includes: The current risk score and the preset risk score threshold are compared in multiple levels to determine the risk level range corresponding to the current risk score. The preset risk assessment threshold includes a normal threshold, an early warning threshold, and a critical threshold. If the current risk score is less than or equal to the normal threshold, the risk level of the preset business objective is determined to be in a normal state. If the current risk score is greater than the normal threshold and less than or equal to the warning threshold, the risk level is determined to be in a state of concern. If the current risk score is greater than the warning threshold and less than or equal to the critical threshold, the risk level is determined to be a low-risk state. If the current risk score is greater than the critical threshold, the risk level is determined to be a high-risk state. Based on the risk level, target risk information is generated according to the time series, which varies with the time series. The target risk information includes risk identification information, risk description information, and risk avoidance strategies.

[0048] Specifically, the current risk score, calculated in real time, is compared step by step with preset normal thresholds, warning thresholds, and critical thresholds to determine its corresponding risk level range.

[0049] If the score does not exceed the normal threshold, it is judged as "normal status"; if the score exceeds the normal threshold but does not exceed the warning threshold, it is judged as "attention status"; if the score exceeds the warning threshold but does not exceed the critical threshold, it is judged as "low-risk status"; if the score exceeds the critical threshold, it is judged as "high-risk status".

[0050] After determining the current risk level, the risk level is analyzed over time by combining the historical scoring sequence of the preset business objectives (e.g., whether it has entered the "low-risk state" for the first time from the "attention" state). Based on the determined risk level and its dynamic trend, a structured target risk information is generated. The target risk information includes risk identifiers (e.g., "Level 2 - Attention"), risk descriptions (e.g., "Cost score has risen for three consecutive periods, reaching the warning line for the first time"), and targeted risk avoidance strategy suggestions.

[0051] In a specific embodiment, in this embodiment, step S40 is followed by: Analyze the preset business objectives, and determine the key causal indicators of the preset business objectives based on the sensitivity of each of the first-level, second-level, and third-level dimension indicators to the preset business objectives; Obtain the progress information of the preset business objectives; The key causative indicators are correlated with the completion progress information, and the target risk information is dynamically adjusted.

[0052] Specifically, the preset business objectives are analyzed, and based on historical data and real-time simulations, the influence of the numerical fluctuations of each third-level dimension indicator in the multi-level evaluation system on the final risk score is analyzed to determine its sensitivity, and then indicators with high sensitivity are selected as key causal indicators.

[0053] Obtain real-time progress information of the preset business objective from relevant business or financial system interfaces, and conduct correlation analysis between key causal indicators and progress information. For example, if the progress is lagging behind, strengthen the risk warning caused by highly sensitive negative indicators, and highlight these key causes and suggest targeted measures in the dynamic description and avoidance strategy of target risk information. Conversely, if the progress is ahead of schedule, the risk level can be downgraded or the warning description can be adjusted accordingly.

[0054] In a specific embodiment, step S30 includes: Based on the linear regression engine and the current state data, calculate the basic risk score for the preset business objective; Based on the rule weight engine and the current state data, calculate the rule risk score for the preset business objective; The current risk score is calculated based on the dynamic weight configuration rules, the basic risk score, and the rule-based risk score.

[0055] Specifically, for each of the three-level dimensions in the multi-level evaluation system, a training sample set is extracted within a preset time window. The training samples include business data fields and their corresponding historical rating labels. Feature engineering is performed on the training samples to construct feature vectors, which at least include policy volume features, geographic coding features, average vehicle age features, and time series features. Based on the statistical correlation between the feature vectors and historical rating labels, a linear regression model is trained to estimate the regression coefficients and error compensation parameters corresponding to each feature dimension, establishing a mapping relationship between feature variables and indicator rating values. The current state data is input into the trained linear regression model, and the linear regression engine outputs the basic risk scores for each of the three-level dimensions.

[0056] Based on a weight tree structure of a multi-level evaluation system, pre-configured interval linear scoring rules are invoked for each third-level dimension indicator. These rules define the correspondence between indicator value intervals and score values. Each indicator value in the current state data is mapped to its corresponding score interval. When an indicator value falls between two threshold nodes, linear interpolation is used to calculate the score for that indicator within that interval. A bottom-up recursive traversal algorithm is employed to aggregate indicators at each level. Specifically, it iterates through all third-level dimension indicator child nodes under each second-level dimension indicator, multiplies the score of each third-level dimension indicator by its configured weight in the multi-level evaluation system, and then sums the results to obtain the aggregated score for each second-level dimension indicator. Similarly, it iterates through all second-level dimension indicator child nodes under each first-level dimension indicator, multiplies the aggregated score of each second-level dimension indicator by its configured weight, and then sums the results to obtain the aggregated score for each first-level dimension indicator. Finally, it multiplies the aggregated score of each first-level dimension indicator by its global weight and sums the results, with the rule weight engine outputting the rule risk score.

[0057] Dynamic weight configuration rules are read from the built-in configuration metadata of the multi-level evaluation system. These rules are used to adjust the output proportion of different computing engines based on data distribution characteristics. The consistency between the current state data and historical training data is evaluated. When the deviation between the feature distribution of the current state data and the feature distribution of historical training data is less than a preset threshold, the data distribution is considered stable. In this case, the basic risk score output by the linear regression engine is assigned a higher fusion weight, and the rule risk score output by the rule weight engine is assigned a lower fusion weight. When the deviation is greater than or equal to the preset threshold, the data distribution is considered to have drifted. In this case, the basic risk score is assigned a lower fusion weight, and the rule risk score is assigned a higher fusion weight, with the sum of the two fusion weights remaining a fixed value. Based on the determined fusion weights, the basic risk score and the rule risk score are weighted and fused to generate a fused risk score, which is then used as the current risk score.

[0058] In a specific embodiment, based on the rule weight engine and the current state data, the rule risk score of the preset business objective is calculated, including: Based on the dynamic weight configuration rules, determine the first fusion weight corresponding to the basic risk score and the second fusion weight corresponding to the rule risk score; Based on the first fusion weight and the second fusion weight, the basic risk score and the rule risk score are weighted and fused to generate a fusion risk score; Based on the hierarchical structure of the multi-level evaluation system, the fusion risk scores are aggregated layer by layer to generate the current risk score.

[0059] Specifically, based on the dynamic weight configuration rules and the determination of the data distribution status, the first fusion weight corresponding to the basic risk score and the second fusion weight corresponding to the rule-based risk score are determined respectively. When the data distribution is stable, the value of the first fusion weight is set higher than the value of the second fusion weight, allowing the output of the linear regression engine to dominate the fusion calculation. When the data distribution drifts, the value of the first fusion weight is set lower than the value of the second fusion weight, allowing the output of the rule-based weight engine to dominate the fusion calculation. The sum of the first and second fusion weights remains a preset fixed value to ensure that the numerical range of the fusion calculation result is controllable.

[0060] The basic risk score is scaled according to the first fusion weight, and the rule-based risk score is scaled according to the second fusion weight. The scaled basic risk score and the scaled rule-based risk score are then merged and summed to generate the fused risk score.

[0061] Based on the hierarchical structure of the multi-level evaluation system, the aforementioned bottom-up recursive traversal method is used to aggregate the fusion risk scores layer by layer, generating the current risk score of the preset business objective.

[0062] Please see Figure 2 , Figure 2 This is a schematic block diagram of a risk prediction information generation device provided in an embodiment of this application. This device is used to execute the aforementioned risk prediction information generation method. The risk prediction information generation device can be configured on a server.

[0063] like Figure 2 As shown, the risk prediction information generation device 400 includes: The multi-level evaluation system construction module 410 is used to construct a multi-level evaluation system based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. The current status data acquisition module 420 is used to acquire the current status data of the preset business target through a message queue; The current risk score calculation module 430 is used to calculate the current risk score of the preset business objective based on the dynamic weight configuration rules, linear regression engine, rule weight engine and the current state data built into the multi-level evaluation system. The target risk information prediction module 440 is used to predict the target risk information of the preset business target based on the current risk score and the preset risk score threshold.

[0064] Furthermore, the multi-level evaluation system construction module 410 includes: A first-level dimension indicator determination unit is used to determine at least one of the first-level dimension indicators from a preset indicator library based on the preset business objectives. The secondary dimension indicator determination unit is used to obtain the detailed decomposition items corresponding to each of the primary dimension indicators, and to determine at least one of the secondary dimension indicators based on the weight of the secondary indicator and each of the detailed decomposition items. The third-level dimension indicator determination unit is used to obtain the computable items corresponding to each of the second-level dimension indicators, and to determine at least one of the third-level dimension indicators based on the weights of the third-level indicators and each of the computable items. A multi-level evaluation system construction unit is used to determine the construction of the multi-level evaluation system based on each of the first-level dimension indicators, each of the second-level dimension indicators, and each of the third-level dimension indicators.

[0065] Furthermore, the current status data acquisition module 420 includes: A streaming data acquisition unit is used to acquire streaming data of the preset business target through a message queue; The current state data determination unit is used to generate a sliding window rating trend curve from the streaming data according to the time series, and use the rating trend curve as the current state data.

[0066] Furthermore, the target risk information prediction module 440 includes: A normal state determination unit is used to determine the risk level of the preset business objective as a normal state when the current risk score is less than or equal to the normal threshold. The attention status determination unit is used to determine the risk level as attention status when the current risk score is greater than the normal threshold and less than or equal to the warning threshold. The low-risk state determination unit is used to determine the risk level as a low-risk state when the current risk score is greater than the warning threshold and less than or equal to the critical threshold. A high-risk status determination unit is used to determine the risk level as a high-risk status when the current risk score is greater than the critical threshold. The target risk information generation unit is used to generate target risk information that changes with the time series according to the risk level and the time series, wherein the target risk information includes risk identification information, risk description information and risk avoidance strategies.

[0067] Furthermore, the risk prediction information generation device 400 also includes: The key causal indicator determination module is used to analyze the preset business objective and determine the key causal indicators of the preset business objective based on the sensitivity of each of the first-level dimension indicators, each of the second-level dimension indicators, and each of the third-level dimension indicators to the preset business objective. The progress information acquisition module is used to acquire the progress information of the preset business objectives. The target risk information adjustment module is used to associate the key causal indicators with the completion progress information and dynamically adjust the target risk information.

[0068] Furthermore, the current risk score calculation module 430 includes: The basic risk score calculation unit is used to calculate the basic risk score of the preset business objective based on the linear regression engine and the current state data. The rule risk scoring calculation unit is used to calculate the rule risk score of the preset business objective based on the rule weight engine and the current state data. The current risk score calculation unit is used to calculate the current risk score based on the dynamic weight configuration rules, the basic risk score, and the rule-based risk score.

[0069] Furthermore, the current risk score calculation unit includes: The fusion weight determination subunit is used to determine the first fusion weight corresponding to the basic risk score and the second fusion weight corresponding to the rule risk score according to the dynamic weight configuration rules. The risk score generation subunit is used to perform weighted fusion of the basic risk score and the rule risk score according to the first fusion weight and the second fusion weight to generate a fusion risk score; The current risk score generation subunit is used to aggregate the fused risk scores layer by layer according to the hierarchical structure of the multi-level evaluation system to generate the current risk score.

[0070] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and modules can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] The aforementioned device can be implemented as a computer program, which can be used in, for example... Figure 3 It runs on the computer device shown.

[0072] Please see Figure 3 , Figure 3This is a schematic block diagram illustrating the structure of a computer device according to an embodiment of this application. The computer device may be a server.

[0073] See Figure 3 The computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0074] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any method for generating risk prediction information.

[0075] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0076] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When these computer programs are executed by a processor, the processor can perform any method for generating risk prediction information.

[0077] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0078] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0079] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: A multi-level evaluation system is constructed based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. Obtain the current status data of the preset business objective through a message queue; The current risk score of the preset business objective is calculated based on the dynamic weight configuration rules, linear regression engine, rule weight engine built into the multi-level evaluation system and the current state data. Based on the current risk score and the preset risk score threshold, predict the target risk information of the preset business objective.

[0080] In one embodiment, a multi-level evaluation system is constructed based on preset business objectives. This multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator, used to achieve: At least one primary dimension indicator is determined from the preset indicator library based on the preset business objectives; Obtain the detailed decomposition items corresponding to each of the first-level dimension indicators, and determine at least one of the second-level dimension indicators based on the weights of the second-level indicators and each of the detailed decomposition items; Obtain the computable items corresponding to each of the secondary dimension indicators, and determine at least one of the tertiary dimension indicators based on the weights of the tertiary indicators and each of the computable items. The multi-level evaluation system is constructed based on the first-level indicators, the second-level indicators, and the third-level indicators.

[0081] In one embodiment, the current status data of the preset business objective is obtained through a message queue, which is used to achieve: Streaming data of the preset business objectives is collected through a message queue; The streaming data is used to generate a sliding window rating trend curve based on the time series, and the rating trend curve is used as the current state data.

[0082] In one embodiment, the target risk information of the preset business objective is predicted based on the current risk score and a preset risk score threshold, for the purpose of: If the current risk score is less than or equal to the normal threshold, the risk level of the preset business objective is determined to be in a normal state. If the current risk score is greater than the normal threshold and less than or equal to the warning threshold, the risk level is determined to be in a state of concern. If the current risk score is greater than the warning threshold and less than or equal to the critical threshold, the risk level is determined to be a low-risk state. If the current risk score is greater than the critical threshold, the risk level is determined to be a high-risk state. Based on the risk level, target risk information is generated according to the time series, which varies with the time series. The target risk information includes risk identification information, risk description information, and risk avoidance strategies.

[0083] In one embodiment, after predicting the target risk information of the preset business objective based on the current risk score and a preset risk score threshold, it is used to achieve: Analyze the preset business objectives, and determine the key causal indicators of the preset business objectives based on the sensitivity of each of the first-level, second-level, and third-level dimension indicators to the preset business objectives; Obtain the progress information of the preset business objectives; The key causative indicators are correlated with the completion progress information, and the target risk information is dynamically adjusted.

[0084] In one embodiment, the current risk score of the preset business objective is calculated based on the dynamic weight configuration rules, linear regression engine, rule weight engine, and current state data built into the multi-level evaluation system, to achieve: Based on the linear regression engine and the current state data, calculate the basic risk score for the preset business objective; Based on the rule weight engine and the current state data, calculate the rule risk score for the preset business objective; The current risk score is calculated based on the dynamic weight configuration rules, the basic risk score, and the rule-based risk score.

[0085] In one embodiment, a rule risk score for the preset business objective is calculated based on the rule weight engine and the current state data, to achieve the following: Based on the dynamic weight configuration rules, determine the first fusion weight corresponding to the basic risk score and the second fusion weight corresponding to the rule risk score; Based on the first fusion weight and the second fusion weight, the basic risk score and the rule risk score are weighted and fused to generate a fusion risk score; Based on the hierarchical structure of the multi-level evaluation system, the fusion risk scores are aggregated layer by layer to generate the current risk score.

[0086] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement any of the risk prediction information generation methods provided in the embodiments of this application.

[0087] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0088] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating risk prediction information, characterized in that, include: A multi-level evaluation system is constructed based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. Obtain the current status data of the preset business objective through a message queue; The current risk score of the preset business objective is calculated based on the dynamic weight configuration rules, linear regression engine, rule weight engine built into the multi-level evaluation system and the current state data. Based on the current risk score and the preset risk score threshold, predict the target risk information of the preset business objective.

2. The risk prediction information generation method according to claim 1, characterized in that, The construction of a multi-level evaluation system based on preset business objectives includes: At least one primary dimension indicator is determined from the preset indicator library based on the preset business objectives; Obtain the detailed decomposition items corresponding to each of the first-level dimension indicators, and determine at least one of the second-level dimension indicators based on the weights of the second-level indicators and each of the detailed decomposition items; Obtain the computable items corresponding to each of the secondary dimension indicators, and determine at least one of the tertiary dimension indicators based on the weights of the tertiary indicators and each of the computable items. The multi-level evaluation system is constructed based on the first-level indicators, the second-level indicators, and the third-level indicators.

3. The risk prediction information generation method according to claim 1, characterized in that, The step of obtaining the current status data of the preset business target through a message queue includes: Streaming data of the preset business objectives is collected through a message queue; The streaming data is used to generate a sliding window rating trend curve based on the time series, and the rating trend curve is used as the current state data.

4. The risk prediction information generation method according to claim 3, characterized in that, The preset risk assessment thresholds include a normal threshold, a warning threshold, and a critical threshold. The step of predicting the target risk information of the preset business objective based on the current risk score and the preset risk score thresholds includes: If the current risk score is less than or equal to the normal threshold, the risk level of the preset business objective is determined to be in a normal state. If the current risk score is greater than the normal threshold and less than or equal to the warning threshold, the risk level is determined to be in a state of concern. If the current risk score is greater than the warning threshold and less than or equal to the critical threshold, the risk level is determined to be a low-risk state. If the current risk score is greater than the critical threshold, the risk level is determined to be a high-risk state. Based on the risk level, target risk information is generated according to the time series, which varies with the time series. The target risk information includes risk identification information, risk description information, and risk avoidance strategies.

5. The risk prediction information generation method according to claim 4, characterized in that, After predicting the target risk information of the preset business objective based on the current risk score and the preset risk score threshold, the process includes: Analyze the preset business objectives, and determine the key causal indicators of the preset business objectives based on the sensitivity of each of the first-level, second-level, and third-level dimension indicators to the preset business objectives; Obtain the progress information of the preset business objectives; The key causative indicators are correlated with the completion progress information, and the target risk information is dynamically adjusted.

6. The risk prediction information generation method according to claim 1, characterized in that, The calculation of the current risk score of the preset business objective based on the dynamic weight configuration rules, linear regression engine, rule weight engine, and current state data built into the multi-level evaluation system includes: Based on the linear regression engine and the current state data, calculate the basic risk score for the preset business objective; Based on the rule weight engine and the current state data, calculate the rule risk score for the preset business objective; The current risk score is calculated based on the dynamic weight configuration rules, the basic risk score, and the rule-based risk score.

7. The risk prediction information generation method according to claim 6, characterized in that, The step of calculating the current risk score based on the dynamic weight configuration rules, the basic risk score, and the rule-based risk score includes: Based on the dynamic weight configuration rules, determine the first fusion weight corresponding to the basic risk score and the second fusion weight corresponding to the rule risk score; Based on the first fusion weight and the second fusion weight, the basic risk score and the rule risk score are weighted and fused to generate a fusion risk score; Based on the hierarchical structure of the multi-level evaluation system, the fusion risk scores are aggregated layer by layer to generate the current risk score.

8. A risk prediction information generation device, characterized in that, include: A multi-level evaluation system construction module is used to construct a multi-level evaluation system based on preset business objectives. The multi-level evaluation system includes at least one primary dimension indicator, at least one secondary dimension indicator, and at least one tertiary dimension indicator. The current status data acquisition module is used to acquire the current status data of the preset business target through a message queue; The current risk score calculation module is used to calculate the current risk score of the preset business objective based on the dynamic weight configuration rules, linear regression engine, rule weight engine and the current state data built into the multi-level evaluation system. The target risk information prediction module is used to predict the target risk information of the preset business target based on the current risk score and the preset risk score threshold.

9. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the risk prediction information generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, causes the processor to implement the risk prediction information generation method as described in any one of claims 1 to 7.

Images