Intelligent group life insurance management system and method based on group dynamic portraits

Abstract

The invention belongs to the technical field of life insurance management, and discloses an intelligent group life insurance management system based on group dynamic portraits, which comprises the following modules: a multi-dimensional data fusion module, a group health risk dynamic modeling module, a self-adaptive group pricing and automatic underwriting engine and a claim settlement anti-fraud collaborative network. Enterprise HR data, financial data, employee health data and industry macroscopic data are integrated through a multi-dimensional data fusion module to construct a group dynamic portrait, and comprehensive and dynamic data support is provided for risk assessment, so that group health risk dynamic modeling based on the portrait can accurately capture health risk changes of an insured group; according to the method, a rate adjustment model is established in combination with dynamic factors such as employee flow rate and enterprise financial stability, high matching of insurance premium and group real risks can be realized, more competitive prices are provided for low-risk enterprises to improve the customer retention rate, and accurate fee adding can be performed on high-risk enterprises to effectively cover risk openness.

Description

Technical Field

[0001] This invention belongs to the field of life insurance management technology, specifically an intelligent group life insurance management system and method based on dynamic group profiling. Background Technology

[0002] Group life insurance, as an important insurance product provided by insurance companies to enterprises or organizations, is based on risk control through scientific pricing and underwriting. It not only protects the protection needs of corporate employees but also maintains the business profitability of insurance companies, thus occupying an important position in both corporate employee welfare systems and the business layout of the insurance industry.

[0003] As companies grow in size and their workforce becomes more complex, the requirements for accuracy in risk assessment, efficiency in underwriting processes, and anti-fraud capabilities in group insurance are gradually increasing. However, current management methods in the group insurance sector still have significant shortcomings. In traditional group insurance management processes, insurance companies mainly rely on static historical data such as employee age, gender, and position provided by the company, combined with macroeconomic data such as industry life tables and average mortality rates, to conduct initial risk assessments. Subsequent manual review of documents by underwriters is also required, making the entire process heavily reliant on historical data and personal experience. Even some group insurance management systems that have introduced digital tools can only connect with a portion of the company's HR departments. The system acquires static data and calculates rates using a simple rule engine. However, complex cases still require manual intervention. Furthermore, existing technologies generally suffer from limited and lagging risk assessment dimensions, making it difficult to capture the impact of dynamic factors such as business conditions, employee health trends, and staff turnover on risk. This results in a mismatch between premiums and actual risks. At the same time, multi-source data processing relies on manual labor, which is not only inefficient and costly but also makes it difficult to uncover deep correlations between data, creating blind spots in risk identification. In addition, the data from various enterprises, insurance companies, and related institutions are isolated, failing to form a coordinated defense mechanism. This makes it difficult to detect organized group insurance fraud in a timely manner, adversely affecting the risk control and business profitability of insurance companies. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent group life insurance management system and method based on dynamic group profiling to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent group life insurance management system based on dynamic group profiling, comprising the following modules: a multi-dimensional data fusion module, a group health risk dynamic modeling module, an adaptive group pricing and automated underwriting engine, and a claims anti-fraud collaborative network; The multidimensional data fusion module is responsible for collecting information from scattered data sources, processing it, and integrating it into a unified and complete dynamic profile of the group. The dynamic modeling module for population health risk is based on the dynamic profile of the population provided by the multi-dimensional data fusion module, which dynamically assesses the health risk of the insured population and establishes a rate adjustment mechanism that matches the risk. The adaptive group pricing and automated underwriting engine is based on the evaluation results of the dynamic modeling module for group health risks, enabling differentiated pricing and automated underwriting processes, thereby improving pricing fairness and underwriting efficiency. The claims anti-fraud collaborative network enhances anti-fraud capabilities by establishing a cross-enterprise claims model knowledge base to assess the fraud risk of claims applications.

[0006] Preferably, the multidimensional data fusion module includes a data source access unit and a data acquisition unit; the data source access unit is responsible for establishing secure connections with various data providers, and adopting corresponding interface protocols and authorization mechanisms for different data sources; when accessing internal enterprise data, it obtains employee age, job level, length of service, department distribution, and employee turnover rate through the enterprise HR system API interface; when accessing enterprise operating data, it obtains enterprise revenue, profit, debt ratio, and cash flow stability through the enterprise financial statement API interface and third-party credit reporting platform interface; when accessing employee health data, it obtains anonymous employee annual physical examination indicators through the data interface of cooperative physical examination institutions; and it accesses industry macro data... According to the data acquisition system, the system obtains benchmark data on mortality and morbidity rates, as well as industry dynamic data, through official data interfaces of industry associations and public data source platforms. The data acquisition unit adopts two methods: real-time acquisition and periodic acquisition, depending on the data update frequency and usage requirements. Real-time acquisition targets employee turnover rate and changes in employee basic information. When relevant changes occur in the enterprise's HR system, the system automatically completes data extraction and transmission. Periodic acquisition targets enterprise financial data on the last working day of each quarter, automatically collects employee annual physical examination data within 15 working days after the end of the physical examination cycle, and automatically collects industry macro data on the 1st of each month. The system automatically sends data acquisition requests and completes the acquisition.

[0007] Preferably, the multidimensional data fusion module includes a data preprocessing unit and a data fusion unit. The data preprocessing unit is responsible for cleaning, standardizing, and anonymizing the collected heterogeneous data. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, desensitizes ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. The data fusion unit establishes multidimensional data associations with the company as the core entity to construct a dynamic profile of the group. The data association process associates operational data with industry data through the company's unique code, associates employee basic information with health data through the employee's unique identifier, and also associates employee turnover rate with the company's financial stability. The dynamic profile of the group includes basic company information, employee structure information, employee health status, company operating status, and employee turnover, and is updated synchronously with the collected data.

[0008] Preferably, the dynamic modeling module for group health risks includes an employee health index heatmap construction unit and a dynamic rate adjustment unit. The employee health index heatmap construction unit first converts key physiological indicators from physical examination reports into health scores according to medical standards, then calculates individual employee health indices based on indicator weights. Subsequently, it divides employees into groups by department and age group, calculates the average health index for each group, and finally generates a heatmap using a gradient of light green, light yellow, light orange, and dark red, labeling the number of employees in each group. The dynamic rate adjustment unit comprehensively considers industry mortality rate (40% weight), morbidity rate (30% weight), and average employee age. The benchmark rate is calculated based on the percentage of employees in high-risk departments (20% weight, with coefficients set according to age range) and the proportion of employees in high-risk departments (10% weight). Then, dynamic factors are determined (employee turnover rate is calculated monthly: monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%; corporate financial stability is scored based on debt-to-equity ratio (30% weight), cash flow ratio (40% weight), and profit growth rate (30% weight). Finally, the formula "final rate = benchmark rate × (1 + 0.3 × f(turnover rate) + 0.2 × g(financial stability))" is established, where f(turnover rate) and g(financial stability) are the influencing functions, and the final rate is automatically calculated every month.

[0009] Preferably, the adaptive group pricing and automated underwriting engine includes a tiered pricing unit and an automated underwriting process unit. The tiered pricing unit uses the K-Means clustering algorithm, selecting individual health index, age, job level, and length of service as feature variables, and iteratively determines stable high, medium, and low-risk groups. For low-risk groups, a 10% reduction in the base premium rate and additional health services are provided; for medium-risk groups, the base premium rate and basic health services are provided; and for high-risk groups, the base premium rate is increased by 20% and a 5% deductible is set. The automated underwriting process unit uses OCR technology to parse scanned employee rosters and NLP technology to parse PDF medical examination reports. The SHA-256 hash values ​​of core insurance information are stored on a Hyperledger Fabric consortium blockchain. Daily monitoring of employee health data, turnover rate, and financial stability changes is conducted. An alert is triggered when the group's average health index increases by more than 0.5, the turnover rate increases by more than 3%, or the financial score decreases by more than 10%, pushing alert information and improvement suggestions to the company's HR department.

[0010] Preferably, the claims anti-fraud collaborative network includes a group claims model knowledge base construction unit. This unit is responsible for collecting closed claims data from partner institutions, including insurance duration, incident time, ICD-10 disease diagnosis code, claim amount, personnel relationships, changes in the number of company employees, and changes in business status, and anonymizing the data. The data undergoes feature extraction, and a knowledge base is constructed using the Neo4j graph database. Nodes include claims cases, companies, employees, and diseases, while edges include ownership, involvement, association, and diagnosis relationships. The knowledge base is updated monthly.

[0011] Preferably, the claims anti-fraud collaborative network includes a collaborative analysis unit and a pattern matching and early warning unit. The collaborative analysis unit receives claims data sent by the insurance company through an encrypted interface, including insurance certificate number, effective date, date of incident, employee identification code, disease diagnosis, claim amount, date of employment, and current number of employees. It queries the blockchain insurance information based on the certificate number, verifies the identity of the employee involved in the incident and the duration of insurance coverage, and extracts features to ensure consistency with the features in the knowledge base. The pattern matching and early warning unit uses the GraphSAGE graph neural network algorithm to calculate the similarity between the new claims feature map and the fraud map in the knowledge base. The similarity score is set from 0 to 100, with a threshold of 80. A score above 80 indicates high risk, 60-80 indicates medium risk, and both trigger an early warning. A score below 60 indicates low risk. Early warning information for high-risk claims is pushed to the insurance company. The feature maps of confirmed fraud cases are added to the knowledge base, while normal cases are used to optimize model parameters.

[0012] A method for intelligent group life insurance management based on dynamic group profiling includes the following steps: S1, Multidimensional Data Acquisition and Fusion S1.1 Data Source Access: For the enterprise HR system, configure HTTPS protocol and OAuth2.0 authorization, complete enterprise identity authentication, and open API interface to obtain employee age, job level, length of service, department distribution, and employee turnover rate; for enterprise operating data, sign a data sharing agreement, open enterprise financial statement API interface and third-party credit reporting platform interface to obtain revenue, profit, debt ratio, cash flow stability, and verify authenticity; for employee health data, connect to the data interface of cooperative medical examination institutions to obtain anonymous medical examination indicators; for industry macro data, connect to industry association and public data source interfaces to obtain mortality rate, morbidity rate benchmark data and industry dynamics, and verify the source; S1.2 Data Collection: For real-time data such as employee turnover rate, job level adjustment, and departmental transfer, an HR system change trigger mechanism is set up so that the system completes data extraction and transmission within 10 seconds of receiving the change notification; for corporate financial data, a request is automatically sent to the financial API interface on the last working day of each quarter to extract the data for the current quarter; for employee annual physical examination data, data is automatically extracted from the physical examination institution interface within 15 working days after the end of the examination cycle; for industry macro data, data from the previous month is automatically extracted from industry association and public data source interfaces on the 1st of each month. S1.3 Data Preprocessing: The collected heterogeneous data is cleaned, standardized, and anonymized. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, desensitizes ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. S1.4 Data Integration Steps: Using the enterprise's unique code as the key, link the enterprise's operating data with the corresponding industry's macro data to clarify the enterprise's industry risk positioning; using the employee's unique identification code as the key, link the employee's basic information and health data to form an employee's personal file; link employee turnover rate with the enterprise's financial stability data to analyze the changing trends of both; integrate the enterprise's basic information, employee structure, health status, operating status, and turnover to construct a dynamic profile of the group; S2, Dynamic Assessment of Population Health Risks S2.1 Employee Health Index Heat Map Construction: Key physiological indicators from physical examination reports are converted into health scores according to medical standards; individual health indices (weights × scores summed) are calculated for each employee, with weights of 0.3 for blood pressure, 0.25 for blood sugar, 0.2 for BMI, and 0.25 for other indicators; groups are formed by department name and age groups of 20-30, 31-40, 41-50, and 51 and above; the average health index for each group is calculated; a heat map is created using a gradient color scheme of light green (average index < 1), light yellow (average index 1-2), light orange (average index 2-3), and dark red (average index > 3), with the group name labeled on the horizontal axis and the average index (average index 0-5) labeled on the vertical axis, with the number of employees labeled for each color block; S2.2 Dynamic Rate Adjustment: Extract the average industry mortality rate and average industry morbidity rate over the past three years, calculate the average age of company employees, and calculate the proportion of employees in high-risk departments; calculate the benchmark rate based on the following weights: industry mortality rate 40%, morbidity rate 30%, average employee age 20% (coefficients of 0.8 for 20-30 years old, 1.0 for 31-40 years old, 1.2 for 41-50 years old, and 1.5 for 51 years old and above), and the proportion of high-risk departments 10%; calculate the employee turnover rate for the previous month on the 1st of each month (monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%); calculate the company's debt-to-equity ratio (total liabilities ÷ total assets × 100%) and cash flow ratio (net operating cash flow). The financial stability score is calculated by standardizing the following three factors: net profit (total amount ÷ current liabilities × 100%), profit growth rate ((current year's net profit - previous year's net profit) ÷ previous year's net profit × 100%), and then weighting them according to 30%, 40%, and 30% respectively. Substituting these factors into the formula "Final Rate = Base Rate × (1 + 0.3 × f(current rate) + 0.2 × g(financial stability))" (f(current rate): <5% is 0, 5%-10% is current rate -5%, >10% is 0.05 + (current rate - 10%) × 1.5; g(financial stability): >80 points is -0.05, 60-80 points is 0, <60 points is 60% - score ÷ 100), yielding the final rate. S3, Adaptive Group Pricing and Automated Underwriting S3.1, Tiered Pricing: Employee individual health index, age, job level, and length of service are selected as clustering feature variables, and each variable is standardized to a 0-1 range. Three random cluster centers are initialized, and the Euclidean distance from each employee sample to the three centers is calculated. The sample is assigned to the group containing the nearest center. The average value of the feature variables for each group is calculated and updated as the new cluster centers. The distance calculation and sample assignment are repeated until the change in adjacent iteration centers is <0.001, resulting in high, medium, and low-risk groups. For the low-risk group (individual health index, age, job level, and length of service), the average value of the feature variables is calculated and updated as the new cluster centers. For individuals with an individual index <1, aged 20-40, and with more than 2 years of service, the base premium rate will be reduced by 10%, and one free annual physical examination and health consultation service will be provided. For individuals with an individual index <2, aged 31-50, and with more than 1 year of service, the base premium rate will be applied, and one free annual physical examination and health knowledge push service will be provided. For individuals with an individual index >2, aged ≥51, and with less than 1 year of service, the base premium rate will be increased by 20%, a deductible of 5% of the annual insured amount will be set, and a health monitoring report will be required to be submitted every six months. S3.2 Automated Underwriting: OCR technology is used to parse scanned employee rosters, and NLP technology is used to parse PDF medical examination reports. The identified information is organized into a structured table containing employee identification codes, indicator names, test values, reference ranges, and whether an abnormality is present. The insured's identification code list, individual health index, health declaration summary, and underwriting conclusion are integrated into a JSON data packet, and a SHA-256 hash value is calculated. The hash value, generation time, and enterprise code are sent to the consortium blockchain endorsement node. After verification by the endorsement node, a signature is generated. The signature information is sent to the sorting node, which sorts the data by timestamp to generate blocks. The blocks are then sent to all ledgers. Nodes are added to the blockchain ledger after verification. Enterprises can query the data by the certificate number, entering the username, password, and verifying the mobile phone verification code to view the hash value, certificate storage time, and participating nodes. Every day at midnight, the changes in individual employee health indices and group average indices are compared with the same period last week, and the differences in turnover rate and financial stability are compared with the same period last month. When the group average index rises by more than 0.5, the turnover rate rises by more than 3%, or the financial score falls by more than 10%, an early warning message is generated. The early warning message includes the warning group / indicator, the magnitude of change, the current value, and the number of high-risk individuals. The early warning message is pushed to the enterprise's HR via email and system messages, along with improvement suggestions. S4, Claims Anti-Fraud Collaboration S4.1 Group Claims Model Knowledge Base Construction: In conjunction with cooperating insurance companies, medical examination institutions, and third-party credit reporting agencies, collect data on group life insurance claims cases that have been settled by each institution; collect data on newly settled cases from cooperating institutions in the previous month before the 10th of each month, extract features and add them to the knowledge base, and delete old data that has been in the database for more than 5 years. S4.2 Collaborative Analysis Steps: After receiving a new claim application, the cooperating insurance company receives data through an encrypted interface, including the insurance certificate number, effective date, date of the incident, employee identification code list, disease diagnosis, claim amount, employee's start date, and current number of employees. The system queries the blockchain insurance information based on the certificate number to obtain the employee identification code list, number of employees at the time of insurance application, and effective date. It verifies whether the employee involved in the incident is in the list at the time of insurance application; if they do not match, it sends an anomaly alert requesting verification. After the data verification is successful, the system extracts features from the claim application data, and the extracted features are consistent with the case features in the knowledge base. S4.3 Pattern Matching and Early Warning: The features of new claims are constructed into a feature map; using the GraphSAGE graph neural network model, the new feature map and all fraud pattern maps in the knowledge base are input, the feature relationships between nodes and edges are learned, and a similarity score (0-100 points) is calculated; 80 points is set as the threshold, scores >80 points are marked as high risk, scores 60-80 points are marked as medium risk (both trigger an early warning), and scores <60 points are marked as low risk (no early warning); early warning information is generated, including the similarity score, the type of the matched fraud pattern, and a description of the high-risk features; the early warning information is pushed to the claims review department of the insurance company that submitted the application through an encrypted message interface, and is also recorded in the system log; if the insurance company confirms that the claim case is fraudulent, the feature map of the case is marked as a new fraud pattern and added to the knowledge base.

[0013] Preferably, as described in S4.3, if the insurance company confirms that there is no fraud in the claim case, it marks the feature map of the case as normal mode for use in optimizing the GraphSAGE model parameters.

[0014] The beneficial effects of this invention are as follows: By integrating enterprise HR data, financial data, employee health data, and industry macro data through a multi-dimensional data fusion module, a dynamic profile of the insured group is constructed. This provides comprehensive and dynamic data support for risk assessment, enabling dynamic modeling of group health risk based on this profile to accurately capture changes in the health risk of the insured group. Combined with a rate adjustment model established using dynamic factors such as employee turnover rate and corporate financial stability, a high degree of matching between premiums and the actual risk of the group can be achieved. This allows for more competitive pricing for low-risk companies to improve customer retention, while also enabling precise premium increases for high-risk companies to effectively cover their risk exposure. Simultaneously, the adaptive group pricing and automated underwriting engine employ K-Means clustering algorithm to achieve employee risk grouping and differentiated pricing, coupled with OCR and NL... P-technology automatically parses insurance application materials and blockchain-stored core information, constructing an automated underwriting process that frees underwriters from tedious and repetitive paperwork, significantly shortening the timeliness of group policy pricing and underwriting, and significantly reducing labor costs. In addition, the claims anti-fraud collaborative network, by jointly building a cross-enterprise group claims model knowledge base by multiple entities, and combining the GraphSAGE graph neural network algorithm to perform pattern matching and fraud risk assessment on new claims applications, can effectively identify organized insurance fraud. Furthermore, the group health risk dynamic modeling module monitors and warns of employee health data in real time, realizing a shift from passive claims to proactive risk control. Overall, this significantly enhances the risk control capabilities of group insurance business, reduces the overall loss ratio, and improves the profitability of insurance companies. Attached Figure Description

[0015] Figure 1 This is a block diagram of the multidimensional data fusion module of the present invention; Figure 2 This is a flowchart of the collaborative network for claims settlement and anti-fraud in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figures 1 to 2 As shown, this embodiment of the invention provides an intelligent group life insurance management system based on dynamic group profiling, including the following modules: a multi-dimensional data fusion module, a group health risk dynamic modeling module, an adaptive group pricing and automated underwriting engine, and a claims anti-fraud collaborative network; The multidimensional data fusion module is responsible for collecting information from scattered data sources, processing it, and integrating it into a unified and complete dynamic profile of the group. The dynamic modeling module for population health risks is based on the dynamic profile of the population provided by the multi-dimensional data fusion module. It dynamically assesses the health risks of the insured population and establishes a rate adjustment mechanism that matches the risks. The adaptive group pricing and automated underwriting engine, based on the evaluation results of the dynamic modeling module for group health risks, enables differentiated pricing and automated underwriting processes, thereby improving pricing fairness and underwriting efficiency. The claims anti-fraud collaborative network enhances anti-fraud capabilities by establishing a cross-enterprise claims model knowledge base to assess the fraud risk of claims applications.

[0018] The multi-dimensional data fusion module collects information from enterprise HR systems, financial systems, medical examination institutions, and industry data sources. After cleaning, standardization, and anonymization, it constructs a unified dynamic profile of the group with the enterprise at its core. The dynamic modeling module for group health risks generates a health index heatmap based on the profile and introduces employee turnover rate and financial stability to establish a dynamic rate adjustment model. The adaptive pricing underwriting engine uses K-Means clustering to classify risk groups and determine differentiated pricing. It uses OCR and NLP to parse data and stores it on the blockchain to achieve automated underwriting. The claims anti-fraud collaborative network unites multiple entities to establish a cross-enterprise claims model knowledge base and matches new applications to assess fraud risk.

[0019] The multi-dimensional data fusion module includes a data source access unit and a data acquisition unit. The data source access unit is responsible for establishing secure connections with various data providers, employing corresponding interface protocols and authorization mechanisms for different data sources. When accessing internal enterprise data, it obtains employee age, job level, length of service, departmental distribution, and employee turnover rate through the enterprise HR system API interface. When accessing enterprise operational data, it obtains enterprise revenue, profit, debt ratio, and cash flow stability through the enterprise financial statement API interface and third-party credit reporting platform interface. When accessing employee health data, it obtains anonymous annual employee health checkup indicators through the data interface of partner medical examination institutions. It also accesses industry macro data. At the same time, the system obtains benchmark data on mortality and morbidity rates, as well as industry dynamic data, through official data interfaces of industry associations and public data source platforms. The data collection unit adopts two methods: real-time collection and periodic collection, depending on the data update frequency and usage needs. Real-time collection targets employee turnover rate and changes in employee basic information. When relevant changes occur in the enterprise's HR system, the system automatically completes data extraction and transmission. Periodic collection targets enterprise financial data on the last working day of each quarter, automatically collects employee annual physical examination data within 15 working days after the end of the physical examination cycle, and automatically collects industry macro data on the 1st of each month. The system automatically sends data acquisition requests and completes the collection.

[0020] The multi-dimensional data fusion module clearly defines the data source access unit, which adopts corresponding interface protocols and authorization mechanisms for different data sources. This allows for the legal and secure acquisition of four key data categories: internal enterprise data, enterprise operational data, employee health data, and industry macro data, ensuring the compliance and security of data acquisition.

[0021] The multidimensional data fusion module includes a data preprocessing unit and a data fusion unit. The data preprocessing unit is responsible for cleaning, standardizing, and anonymizing the collected heterogeneous data. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, de-identifies ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. The data fusion unit establishes multidimensional data associations with the company as the core entity to construct a dynamic profile of the group. The data association process associates operational data with industry data through the company's unique code, and associates employee basic information with health data through the employee's unique identifier, while also associating employee turnover rate with the company's financial stability. The dynamic profile of the group includes basic company information, employee structure information, employee health status, company operating status, and employee turnover, and is updated synchronously with the collected data.

[0022] The data fusion unit establishes multi-dimensional data associations with enterprises as the core entity, and constructs a dynamic profile of the group that includes basic enterprise information, employee structure information, employee health status, enterprise operation status, and employee turnover, and is updated in real time. It can integrate scattered data to form a complete data view, providing accurate and comprehensive data basis for subsequent risk assessment.

[0023] The dynamic modeling module for group health risks includes an employee health index heatmap construction unit and a dynamic rate adjustment unit. The employee health index heatmap construction unit first converts key physiological indicators from physical examination reports into health scores according to medical standards, then calculates individual employee health indices based on indicator weights. Subsequently, it divides employees into groups by department and age group, calculates the average health index for each group, and finally generates a heatmap using a gradient of light green, light yellow, light orange, and dark red, labeling the number of employees in each group. The dynamic rate adjustment unit comprehensively considers industry mortality rate (40% weight), morbidity rate (30% weight), and average employee age (20% weight). The system calculates the benchmark rate based on the following factors: weights (e.g., coefficients set according to age ranges) and the proportion of employees in high-risk departments (10% weight); dynamic factors are then determined (employee turnover rate is calculated monthly: monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%; corporate financial stability is scored based on debt-to-equity ratio (30% weight), cash flow ratio (40% weight), and profit growth rate (30% weight); finally, the formula "final rate = benchmark rate × (1 + 0.3 × f(turnover rate) + 0.2 × g(financial stability))" is established, where f(turnover rate) and g(financial stability) are the influencing functions, and the final rate is automatically calculated monthly.

[0024] The employee health index heatmap construction unit in the dynamic modeling module of group health risk converts physical examination indicators into health scores, calculates the average health index of individuals and groups, and generates a gradient color heatmap, which can intuitively display the health risk level of employees in different departments and age groups, making it easy to quickly identify high-risk groups.

[0025] The adaptive group pricing and automated underwriting engine includes a tiered pricing unit and an automated underwriting process unit. The tiered pricing unit uses the K-Means clustering algorithm, selecting individual health index, age, job level, and length of service as feature variables, and iteratively determines stable high, medium, and low-risk groups. For low-risk groups, the benchmark premium is reduced by 10% and additional health services are provided; for medium-risk groups, the benchmark premium and basic health services are provided; and for high-risk groups, the benchmark premium is increased by 20% and a 5% deductible is set. The automated underwriting process unit uses OCR technology to parse scanned employee rosters and NLP technology to parse PDF medical examination reports. The SHA-256 hash values ​​of core insurance information are stored on a consortium blockchain based on the Hyperledger Fabric architecture. The unit monitors changes in employee health data, turnover rate, and financial stability daily. When the average health index of a group increases by more than 0.5, the turnover rate increases by more than 3%, or the financial score decreases by more than 10%, an alert is triggered, and alert information and improvement suggestions are pushed to the company's HR department.

[0026] The adaptive group pricing and automated underwriting engine uses the K-Means clustering algorithm to divide high, medium and low risk groups, and formulates differentiated pricing strategies and health service plans for different groups, which can achieve fair pricing and meet the needs of groups with different risk levels.

[0027] The claims anti-fraud collaborative network includes a group claims model knowledge base construction unit. This unit is responsible for collecting closed claims data from partner institutions, including insurance duration, incident time, ICD-10 disease diagnosis code, claim amount, personnel relationships, changes in the number of employees, and changes in business status, and anonymizes the data. The unit also extracts features from the data and uses the Neo4j graph database to build the knowledge base. Nodes include claims cases, companies, employees, and diseases, while edges include ownership, involvement, association, and diagnosis relationships. The knowledge base is updated monthly.

[0028] The knowledge base building unit for group claims models in the claims anti-fraud collaborative network collects and anonymizes closed claims data from partner institutions, protecting privacy while accumulating rich case data.

[0029] The claims anti-fraud collaborative network includes a collaborative analysis unit and a pattern matching and early warning unit. The collaborative analysis unit receives claims data from insurance companies via an encrypted interface, including insurance certificate number, effective date, date of incident, employee identification code, disease diagnosis, claim amount, start date, and current number of employees. It queries blockchain insurance information based on the certificate number to verify the identity of the employee involved in the incident and the duration of their insurance coverage, extracting features that match those in the knowledge base. The pattern matching and early warning unit uses the GraphSAGE graph neural network algorithm to calculate the similarity between the new claims feature graph and the fraud graph in the knowledge base. The similarity score is set from 0 to 100, with a threshold of 80. A score above 80 indicates high risk, 60-80 indicates medium risk, both triggering an early warning; a score below 60 indicates low risk. Early warning information for high-risk claims is pushed to the insurance company. Feature graphs of confirmed fraud cases are added to the knowledge base, while normal cases are used to optimize model parameters.

[0030] The pattern matching and early warning unit uses the GraphSAGE graph neural network algorithm to calculate similarity, distinguishes risk levels according to thresholds and triggers early warnings, which can accurately identify claims fraud risks and push early warning information to insurance companies to help them conduct efficient reviews.

[0031] A method for intelligent group life insurance management based on dynamic group profiling includes the following steps: S1, Multidimensional Data Acquisition and Fusion S1.1 Data Source Access: For the enterprise HR system, configure HTTPS protocol and OAuth2.0 authorization, complete enterprise identity authentication, and open API interface to obtain employee age, job level, length of service, department distribution, and employee turnover rate; for enterprise operating data, sign a data sharing agreement, open enterprise financial statement API interface and third-party credit reporting platform interface to obtain revenue, profit, debt ratio, cash flow stability, and verify authenticity; for employee health data, connect to the data interface of cooperative medical examination institutions to obtain anonymous medical examination indicators; for industry macro data, connect to industry association and public data source interfaces to obtain mortality rate, morbidity rate benchmark data and industry dynamics, and verify the source; S1.2 Data Collection: For real-time data such as employee turnover rate, job level adjustment, and departmental transfer, an HR system change trigger mechanism is set up so that the system completes data extraction and transmission within 10 seconds of receiving the change notification; for corporate financial data, a request is automatically sent to the financial API interface on the last working day of each quarter to extract the data for the current quarter; for employee annual physical examination data, data is automatically extracted from the physical examination institution interface within 15 working days after the end of the examination cycle; for industry macro data, data from the previous month is automatically extracted from industry association and public data source interfaces on the 1st of each month. S1.3 Data Preprocessing: The collected heterogeneous data is cleaned, standardized, and anonymized. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, desensitizes ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. S1.4 Data Integration Steps: Using the enterprise's unique code as the key, link the enterprise's operating data with the corresponding industry's macro data to clarify the enterprise's industry risk positioning; using the employee's unique identification code as the key, link the employee's basic information and health data to form an employee's personal file; link employee turnover rate with the enterprise's financial stability data to analyze the changing trends of both; integrate the enterprise's basic information, employee structure, health status, operating status, and turnover to construct a dynamic profile of the group; S2, Dynamic Assessment of Population Health Risks S2.1 Employee Health Index Heat Map Construction: Key physiological indicators from physical examination reports are converted into health scores according to medical standards; individual health indices (weights × scores summed) are calculated for each employee, with weights of 0.3 for blood pressure, 0.25 for blood sugar, 0.2 for BMI, and 0.25 for other indicators; groups are formed by department name and age groups of 20-30, 31-40, 41-50, and 51 and above; the average health index for each group is calculated; a heat map is created using a gradient color scheme of light green (average index < 1), light yellow (average index 1-2), light orange (average index 2-3), and dark red (average index > 3), with the group name labeled on the horizontal axis and the average index (average index 0-5) labeled on the vertical axis, with the number of employees labeled for each color block; S2.2 Dynamic Rate Adjustment: Extract the average industry mortality rate and average industry morbidity rate over the past three years, calculate the average age of company employees, and calculate the proportion of employees in high-risk departments; calculate the benchmark rate based on the following weights: industry mortality rate 40%, morbidity rate 30%, average employee age 20% (coefficients of 0.8 for 20-30 years old, 1.0 for 31-40 years old, 1.2 for 41-50 years old, and 1.5 for 51 years old and above), and the proportion of high-risk departments 10%; calculate the employee turnover rate for the previous month on the 1st of each month (monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%); calculate the company's debt-to-equity ratio (total liabilities ÷ total assets × 100%) and cash flow ratio (net operating cash flow). The financial stability score is calculated by standardizing the following three factors: net profit (total amount ÷ current liabilities × 100%), profit growth rate ((current year's net profit - previous year's net profit) ÷ previous year's net profit × 100%), and then weighting them according to 30%, 40%, and 30% respectively. Substituting these factors into the formula "Final Rate = Base Rate × (1 + 0.3 × f(current rate) + 0.2 × g(financial stability))" (f(current rate): <5% is 0, 5%-10% is current rate -5%, >10% is 0.05 + (current rate - 10%) × 1.5; g(financial stability): >80 points is -0.05, 60-80 points is 0, <60 points is 60% - score ÷ 100), yielding the final rate. S3, Adaptive Group Pricing and Automated Underwriting S3.1, Tiered Pricing: Employee individual health index, age, job level, and length of service are selected as clustering feature variables, and each variable is standardized to a 0-1 range. Three random cluster centers are initialized, and the Euclidean distance from each employee sample to the three centers is calculated. The sample is assigned to the group containing the nearest center. The average value of the feature variables for each group is calculated and updated as the new cluster centers. The distance calculation and sample assignment are repeated until the change in adjacent iteration centers is <0.001, resulting in high, medium, and low-risk groups. For the low-risk group (individual health index, age, job level, and length of service), the average value of the feature variables is calculated and updated as the new cluster centers. For individuals with an individual index <1, aged 20-40, and with more than 2 years of service, the base premium rate will be reduced by 10%, and one free annual physical examination and health consultation service will be provided. For individuals with an individual index <2, aged 31-50, and with more than 1 year of service, the base premium rate will be applied, and one free annual physical examination and health knowledge push service will be provided. For individuals with an individual index >2, aged ≥51, and with less than 1 year of service, the base premium rate will be increased by 20%, a deductible of 5% of the annual insured amount will be set, and a health monitoring report will be required to be submitted every six months. S3.2 Automated Underwriting: OCR technology is used to parse scanned employee rosters, and NLP technology is used to parse PDF medical examination reports. The identified information is organized into a structured table containing employee identification codes, indicator names, test values, reference ranges, and whether an abnormality is present. The insured's identification code list, individual health index, health declaration summary, and underwriting conclusion are integrated into a JSON data packet, and a SHA-256 hash value is calculated. The hash value, generation time, and enterprise code are sent to the consortium blockchain endorsement node. After verification by the endorsement node, a signature is generated. The signature information is sent to the sorting node, which sorts the data by timestamp to generate blocks. The blocks are then sent to all ledgers. Nodes are added to the blockchain ledger after verification. Enterprises can query the data by the certificate number, entering the username, password, and verifying the mobile phone verification code to view the hash value, certificate storage time, and participating nodes. Every day at midnight, the changes in individual employee health indices and group average indices are compared with the same period last week, and the differences in turnover rate and financial stability are compared with the same period last month. When the group average index rises by more than 0.5, the turnover rate rises by more than 3%, or the financial score falls by more than 10%, an early warning message is generated. The early warning message includes the warning group / indicator, the magnitude of change, the current value, and the number of high-risk individuals. The early warning message is pushed to the enterprise's HR via email and system messages, along with improvement suggestions. S4, Claims Anti-Fraud Collaboration S4.1 Group Claims Model Knowledge Base Construction: In conjunction with cooperating insurance companies, medical examination institutions, and third-party credit reporting agencies, collect data on group life insurance claims cases that have been settled by each institution; collect data on newly settled cases from cooperating institutions in the previous month before the 10th of each month, extract features and add them to the knowledge base, and delete old data that has been in the database for more than 5 years. S4.2 Collaborative Analysis Steps: After receiving a new claim application, the cooperating insurance company receives data through an encrypted interface, including the insurance certificate number, effective date, date of the incident, employee identification code list, disease diagnosis, claim amount, employee's start date, and current number of employees. The system queries the blockchain insurance information based on the certificate number to obtain the employee identification code list, number of employees at the time of insurance application, and effective date. It verifies whether the employee involved in the incident is in the list at the time of insurance application; if they do not match, it sends an anomaly alert requesting verification. After the data verification is successful, the system extracts features from the claim application data, and the extracted features are consistent with the case features in the knowledge base. S4.3 Pattern Matching and Early Warning: The features of new claims are constructed into a feature map; using the GraphSAGE graph neural network model, the new feature map and all fraud pattern maps in the knowledge base are input, the feature relationships between nodes and edges are learned, and a similarity score (0-100 points) is calculated; 80 points is set as the threshold, scores >80 points are marked as high risk, scores 60-80 points are marked as medium risk (both trigger an early warning), and scores <60 points are marked as low risk (no early warning); early warning information is generated, including the similarity score, the type of the matched fraud pattern, and a description of the high-risk features; the early warning information is pushed to the claims review department of the insurance company that submitted the application through an encrypted message interface, and is also recorded in the system log; if the insurance company confirms that the claim case is fraudulent, the feature map of the case is marked as a new fraud pattern and added to the knowledge base.

[0032] By systematically integrating corporate HR data, financial data, employee health data, and industry macro data, a dynamic and comprehensive risk profile of the insured group is formed. "Employee turnover rate" and "corporate financial stability" are used as core dynamic factors to adjust insurance rates in real time, so that the rates can fluctuate with changes in risk.

[0033] In S4.3, if the insurance company confirms that there is no fraud in the claim case, it marks the feature map of the case as normal mode, which is used to optimize the parameters of the GraphSAGE model.

[0034] The feature maps of claims that are not fraudulent are clearly identified and marked as normal patterns and used to optimize the parameters of the GraphSAGE model. This allows the model to be iterated using real normal case data, correct model biases, and improve the model's ability to distinguish between normal and fraudulent claims.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 process, method, article, or apparatus.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent group life insurance management system based on dynamic group profiling, characterized in that, It includes the following modules: multi-dimensional data fusion module, dynamic modeling module for group health risks, adaptive group pricing and automated underwriting engine, and claims anti-fraud collaborative network; The multidimensional data fusion module is responsible for collecting information from scattered data sources, processing it, and integrating it into a unified and complete dynamic profile of the group. The dynamic modeling module for population health risk is based on the dynamic profile of the population provided by the multi-dimensional data fusion module, which dynamically assesses the health risk of the insured population and establishes a rate adjustment mechanism that matches the risk. The adaptive group pricing and automated underwriting engine is based on the evaluation results of the dynamic modeling module for group health risks, enabling differentiated pricing and automated underwriting processes, thereby improving pricing fairness and underwriting efficiency. The claims anti-fraud collaborative network enhances anti-fraud capabilities by establishing a cross-enterprise claims model knowledge base to assess the fraud risk of claims applications.

2. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The multi-dimensional data fusion module includes a data source access unit and a data acquisition unit. The data source access unit is responsible for establishing secure connections with various data providers, employing corresponding interface protocols and authorization mechanisms for different data sources. When accessing internal enterprise data, it obtains employee age, job level, length of service, departmental distribution, and employee turnover rate through the enterprise HR system API interface. When accessing enterprise operational data, it obtains enterprise revenue, profit, debt ratio, and cash flow stability through the enterprise financial statement API interface and third-party credit reporting platform interface. When accessing employee health data, it obtains anonymous annual employee health checkup indicators through the data interface of cooperating medical examination institutions. When accessing industry macro data... The system acquires benchmark data on mortality and morbidity rates, as well as industry dynamics data, through official data interfaces of industry associations and public data source platforms. The data acquisition unit adopts two methods: real-time acquisition and periodic acquisition, depending on the data update frequency and usage requirements. Real-time acquisition targets employee turnover rate and changes in employee basic information. When relevant changes occur in the enterprise's HR system, the system automatically completes data extraction and transmission. Periodic acquisition targets enterprise financial data on the last working day of each quarter, automatically collects employee annual physical examination data within 15 working days after the end of the physical examination cycle, and automatically collects industry macro data on the 1st of each month. The system automatically sends data acquisition requests and completes the acquisition.

3. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The multidimensional data fusion module includes a data preprocessing unit and a data fusion unit. The data preprocessing unit is responsible for cleaning, standardizing, and anonymizing the collected heterogeneous data. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, desensitizes ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. The data fusion unit establishes multidimensional data associations with the company as the core entity to construct a dynamic profile of the group. The data association process associates operational data with industry data through the company's unique code, associates employee basic information with health data through the employee's unique identifier, and also associates employee turnover rate with the company's financial stability. The dynamic profile of the group includes basic enterprise information, employee structure information, employee health status, enterprise operation status, and employee turnover, and is updated synchronously with the collected data.

4. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The dynamic modeling module for group health risks includes an employee health index heatmap construction unit and a dynamic rate adjustment unit. The employee health index heatmap construction unit first converts key physiological indicators from physical examination reports into health scores according to medical standards, then calculates individual employee health indices based on indicator weights. Subsequently, it divides employees into groups by department and age group, calculates the average health index for each group, and finally generates a heatmap using a gradient of light green, light yellow, light orange, and dark red, labeling the number of employees in each group. The dynamic rate adjustment unit comprehensively considers industry mortality rate (40% weight), morbidity rate (30% weight), and average employee age (2... The benchmark rate is calculated based on the following factors: 0% weight (with coefficients set according to age range) and the proportion of employees in high-risk departments (10% weight); then dynamic factors are determined (employee turnover rate is calculated monthly: monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%; corporate financial stability is scored based on debt-to-equity ratio (30% weight), cash flow ratio (40% weight), and profit growth rate (30% weight); finally, the formula "final rate = benchmark rate × (1 + 0.3 × f(turnover rate) + 0.2 × g(financial stability))" is established, where f(turnover rate) and g(financial stability) are the influencing functions, and the final rate is automatically calculated monthly.

5. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The adaptive group pricing and automated underwriting engine includes a tiered pricing unit and an automated underwriting process unit. The tiered pricing unit uses the K-Means clustering algorithm to select individual health index, age, job level, and length of service as feature variables, and iteratively determines stable high, medium, and low risk groups. For low-risk groups, the base premium rate is reduced by 10% and additional health services are provided. For medium-risk groups, the base premium rate and basic health services are provided. For high-risk groups, the base premium rate is increased by 20% and a 5% deductible is set. The automated underwriting process unit uses OCR technology to parse scanned copies of employee rosters and NLP technology to parse PDF medical examination reports. The SHA-256 hash values ​​of core insurance information are stored on a consortium blockchain based on the Hyperledger Fabric architecture. The system monitors changes in employee health data, turnover rate, and financial stability daily. When the average health index of a group increases by more than 0.5, the turnover rate increases by more than 3%, or the financial score decreases by more than 10%, an alert is triggered, and alert information and improvement suggestions are pushed to the company's HR department.

6. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The claims anti-fraud collaborative network includes a group claims model knowledge base construction unit. This unit is responsible for collecting closed claims data from partner institutions, including insurance duration, incident time, ICD-10 disease diagnosis code, claim amount, personnel relationships, changes in the number of company employees, and changes in business status, and anonymizes the data. Features are extracted from the data, and a knowledge base is constructed using the Neo4j graph database. Nodes include claims cases, companies, employees, and diseases, while edges include ownership, involvement, association, and diagnosis relationships. The knowledge base is updated monthly.

7. The intelligent group life insurance management system based on dynamic group profiling according to claim 1, characterized in that: The claims anti-fraud collaborative network includes a collaborative analysis unit and a pattern matching and early warning unit. The collaborative analysis unit receives claims data from the insurance company via an encrypted interface, including insurance certificate number, effective date, date of incident, employee identification code, disease diagnosis, claim amount, start date, and current number of employees. It queries the blockchain insurance information based on the certificate number, verifies the identity of the employee involved in the incident and the duration of the insurance policy, and extracts features to ensure consistency with the features in the knowledge base. The pattern matching and early warning unit uses the GraphSAGE graph neural network algorithm to calculate the similarity between the new claims feature map and the fraud map in the knowledge base. The similarity score is set from 0 to 100, with a threshold of 80. A score above 80 indicates high risk, 60-80 indicates medium risk, both triggering an early warning; a score below 60 indicates low risk. Early warning information for high-risk claims is pushed to the insurance company. The feature maps of confirmed fraud cases are added to the knowledge base, while normal cases are used to optimize model parameters.

8. A method for intelligent group life insurance management based on dynamic group profiling, characterized in that: Includes the following steps: S1, Multidimensional Data Acquisition and Fusion S1.1 Data Source Access: For the enterprise HR system, configure HTTPS protocol and OAuth2.0 authorization, complete enterprise identity authentication, and open API interface to obtain employee age, job level, length of service, department distribution, and employee turnover rate; for enterprise operating data, sign a data sharing agreement, open enterprise financial statement API interface and third-party credit reporting platform interface to obtain revenue, profit, debt ratio, cash flow stability, and verify authenticity; for employee health data, connect to the data interface of cooperative medical examination institutions to obtain anonymous medical examination indicators; for industry macro data, connect to industry association and public data source interfaces to obtain mortality rate, morbidity rate benchmark data and industry dynamics, and verify the source; S1.2 Data Collection: For real-time data such as employee turnover rate, job level adjustment, and departmental transfer, an HR system change trigger mechanism is set up so that the system completes data extraction and transmission within 10 seconds of receiving the change notification; for corporate financial data, a request is automatically sent to the financial API interface on the last working day of each quarter to extract the data for the current quarter; for employee annual physical examination data, data is automatically extracted from the physical examination institution interface within 15 working days after the end of the examination cycle; for industry macro data, data from the previous month is automatically extracted from industry association and public data source interfaces on the 1st of each month. S1.3 Data Preprocessing: The collected heterogeneous data is cleaned, standardized, and anonymized. The cleaning process removes invalid data, corrects erroneous data, deletes duplicate data, and fills in missing data. The standardization process converts data of different formats and units into the system's specified format. The anonymization process replaces employee names with unique identifiers, desensitizes ID numbers, encrypts company names, and generates temporary medical examination codes to protect privacy information. S1.4 Data Integration Steps: Using the enterprise's unique code as the key, link the enterprise's operating data with the corresponding industry's macro data to clarify the enterprise's industry risk positioning; using the employee's unique identification code as the key, link the employee's basic information and health data to form an employee's personal file; link employee turnover rate with the enterprise's financial stability data to analyze the changing trends of both; integrate the enterprise's basic information, employee structure, health status, operating status, and turnover to construct a dynamic profile of the group; S2, Dynamic Assessment of Population Health Risks S2.1 Employee Health Index Heat Map Construction: Key physiological indicators from physical examination reports are converted into health scores according to medical standards; individual health indices (weights × scores summed) are calculated for each employee, with weights of 0.3 for blood pressure, 0.25 for blood sugar, 0.2 for BMI, and 0.25 for other indicators; groups are formed by department name and age groups of 20-30, 31-40, 41-50, and 51 and above; the average health index for each group is calculated; a heat map is created using a gradient color scheme of light green (average index < 1), light yellow (average index 1-2), light orange (average index 2-3), and dark red (average index > 3), with the group name labeled on the horizontal axis and the average index (average index 0-5) labeled on the vertical axis, with the number of employees labeled for each color block; S2.2 Dynamic Rate Adjustment: Extract the average industry mortality rate and average industry morbidity rate over the past three years, calculate the average age of company employees, and calculate the proportion of employees in high-risk departments; calculate the benchmark rate based on the following weights: industry mortality rate 40%, morbidity rate 30%, average employee age 20% (coefficients of 0.8 for 20-30 years old, 1.0 for 31-40 years old, 1.2 for 41-50 years old, and 1.5 for 51 years old and above), and the proportion of high-risk departments 10%; calculate the employee turnover rate for the previous month on the 1st of each month (monthly turnover rate = (number of employees leaving in the current month ÷ total number of employees at the beginning of the current month) × 100%); calculate the company's debt-to-equity ratio (total liabilities ÷ total assets × 100%) and cash flow ratio (net operating cash flow). The financial stability score is calculated by standardizing the following three factors: net profit (total amount ÷ current liabilities × 100%), profit growth rate ((current year's net profit - previous year's net profit) ÷ previous year's net profit × 100%), and then weighting them according to 30%, 40%, and 30% respectively. Substituting these factors into the formula "Final Rate = Base Rate × (1 + 0.3 × f(current rate) + 0.2 × g(financial stability))" (f(current rate): <5% is 0, 5%-10% is current rate -5%, >10% is 0.05 + (current rate - 10%) × 1.5; g(financial stability): >80 points is -0.05, 60-80 points is 0, <60 points is 60% - score ÷ 100), to obtain the final rate. S3, Adaptive Group Pricing and Automated Underwriting S3.1, Tiered Pricing: Select individual employee health index, age, job level, and length of service as clustering feature variables, and standardize each variable to a value between 0 and 1; initialize 3 random cluster centers, calculate the Euclidean distance of each employee sample to the 3 centers, and assign the sample to the group containing the nearest center; calculate the average value of the feature variables of each group and update it as the new cluster center; repeat the distance calculation and sample assignment until the change in adjacent iteration centers is <0.001, to obtain high, medium, and low risk groups; For low-risk groups (individual index < 1, age 20-40, length of service > 2 years), the base premium rate will be reduced by 10%, and one free annual physical examination and health consultation service will be provided; for medium-risk groups (individual index 1-2, age 31-50, length of service 1-2 years), the base premium rate will be applied, and one free annual physical examination and health knowledge push service will be provided; for high-risk groups (individual index > 2, age ≥ 51, length of service < 1 year), the base premium rate will be increased by 20%, a deductible of 5% of the annual insured amount will be set, and a health monitoring report will be required to be submitted every six months. S3.2 Automated Underwriting: OCR technology is used to parse scanned employee rosters, and NLP technology is used to parse PDF medical examination reports. The identified information is organized into a structured table containing employee identification codes, indicator names, test values, reference ranges, and whether they are abnormal. The list of insured personnel identification codes, individual health indices, health declaration summaries, and underwriting conclusions are integrated into a JSON data packet, and the SHA-256 hash value is calculated. The hash value, generation time, and enterprise code are sent to the consortium blockchain endorsement node. After the endorsement node verifies the data, a signature is generated. The signature information is sent to the sorting node, and blocks are generated by sorting the data by timestamp. The block is sent to all accounting nodes, verified, and added to the blockchain ledger. Enterprises can query the evidence by the evidence number, enter the username, password, and verify the mobile phone verification code to view the hash value, evidence storage time, and participating nodes. Every day at midnight, the changes in individual employee health index and group average index are compared with the same period last week, and the differences in turnover rate and financial stability are compared with the same period last month. When the group average index rises by more than 0.5, the turnover rate rises by more than 3%, and the financial score falls by more than 10%, an early warning message is generated. The early warning message includes the warning group / indicator, the magnitude of change, the current value, and the number of high-risk individuals. Warning information is pushed to the company's HR department via email and system messages, along with suggestions for improvement. S4, Claims Anti-Fraud Collaboration S4.1 Group Claims Model Knowledge Base Construction: In conjunction with cooperating insurance companies, medical examination institutions, and third-party credit reporting agencies, collect data on group life insurance claims cases that have been settled by each institution; collect data on newly settled cases from cooperating institutions in the previous month before the 10th of each month, extract features and add them to the knowledge base, and delete old data that has been in the database for more than 5 years. S4.2 Collaborative Analysis Steps: After receiving a new claim application, the cooperating insurance company receives data through an encrypted interface, including the insurance certificate number, effective date, date of the incident, employee identification code list, disease diagnosis, claim amount, employee's start date, and current number of employees. The system queries the blockchain insurance information based on the certificate number to obtain the employee identification code list, number of employees at the time of insurance application, and effective date. It verifies whether the employee involved in the incident is in the list at the time of insurance application; if they do not match, it sends an anomaly alert requesting verification. After the data verification is successful, the system extracts features from the claim application data, and the extracted features are consistent with the case features in the knowledge base. S4.3 Pattern Matching and Early Warning: The features of new claim applications are constructed into a feature map; the GraphSAGE graph neural network model is used as input, the new feature map and all fraud pattern maps in the knowledge base are input, the feature relationship between nodes and edges is learned, and the similarity score (0-100 points) is calculated; 80 points is set as the threshold, a score >80 points is marked as high risk, 60-80 points is marked as medium risk (both trigger an early warning), and <60 points is marked as low risk (no early warning). Generate early warning information, including similarity scores, matched fraud pattern types, and high-risk feature descriptions; push the early warning information to the claims review department of the insurance company that submitted the application via an encrypted message interface, and record it in the system log; if the insurance company confirms that the claim case involves fraud, mark the feature map of the case as a new fraud pattern and add it to the knowledge base.

9. The intelligent group life insurance management method based on dynamic group profiling according to claim 8, characterized in that: As described in S4.3, if the insurance company confirms that there is no fraud in the claim case, it will mark the feature map of the case as normal mode for use in optimizing the GraphSAGE model parameters.

Images