Health insurance risk control method and system based on big data

Abstract

The present invention provides a health insurance risk control method and system based on big data, which relates to the field of big data health, including obtaining basic attribute information and medical record information of a target object, performing data preprocessing on the basic attribute information and the medical record information, and extracting basic attribute features corresponding to the basic attribute information after data preprocessing and medical record features corresponding to the medical record information; determining a personal risk score corresponding to the target object through a pre-constructed risk assessment model according to the basic attribute features and the medical record features, and determining an insurance risk level corresponding to the target object based on the personal risk score; determining an average claim cost corresponding to the insurance risk level through the insurance risk level, and determining the insurance product rate corresponding to the target object in combination with the commercial cost of a candidate insurance product corresponding to the insurance risk level.

Description

Technical Field

[0001] The present invention relates to big data health technology, and in particular to a health insurance risk control method and system based on big data. Background Art

[0002] In the process of promoting insurance business, salesmen can usually only recommend insurance products to customers based on their experience. However, due to the different personal situations of social personnel and the wide variety of insurance products, each insurance product has different product terms and applicable scenarios. It is difficult for salesmen to determine the insurance products suitable for customers through self-analysis, and the effect of business promotion is not good.

[0003] In the current practice of recommending suitable insurance products to users, sales managers usually communicate with users to understand their personal information and then manually determine suitable insurance products for them to make personalized recommendations. However, when the suitable insurance products are manually determined, they are greatly influenced by subjective factors such as human experience, and it is inevitable that imperfect decisions will occur, resulting in low efficiency and accuracy. Summary of the invention

[0004] The embodiments of the present invention provide a health insurance risk control method and system based on big data, which can at least solve some problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] A health insurance risk control method based on big data is provided, comprising:

[0007] Acquire basic attribute information and medical record information of the target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information after data preprocessing and medical record features corresponding to the medical record information;

[0008] According to the basic attribute characteristics and the medical record characteristics, a personal risk score corresponding to the target object is determined through a pre-constructed risk assessment model, and based on the personal risk score, an insurance risk level corresponding to the target object is determined, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism;

[0009] The average claim cost corresponding to the insurance risk level is determined through the insurance risk level, and the insurance product rate corresponding to the target object is determined in combination with the commercial cost of the candidate insurance products corresponding to the insurance risk level.

[0010] In an optional embodiment,

[0011] Before determining the personal risk score corresponding to the target object by using a pre-constructed risk assessment model according to the basic attribute characteristics and the medical record characteristics, the method further includes training the risk assessment model:

[0012] Initializing multiple hyperparameters of the risk assessment model, setting initial parameter ranges for the multiple hyperparameters, and randomly combining the multiple hyperparameters through random search to determine an initial hyperparameter combination;

[0013] Determine the parameter performance of each initial hyperparameter combination through cross-validation, and take the initial hyperparameter combinations with the top 20% of parameter performance as the hotspot hyperparameter combinations;

[0014] The prior distribution corresponding to the hotspot hyperparameter combination is determined by a Gaussian function, and the search step size is iteratively adjusted. The new parameters searched each time are added to the observation parameter set. Based on the observation parameter set and the risk assessment objective function of the risk assessment model, the posterior distribution corresponding to the hotspot hyperparameter combination is determined, and the search step size is iteratively adjusted repeatedly until a preset number of iterations is met, wherein the risk assessment objective function is used to indicate the error between the risk assessment value output by the risk assessment model and the risk label value.

[0015] In an optional embodiment,

[0016] The risk assessment objective function is shown in the following formula:

[0017] TAR=α·Acc(θ)-λ·TrainTime(θ)-(1-α-λ)Cmp(θ);

[0018] Among them, TAR represents the target value of the risk assessment objective function, α and λ represent the accuracy weight and training time weight respectively, θ represents the hotspot hyperparameter combination, Acc(θ), TrainTime(θ), and Cmp(θ) represent the verification accuracy function, training time function, and complexity function respectively under a given hotspot hyperparameter combination.

[0019] In an optional embodiment,

[0020] Determining the average claim cost corresponding to the insurance risk level by the insurance risk level, and determining the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level includes:

[0021] Determining, by using the insurance risk level, an average claim cost corresponding to the insurance risk level;

[0022] Determine the pure risk premium corresponding to the insurance risk level based on the claim probability corresponding to the insurance risk level and the average claim cost;

[0023] The insurance product rate corresponding to the target object is determined based on the pure risk premium, combined with the commercial cost of the candidate insurance product corresponding to the insurance risk level, and the expected profit ratio corresponding to the candidate insurance product.

[0024] In an optional embodiment,

[0025] The insurance product rate corresponding to the target object is determined as shown in the following formula:

[0026] R=(P×q×(1+C))×(1+L)×F;

[0027] Among them, R represents the insurance product rate, P and q represent the average claim cost and claim probability respectively, C represents the operating cost ratio, L represents the expected profit ratio, and F represents the commercial cost of the candidate insurance product.

[0028] In an optional embodiment,

[0029] After determining the insurance product rate corresponding to the target object, the method further includes:

[0030] Taking the insurance product rate corresponding to the target object as the target insurance product rate, determining multiple insurance product combinations that match the insurance product rate, and searching for a candidate insurance product that is closest to the target insurance product rate from the multiple insurance product combinations;

[0031] If the premium cap of the candidate insurance product reaches or exceeds the target insured amount corresponding to the insurance risk level, then it is determined as an insurance product that meets the target insured amount requirements;

[0032] If the premium cap of the candidate insurance product is lower than the target insurance amount corresponding to the insurance risk level, then the amount difference between the two is calculated, and a secondary insurance product that can best make up the amount difference is found among all the candidate insurance products;

[0033] If the premium cap of the secondary insurance product is equal to or exceeds the amount difference, the candidate insurance product is combined with the secondary insurance product to determine it as an insurance product that meets the target insurance amount requirement;

[0034] If the premium cap of the secondary insurance product is lower than the amount gap, the amount gap will continue to be calculated, and an insurance product that can best fill the amount gap will be sought until the premium cap meets or exceeds the target insurance amount.

[0035] According to a second aspect of the embodiments of the present invention,

[0036] Provided is a health insurance risk control system based on big data, including:

[0037] The first unit is used to obtain basic attribute information and medical record information of a target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information and medical record features corresponding to the medical record information after data preprocessing;

[0038] The second unit is used to determine the personal risk score corresponding to the target object according to the basic attribute characteristics and the medical record characteristics through a pre-constructed risk assessment model, and determine the insurance risk level corresponding to the target object based on the personal risk score, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism;

[0039] The third unit is used to determine the average claim cost corresponding to the insurance risk level through the insurance risk level, and determine the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level.

[0040] According to a third aspect of the embodiments of the present invention,

[0041] An electronic device is provided, comprising:

[0042] processor;

[0043] a memory for storing processor-executable instructions;

[0044] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0045] A fourth aspect of the embodiments of the present invention is:

[0046] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0047] The prior distribution determined by the Gaussian function used in this application can provide a reasonable starting point based on historical data for the search of hyperparameters. This method assumes that the effective value of the hyperparameter follows a Gaussian distribution, which helps to start the search from the area with higher probability, rather than blindly searching the entire parameter space; by adjusting the search step size in each iteration, the search range can be gradually narrowed, thereby focusing on the most promising hyperparameter combination more quickly. Adding the new parameters searched each time to the set of observed parameters allows the model to dynamically learn from new data, so that the hyperparameter optimization process can continuously adjust its search strategy based on the latest information, increasing the probability of finding the optimal solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of a health insurance risk control method based on big data according to an embodiment of the present invention;

[0049] Figure 2 It is a schematic diagram of the structure of a health insurance risk control system based on big data according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0052] Figure 1 FIG. 1 is a flow chart of a health insurance risk control method based on big data according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0053] S101. Obtain basic attribute information and medical record information of a target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information and medical record features corresponding to the medical record information after data preprocessing;

[0054] Exemplarily, the basic attribute information of the target object may include age, gender, occupation, place of residence, etc., and the medical record information of the target object may include medical history information, diagnosis results, treatment history, drug use records, etc. Check data quality, handle missing values, outliers and duplicate records, ensure that all data formats are unified, and standardize numerical data, such as converting height and weight into a unified unit of measurement; convert non-numeric categorical data into numerical data, such as using one-hot encoding.

[0055] Extract features from preprocessed basic attribute information, such as age group, occupation type, etc. Analyze medical record data to extract key features, such as common disease types, treatment response patterns, drug sensitivity, etc. Integrate the extracted basic attribute features and medical record features into a unified data set for subsequent analysis or model training.

[0056] Assuming the target group is a group of diabetic patients, the steps are as follows:

[0057] Data collection: The patients’ age, sex, weight, height, and their medical records, including blood glucose levels, HbA1c (glycosylated hemoglobin) results, and glucose-lowering medications used, were collected;

[0058] Data preprocessing: cleaning the data, such as excluding incomplete medical records, standardizing the measurement units of height and weight, and encoding gender as a binary variable;

[0059] Feature extraction: extract age group and BMI (body mass index) from basic attributes, and extract features such as blood sugar control and drug response from medical records;

[0060] Data integration: Integrate the extracted features into a data set in preparation for further statistical analysis or machine learning model training to predict efficacy or discover treatment patterns, so as to provide corresponding reference information for subsequent insurance product recommendations.

[0061] S102. According to the basic attribute characteristics and the medical record characteristics, a personal risk score corresponding to the target object is determined by a pre-constructed risk assessment model, and based on the personal risk score, an insurance risk level corresponding to the target object is determined, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism;

[0062] Exemplarily, a risk assessment model constructed based on an improved gradient boosting machine (GBM) is used to determine the personal risk score of a target object and to determine its insurance risk level accordingly. The present application makes the following improvements over the traditional gradient boosting machine: Based on the traditional gradient boosting machine, the present application determines the final hyperparameter combination by combining multiple hyperparameter search algorithms, so that the trained risk assessment model can perform risk assessment more accurately.

[0063] Specifically, the basic attribute characteristics and medical record characteristics of the target object are input into the risk assessment model. The model calculates a personal risk score based on the input characteristics. The score reflects the potential risk of the target object in insurance. The risk score is interpreted according to the pre-defined standards and converted into a specific insurance risk level. The risk score is divided into different levels, such as low risk, medium risk and high risk. Each level corresponds to different insurance needs and pricing strategies. According to the determined risk level, the corresponding insurance products and pricing are formulated. Provide more comprehensive protection for high-risk individuals, and provide more economical options for low-risk individuals.

[0064] In an optional embodiment,

[0065] Before determining the personal risk score corresponding to the target object by using a pre-constructed risk assessment model according to the basic attribute characteristics and the medical record characteristics, the method further includes training the risk assessment model:

[0066] Initializing multiple hyperparameters of the risk assessment model, setting initial parameter ranges for the multiple hyperparameters, and randomly combining the multiple hyperparameters through random search to determine an initial hyperparameter combination;

[0067] Determine the parameter performance of each initial hyperparameter combination through cross-validation, and take the initial hyperparameter combinations with the top 20% of parameter performance as the hotspot hyperparameter combinations;

[0068] The prior distribution corresponding to the hotspot hyperparameter combination is determined by a Gaussian function, and the search step size is iteratively adjusted. The new parameters searched each time are added to the observation parameter set. Based on the observation parameter set and the risk assessment objective function of the risk assessment model, the posterior distribution corresponding to the hotspot hyperparameter combination is determined, and the search step size is iteratively adjusted repeatedly until a preset number of iterations is met, wherein the risk assessment objective function is used to indicate the error between the risk assessment value output by the risk assessment model and the risk label value.

[0069] Exemplarily, an initial parameter range is set for each hyperparameter of the risk assessment model (such as learning rate, number and depth of trees, regularization parameter, etc.), and a hyperparameter combination is randomly selected within the set parameter range through random search, which helps to quickly explore a large range of parameter space and find an effective initial point. Each randomly selected hyperparameter combination is cross-validated to determine its performance. The performance indicator can be the accuracy of the model, F1 score, mean square error, etc. The top 20% of the hyperparameter combinations in performance are selected as hot combinations for subsequent optimization.

[0070] Use Gaussian function to determine the prior distribution for these hot spot combinations. The prior distribution reflects the impact of each parameter on the model performance. In each iteration, the search step size is adjusted based on the prior distribution to find new hyperparameter combinations. Every time a new combination is found, it is added to the observed parameter set. The observed parameter set is used to indicate the hyperparameter combinations that have been evaluated in the hyperparameter optimization process and their corresponding performance indicators. In the training process of machine learning models, hyperparameters are those parameters set before the learning process begins, such as learning rate, tree depth (for tree models) or the number of hidden layers (for neural networks). Each specific set of hyperparameters will have an impact on the training and final performance of the model.

[0071] Using the risk assessment objective function and the observed parameter performance, the posterior distribution is updated, which helps understand the impact of different parameters on the prediction error. The iterative search for new hyperparameter combinations continues according to the posterior distribution. The iterative process continues until the preset number of iterations is reached, and the hyperparameter combination with the best performance in the final iteration is selected as the final configuration of the model. The risk assessment model is trained using the determined optimal hyperparameter combination.

[0072] For example, the prior distribution determined by the Gaussian function can provide a reasonable starting point for the search of hyperparameters based on historical data. This method assumes that the effective value of the hyperparameter follows a Gaussian distribution, which helps to start the search from the area with higher probability instead of blindly searching the entire parameter space; by adjusting the search step size in each iteration, the search range can be gradually narrowed, so as to focus on the most promising hyperparameter combination more quickly. Adding the new parameters searched each time to the set of observed parameters allows the model to dynamically learn from new data. In this way, the hyperparameter optimization process can continuously adjust its search strategy based on the latest information, increasing the probability of finding the optimal solution.

[0073] In addition, by observing the parameter set and the objective function feedback, the posterior distribution of the hotspot hyperparameter combination is continuously updated, providing a more accurate estimate of the parameter effect. This dynamic update mechanism makes the posterior distribution closer to the actual parameter performance. By precisely controlling the search step size and using the information of the posterior distribution, the optimal hyperparameter combination can be found more quickly, thereby accelerating the model convergence.

[0074] In an optional embodiment,

[0075] The risk assessment objective function is shown in the following formula:

[0076] TAR=α·Acc(θ)-λ·TrainTime(θ)-(1-α-λ)Cmp(θ);

[0077] Among them, TAR represents the target value of the risk assessment objective function, α and λ represent the accuracy weight and training time weight respectively, θ represents the hotspot hyperparameter combination, Acc(θ), TrainTime(θ), and Cmp(θ) represent the verification accuracy function, training time function, and complexity function respectively under a given hotspot hyperparameter combination.

[0078] For example, by weighing the accuracy, training time and complexity of the model, the formula supports the optimization of multiple objectives rather than a single indicator, which is more in line with the needs of practical applications; by adjusting α and λ, the importance of accuracy and training time in model optimization can be adjusted according to specific needs, thereby realizing a customized model optimization strategy.

[0079] S103. Determine the average claim cost corresponding to the insurance risk level through the insurance risk level, and determine the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level.

[0080] Exemplarily, historical claims data of similar risk levels to the target object are collected and analyzed, and the average claim cost at that risk level is calculated, which is usually obtained by calculating the total amount of all claims at a specific risk level and dividing it by the number of claim cases. The business costs associated with providing insurance products are evaluated, including management, operation, sales and marketing expenses, and the average claim cost is combined with the business cost to obtain the total cost. The insurance premium rate of the target object is determined based on the total cost, which usually involves adding a predetermined profit margin to the total cost.

[0081] In an optional embodiment,

[0082] Determining the average claim cost corresponding to the insurance risk level by the insurance risk level, and determining the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level includes:

[0083] Determining, by using the insurance risk level, an average claim cost corresponding to the insurance risk level;

[0084] Determine the pure risk premium corresponding to the insurance risk level based on the claim probability corresponding to the insurance risk level and the average claim cost;

[0085] The insurance product rate corresponding to the target object is determined based on the pure risk premium, combined with the commercial cost of the candidate insurance product corresponding to the insurance risk level, and the expected profit ratio corresponding to the candidate insurance product.

[0086] For example, pure risk premium: the portion of the insurance premium set to cover the expected claim costs after considering the claim probability and the average claim cost; expected profit ratio: the profit margin set by the insurance company to achieve profitability;

[0087] Among them, historical claim data under the same risk level is analyzed, and the frequency of claims occurring at this risk level is calculated. The claim probability can be determined by the ratio of the number of claim cases at a specific risk level to the total number of policies at a specific risk level; the pure risk premium can be determined by the product of the claim probability and the average claim cost.

[0088] In an optional embodiment,

[0089] The insurance product rate corresponding to the target object is determined as shown in the following formula:

[0090] R=(P×2×(1+C))×(1+L)×F;

[0091] Among them, R represents the insurance product rate, P and q represent the average claim cost and claim probability respectively, C represents the operating cost ratio, L represents the expected profit ratio, and F represents the commercial cost of the candidate insurance product.

[0092] In an optional embodiment,

[0093] After determining the insurance product rate corresponding to the target object, the method further includes:

[0094] Taking the insurance product rate corresponding to the target object as the target insurance product rate, determining multiple insurance product combinations that match the insurance product rate, and searching for a candidate insurance product that is closest to the target insurance product rate from the multiple insurance product combinations;

[0095] If the premium cap of the candidate insurance product reaches or exceeds the target insured amount corresponding to the insurance risk level, then it is determined as an insurance product that meets the target insured amount requirements;

[0096] If the premium cap of the candidate insurance product is lower than the target insurance amount corresponding to the insurance risk level, then the amount difference between the two is calculated, and a secondary insurance product that can best make up the amount difference is found among all the candidate insurance products;

[0097] If the premium cap of the secondary insurance product is equal to or exceeds the amount difference, the candidate insurance product is combined with the secondary insurance product to determine it as an insurance product that meets the target insurance amount requirement;

[0098] If the premium cap of the secondary insurance product is lower than the amount gap, the amount gap will continue to be calculated, and an insurance product that can best fill the amount gap will be sought until the premium cap meets or exceeds the target insurance amount.

[0099] Exemplarily, the target insurance product rate is determined based on the insurance product rate corresponding to the target object calculated previously; all available insurance products and their rates are listed; the product closest to the target insurance product rate is selected from the candidate products; and it is checked whether the premium cap of the selected candidate insurance product reaches or exceeds the target insurance amount corresponding to the target insurance risk level;

[0100] If the premium cap meets the requirements, the product is identified as an insurance product that meets the target insurance amount requirements;

[0101] If the premium cap is lower than the target insurance amount, calculate the difference between the two and find the secondary insurance product that can best make up the difference among all candidate insurance products; if the premium cap of the secondary insurance product is equal to or exceeds the difference, combine it with the candidate insurance product.

[0102] By accurately calculating and matching insurance products, we can provide customers with more personalized insurance solutions and ensure that their specific risks are appropriately covered; by considering premium caps and insurance sum insured, we can help insurance companies find a balance between taking risks and maintaining cost-effectiveness; ensure that the insurance products provided meet the customer's expected insurance sum insured, thereby improving customer satisfaction and loyalty; by considering different insurance product combinations, we can provide customers with a variety of choices, allowing them to choose the most suitable insurance products based on their needs and budgets; ensure that the insurance products provided will neither lead to over-protection (thereby increasing unnecessary costs) nor under-protection (thereby leaving risks uncovered).

[0103] Figure 2 FIG. 1 is a schematic diagram of the structure of a health insurance risk control system based on big data according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0104] The first unit is used to obtain basic attribute information and medical record information of a target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information and medical record features corresponding to the medical record information after data preprocessing;

[0105] The second unit is used to determine the personal risk score corresponding to the target object according to the basic attribute characteristics and the medical record characteristics through a pre-constructed risk assessment model, and determine the insurance risk level corresponding to the target object based on the personal risk score, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism;

[0106] The third unit is used to determine the average claim cost corresponding to the insurance risk level through the insurance risk level, and determine the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level.

[0107] According to a third aspect of the embodiments of the present invention,

[0108] An electronic device is provided, comprising:

[0109] processor;

[0110] a memory for storing processor-executable instructions;

[0111] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0112] A fourth aspect of the embodiments of the present invention is:

[0113] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0114] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A health insurance risk control method based on big data, characterized in that: include: Acquire basic attribute information and medical record information of the target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information after data preprocessing and medical record features corresponding to the medical record information; According to the basic attribute characteristics and the medical record characteristics, a personal risk score corresponding to the target object is determined through a pre-constructed risk assessment model, and based on the personal risk score, an insurance risk level corresponding to the target object is determined, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism; Determine the average claim cost corresponding to the insurance risk level through the insurance risk level, and determine the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level; The risk assessment objective function is shown in the following formula: TAR=α·Acc(θ)-λ·TrainTime(θ)-(1-α-λ)Cmp(θ); Wherein, TAR represents the target value of the risk assessment objective function, α and λ represent the accuracy weight and the training time weight respectively, θ represents the hotspot hyperparameter combination, Acc(θ), TrainTime(θ), and Cmp(θ) represent the verification accuracy function, training time function, and complexity function respectively under a given hotspot hyperparameter combination; Determining the average claim cost corresponding to the insurance risk level by the insurance risk level, and determining the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level includes: Determining, by using the insurance risk level, an average claim cost corresponding to the insurance risk level; Determine the pure risk premium corresponding to the insurance risk level based on the claim probability corresponding to the insurance risk level and the average claim cost; Determine the insurance product rate corresponding to the target object according to the pure risk premium, combined with the commercial cost of the candidate insurance product corresponding to the insurance risk level, and the expected profit ratio corresponding to the candidate insurance product; The insurance product rate corresponding to the target object is determined as shown in the following formula: R=(P×q×(1+C))×(1+L)×F; Among them, R represents the insurance product rate, P and q represent the average claim cost and claim probability respectively, C represents the operating cost ratio, L represents the expected profit ratio, and F represents the commercial cost of the candidate insurance product; After determining the insurance product rate corresponding to the target object, the method further includes: Taking the insurance product rate corresponding to the target object as the target insurance product rate, determining multiple insurance product combinations that match the insurance product rate, and searching for a candidate insurance product that is closest to the target insurance product rate from the multiple insurance product combinations; If the premium cap of the candidate insurance product reaches or exceeds the target insured amount corresponding to the insurance risk level, then it is determined as an insurance product that meets the target insured amount requirements; If the premium cap of the candidate insurance product is lower than the target insurance amount corresponding to the insurance risk level, then the amount difference between the two is calculated, and a secondary insurance product that can best make up the amount difference is found among all the candidate insurance products; If the premium cap of the secondary insurance product is equal to or exceeds the amount difference, the candidate insurance product is combined with the secondary insurance product to determine it as an insurance product that meets the target insurance amount requirement; If the premium cap of the secondary insurance product is lower than the amount gap, the amount gap will continue to be calculated, and an insurance product that can best fill the amount gap will be sought until the premium cap meets or exceeds the target insurance amount.

2. The method according to claim 1, characterized in that Before determining the personal risk score corresponding to the target object by using a pre-constructed risk assessment model according to the basic attribute characteristics and the medical record characteristics, the method further includes training the risk assessment model: Initializing multiple hyperparameters of the risk assessment model, setting initial parameter ranges for the multiple hyperparameters, and randomly combining the multiple hyperparameters through random search to determine an initial hyperparameter combination; Determine the parameter performance of each initial hyperparameter combination through cross-validation, and take the initial hyperparameter combinations with the top 20% of parameter performance as the hotspot hyperparameter combinations; The prior distribution corresponding to the hotspot hyperparameter combination is determined by a Gaussian function, and the search step size is iteratively adjusted. The new parameters searched each time are added to the observation parameter set. Based on the observation parameter set and the risk assessment objective function of the risk assessment model, the posterior distribution corresponding to the hotspot hyperparameter combination is determined, and the search step size is iteratively adjusted repeatedly until a preset number of iterations is met, wherein the risk assessment objective function is used to indicate the error between the risk assessment value output by the risk assessment model and the risk label value.

3. A health insurance risk control system based on big data, used to implement the health insurance risk control method based on big data as described in any one of claims 1-2, characterized in that: include: The first unit is used to obtain basic attribute information and medical record information of a target object, perform data preprocessing on the basic attribute information and the medical record information, and extract basic attribute features corresponding to the basic attribute information and medical record features corresponding to the medical record information after data preprocessing; The second unit is used to determine the personal risk score corresponding to the target object according to the basic attribute characteristics and the medical record characteristics through a pre-constructed risk assessment model, and determine the insurance risk level corresponding to the target object based on the personal risk score, wherein the risk assessment model is constructed based on an improved gradient boosting mechanism; The third unit is used to determine the average claim cost corresponding to the insurance risk level through the insurance risk level, and determine the insurance product rate corresponding to the target object in combination with the commercial cost of the candidate insurance product corresponding to the insurance risk level.

4. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 2.

5. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 2 is implemented.

Images