Insurance risk assessment method and device, training method and device, equipment and storage medium

Abstract

The invention relates to the technical field of data processing and risk assessment, and provides an insurance risk assessment method, a training method, a training device, equipment and a storage medium, and the method comprises the steps: obtaining health state data of a user, the health state data comprising multiple types of archive data; extracting features corresponding to the health state data based on a preset feature extraction network; based on a preset portrait construction model, fusing features corresponding to the health state data to obtain a health portrait of the user; and based on a preset risk assessment model, according to the health portrait of the user, assessing the risk that the user purchases the life insurance to obtain an assessment result. According to the method, the specific health state of the user is combined, so that the accuracy of insurance risk assessment is improved. By taking a medical scene as an example, physiological indexes of a user and health status data such as medical images and blood test reports in a medical institution can be obtained, and different types of health status data are fused, so that the comprehensiveness of risk assessment is improved.

Description

Technical Field

[0001] This application relates to the fields of data processing and risk assessment technology, and in particular to insurance risk assessment methods, training methods, devices, equipment and storage media. Background Technology

[0002] Conducting risk assessments for users purchasing insurance to determine reasonable premiums is a core part of an insurance company's business. For example, life insurance risk assessment involves evaluating a user's risk of death to determine the premium for that life insurance policy. Therefore, different users will have different risk assessment results, and consequently, different premiums.

[0003] The relevant technologies do not take into account the user's specific circumstances to conduct risk assessments, resulting in inaccurate risk assessments for users purchasing life insurance and causing users to bear unreasonable insurance costs. Summary of the Invention

[0004] The main purpose of this application is to provide an insurance risk assessment method, training method, device, equipment, and storage medium that incorporates the user's specific health status to improve the accuracy of insurance risk assessment.

[0005] Firstly, this application provides an insurance risk assessment method, including:

[0006] Acquire user health status data, which includes various types of profile data;

[0007] Based on a preset feature extraction network, features corresponding to the health status data are extracted;

[0008] Based on a preset profile building model, the features corresponding to the health status data are fused to obtain the user's health profile.

[0009] Based on a preset risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment result is obtained.

[0010] Secondly, this application also provides a training method for an insurance risk assessment model, wherein the insurance risk assessment model includes a feature extraction network, a profile construction model, and a risk assessment model; the training method includes:

[0011] Obtain a training sample set, which includes user health status data and corresponding evaluation result labels for the health status data. The health status data includes various types of profile data.

[0012] Based on the feature extraction network, features corresponding to the health status data are extracted;

[0013] Based on the profile, a model is constructed, and the features corresponding to the health status data are fused to obtain the user's health profile;

[0014] Based on the risk assessment model, the risk of the user purchasing life insurance is assessed according to the user's health profile, and the assessment result is obtained.

[0015] Based on a preset loss function, the loss value is determined according to the evaluation result label and the evaluation result;

[0016] Based on the loss value, at least the model parameters of the risk assessment model should be adjusted.

[0017] Thirdly, this application also provides an insurance risk assessment device, comprising:

[0018] The acquisition module is used to acquire the user's health status data, which includes various types of profile data;

[0019] The extraction module is used to extract features corresponding to the health status data based on a preset feature extraction network;

[0020] The construction module is used to build a model based on a preset profile, and to fuse the features corresponding to the health status data to obtain the user's health profile.

[0021] The assessment module is used to assess the risk of the user purchasing life insurance based on a preset risk assessment model and the user's health profile, and obtain the assessment result.

[0022] Fourthly, this application also provides a computer device, the computer device including a memory and a processor;

[0023] The memory is used to store computer programs;

[0024] The processor is used to execute the computer program and, in executing the computer program, implement the insurance risk assessment method and the training method for the insurance risk assessment model as described above.

[0025] Fifthly, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the insurance risk assessment method and the training method of the insurance risk assessment model as described above.

[0026] This application provides an insurance risk assessment method, training method, device, equipment, and storage medium. The method includes: acquiring a user's health status data, which includes various types of archival data; extracting features corresponding to the health status data based on a preset feature extraction network; fusing the features corresponding to the health status data based on a preset profile construction model to obtain a health profile of the user; and assessing the risk of the user purchasing life insurance based on the user's health profile using a preset risk assessment model, obtaining an assessment result. This application uses a feature extraction network to extract different types of features corresponding to the user's health status data; and uses a profile construction model to fuse different types of features to obtain a health profile of the user; and assesses the risk of the user purchasing life insurance based on the user's health profile, combining the user's specific health status to improve the accuracy of insurance risk assessment. Taking a medical scenario as an example, it can acquire the user's physiological indicators and health status data such as medical images and blood test reports from medical institutions, and fuse different types of health status data to improve the comprehensiveness of risk assessment. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart illustrating an insurance risk assessment method provided in this application embodiment;

[0029] Figure 2 This is a schematic diagram illustrating the connection between the server and the terminal device provided in an embodiment of this application;

[0030] Figure 3 A flowchart illustrating a training method for an insurance risk assessment model provided in an embodiment of this application;

[0031] Figure 4 A schematic block diagram of an insurance risk assessment device provided in this application embodiment;

[0032] Figure 5 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0033] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0034] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the described order. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0035] This application provides an insurance risk assessment method, training method, apparatus, device, and storage medium. The insurance risk assessment method can be applied to terminal devices, such as mobile phones, tablets, laptops, and desktop computers. It can also be applied to servers, which can be standalone servers or cloud servers providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0036] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0037] Please see Figure 1 , Figure 1 This is a flowchart illustrating an insurance risk assessment method provided in an embodiment of this application. It should be noted that the insurance risk assessment method provided in this embodiment can be used on terminal devices, and of course, it can also be used on servers.

[0038] like Figure 2 As shown, the insurance risk assessment method is applied to the server. The server and terminal devices communicate with each other, and the server can send the assessment results obtained through the insurance risk assessment method to the terminal devices. Of course, it is not limited to this, and no restrictions are imposed here.

[0039] In practice, terminal devices include, but are not limited to, any of the following: mobile phones, tablets, laptops, and desktop computers; servers can be standalone servers, server clusters, or cloud servers that provide cloud computing services.

[0040] like Figure 1As shown, the insurance risk assessment method includes steps S101 to S104.

[0041] Step S101: Obtain the user's health status data, which includes various types of profile data.

[0042] Generally, before a user purchases life insurance, the insurance company may request relevant health data from the user. Alternatively, the user may undergo a medical examination at a designated medical institution to obtain their health data.

[0043] For example, a user's health status data includes various types of archival data. These include electronic health record data from electronic health record systems in the medical field, and physiological indicators collected by wearable physiological devices. Electronic health record data can include X-ray images, CT images, MRI images, gene sequencing reports, blood test reports, and ultrasound images. Physiological indicators can include sleep depth, heart rate, and blood oxygenation.

[0044] This application embodiment also requires preprocessing of the health status data, such as removing abnormal values ​​and noise data from the health status data, and unifying different types of archive data in the health status data into the same format to facilitate subsequent feature extraction.

[0045] Step S102: Extract features corresponding to health status data based on a preset feature extraction network.

[0046] It is understood that health status data includes various types of archival data; that is, health status data can include data in different modalities such as images, text, audio, and video. In this application embodiment, different feature extraction networks can be used for different modalities of data to extract different types of features corresponding to the health status data.

[0047] For example, in the medical field, image features of medical images can be extracted based on convolutional neural networks or Transformer models; text features of medical image text reports and laboratory reports can be extracted based on word embedding models, convolutional neural networks, and recurrent neural networks; and video features such as vascular dynamic imaging videos and sleep monitoring videos can be extracted based on 3D convolutional neural networks.

[0048] Step S103: Based on the preset profile construction model, the features corresponding to the health status data are fused to obtain the user's health profile.

[0049] For example, the profile construction model in this application embodiment may include a Transformer model. It should be noted that the Transformer model captures long-term dependencies in a sequence through a self-attention mechanism, mainly including an encoder and a decoder, with each layer including a self-attention mechanism and a feedforward neural network. A health profile is a digital, personalized personal health record that integrates various aspects of health status data to form a comprehensive, intuitive, and dynamic health portrait. Generally, the health profile is displayed intuitively in the form of charts, dashboards, radar charts, etc., allowing people to clearly see their health status at a glance, such as scores in various dimensions and health levels. Therefore, a user's health profile allows for comprehensive monitoring of the user's physical condition.

[0050] This application embodiment can fuse features corresponding to health status data based on the Transformer model to obtain a user's health profile.

[0051] Specifically, the importance weights of different features or locations can be calculated based on a preset attention weight formula. The attention weight formula is:

[0052]

[0053] Where, α ij Here, represents the attention weights, i.e., the degree of attention the i-th query pays to the j-th key; softmax is an activation function used to transform the original scores into a probability distribution, ensuring that the sum of all weights is 1; Q i The i-th row of the query matrix Q represents the query vector for the feature or location to be processed; K j T is the transpose of the j-th column of the key matrix K, representing the key vector of the feature or position to be compared; √d represents the scaling factor, where d represents the dimension of the key vector. The scaling factor is used to prevent the dot product result from becoming too large, which could lead to gradient vanishing or exploding.

[0054] This application embodiment is based on the Transformer model to fuse features corresponding to the user's health status data, that is, features corresponding to different types of profile data, and calculates attention weights based on the attention weight formula, so that the model can focus on different parts of the sequence of input health status data, thereby capturing richer contextual information and improving the accuracy of the constructed user profile.

[0055] Step S104: Based on the preset risk assessment model and the user's health profile, assess the risk of the user purchasing life insurance and obtain the assessment results.

[0056] It is understood that, after constructing a user's health profile, this application embodiment can obtain the user's health status from the constructed user profile, and assess the risk of the user purchasing life insurance based on a preset risk assessment model, and obtain the assessment result.

[0057] The insurance risk assessment method provided in the above embodiments includes: acquiring a user's health status data, which includes various types of archival data; extracting features corresponding to the health status data based on a preset feature extraction network; fusing the features corresponding to the health status data based on a preset profile construction model to obtain a user's health profile; and assessing the risk of the user purchasing life insurance based on the user's health profile using a preset risk assessment model to obtain an assessment result. This embodiment of the application extracts different types of features corresponding to the user's health status data based on a feature extraction network; fuses different types of features based on a profile construction model to obtain a user's health profile; and assesses the risk of the user purchasing life insurance based on the user's health profile, taking into account the user's specific health status to improve the accuracy of insurance risk assessment. Taking a medical scenario as an example, it is possible to acquire the user's physiological indicators and health status data such as medical images and blood test reports from medical institutions, and fuse different types of health status data to improve the comprehensiveness of risk assessment.

[0058] In one exemplary embodiment, the health status data includes medical images and gene sequencing reports; the feature extraction network includes a convolutional neural network and a bidirectional long short-term memory network; step S102 includes steps S1021 to S1023.

[0059] Step S1021: Extract the first feature of the medical image based on the convolutional neural network.

[0060] Step S1022: Extract the second feature from the gene sequencing report based on the bidirectional long short-term memory network.

[0061] Step S1023: Map the first feature and the second feature to the same feature space for alignment to obtain the features corresponding to the health status data.

[0062] In a medical setting, this application embodiment can obtain health status data such as medical images (CT images, MRI images) and gene sequencing reports from the business management system of a medical institution, and use different feature extraction networks to extract features from the medical images and gene sequencing reports.

[0063] For example, image features (such as lung nodule volume and vascular stenosis) from medical images like CT and MRI images can be extracted based on convolutional neural networks, such as 3D convolutional neural networks. These features are the first features. Specifically, the image feature extraction formula is:

[0064] F image =3D-CNN(I CT I MRI );

[0065] Among them, F image For image features, I CT For CT images, I MRI This is an MRI image.

[0066] In another embodiment, a bidirectional long short-term memory network can be used to identify the gene characteristics of disease susceptibility sites in the gene sequencing report, i.e., the second feature. Specifically, the feature encoding formula is:

[0067] F genetic =BiLSTM(S) DNA );

[0068] Among them, F genetic Genetic characteristics of disease susceptibility loci, S DNA This is a gene sequencing report.

[0069] It is understood that the embodiments of this application can employ a contrastive learning method to map image features corresponding to medical images such as CT images and MRI images, as well as gene features of disease susceptibility sites, into the same feature space for alignment, thereby obtaining features corresponding to health status data. Specifically, after extracting image features and gene features, the image features and gene features are mapped into the same feature space; based on the semantic relevance of multiple features, positive sample pairs and negative sample pairs are constructed, where positive sample pairs contain features of the same semantic concept in different modalities, and negative sample pairs are features of randomly combined different semantic concepts; the similarity between the latent representations of different features is calculated; a contrastive loss function is designed, based on which the latent representations of positive sample pairs are made as close as possible, and the latent representations of negative sample pairs are made as far apart as possible; by continuously optimizing the contrastive loss function, the latent spaces of different features are aligned, thereby obtaining features corresponding to health status data.

[0070] The contrastive loss function used in the contrastive learning method is:

[0071] L=∑||F image (x)-F genetic (y)|| 2 ;

[0072] Where x is the input variable for image features and y is the input variable for gene features.

[0073] This application embodiment aligns the latent spaces of different features using a contrastive learning method, so that semantically related features are close to each other in the latent space, thereby extracting features corresponding to health status data that are more accurate and richer in information representation, thus improving the accuracy of subsequent assessments.

[0074] In one exemplary embodiment, the risk assessment model includes a decision model, a prediction model, and an evaluation model; step S104 includes steps S1041 to S1043.

[0075] Step S1041: Based on the decision model and the user's health profile, obtain the corresponding health management content for the user. The health management content is used to indicate ways to improve the user's health status.

[0076] Step S1042: Based on the prediction model, predict the user's corresponding health events according to the user's health profile.

[0077] Step S1043: Based on the assessment model, the risk of the user purchasing life insurance is assessed according to the user's health profile, the user's corresponding health management content, and health events, and the assessment results are obtained.

[0078] The risk assessment model in this application embodiment may include a decision-making model, a prediction model, and an evaluation model. The decision-making module may include a deep reinforcement learning model, which is used to develop personalized health management content aimed at improving the user's health status and reducing future health risks. Based on the deep reinforcement learning model and the user's health profile, corresponding health management content can be obtained. This health management content may include interventions such as dietary recommendations, exercise recommendations, and medication treatments, enabling users to improve their health status and thus reduce future health risks.

[0079] The reward function for the deep reinforcement learning model is:

[0080] R = L - λ *C;

[0081] Where R is the net risk reduction indicator, L is the risk reduction rate, λ is the weighting coefficient, and C is the health management cost.

[0082] Understandably, personalized health management content generated through deep reinforcement learning models can reduce the deterioration of users' health problems, thereby reducing the incidence of life insurance claims disputes; it can also optimize protection plans and reduce loss ratios, thereby increasing the profitability of insurance companies in the life insurance business.

[0083] For example, the prediction model in this application embodiment may include a time series Transformer model, namely the TST model, which is a variant of the Transformer architecture in the time series domain and is mainly used for time series prediction tasks. This application embodiment can first sort different modalities of data obtained from the user's health profile in chronological order to construct a time series dataset, and then align the different modalities of data along the time dimension so that the model can capture the correlation between different modalities. Then, the time series dataset aligned along the time dimension is input into the time series Transformer model to predict the user's corresponding health events. Specifically, it can predict health events that will occur at a preset time in the future, such as myocardial infarction and its probability, hospitalization and its probability. The preset expression for predicting health events in this application embodiment is as follows:

[0084] P event =σ*(W*h t +b);

[0085] Among them, P event Let σ be the probability of a health event occurring, W be the activation function, and h be the weight matrix. t Let b be the hidden state at time step t, and b be the bias term.

[0086] The time series Transformer model used in this application embodiment performs well in handling long sequence dependencies and capturing time patterns. When training the time series Transformer model, the model performance can be optimized by adjusting parameters such as learning rate, data size, and number of model layers. Furthermore, the model's ability to predict health events can be evaluated through metrics such as accuracy and recall.

[0087] It is understood that, in the embodiments of this application, the historical health management content output by the decision model and the historical health events output by the prediction model can be input into the initial evaluation model for training until the initial evaluation model is trained, so as to obtain an evaluation model with a high evaluation accuracy.

[0088] Specifically, static and dynamic data can be obtained from the user's health profile. For example, static data includes the user's basic information, past medical history, and gene sequencing reports; dynamic data includes changes in blood pressure, blood sugar, blood lipids, and liver and kidney function over time. Health management content, such as medication intake and exercise duration for treating related diseases, output from the decision-making model, can be obtained; and health events, such as sudden illnesses, hospitalizations, and their probabilities, output from the prediction model, can be obtained. Then, the obtained static and dynamic data, health management content, and health events are input into a trained evaluation model to assess the risk of the user purchasing life insurance and obtain the evaluation results.

[0089] This application embodiment uses a time series Transformer model to predict a user's health events at a preset future time, and combines this with health management content generated by a deep reinforcement learning model to dynamically adjust the risk assessment for the user's life insurance purchase, thereby improving the accuracy of the assessment.

[0090] In one exemplary embodiment, step S1041 may specifically include: based on the decision model, determining the user's corresponding health management content from a preset health knowledge graph according to the user's health profile.

[0091] For example, embodiments of this application can obtain health and wellness data (medications, exercise, acupuncture, massage, etc.) from various medical literature and health and wellness guidelines recommended by medical experts, and construct a health knowledge graph based on this data to improve the professionalism of the health knowledge graph, thereby providing health management content suitable for most users. The health knowledge graph can include multiple nodes, and each node includes entities such as diseases, symptoms, medications, and treatment plans, as well as the relationships between these entities.

[0092] This application embodiment uses a decision model to search the health knowledge graph based on the user's health profile to match health management content that is more relevant to the user's current health status. By combining the decision-making ability of the deep reinforcement learning model with the semantic association of the health knowledge graph, dynamic, personalized, and interpretable health management content can be generated, while continuously optimizing the user's long-term health.

[0093] In an exemplary embodiment, the evaluation model includes an input layer, a fusion layer, and an evaluation layer; step S1043 includes steps S10431 to S10433.

[0094] Step S10431: Based on the input layer, input the user's health profile, the user's corresponding health management content, and health events.

[0095] Step S10432: Based on the fusion layer, the health profile, health management content, and health events are fused to obtain fused features.

[0096] Step S10433: Based on the evaluation layer, output the risk assessment results of the user's purchase of life insurance according to the fusion characteristics.

[0097] Understandably, the user's health profile, the health management content output by the decision-making model, and the health events output by the prediction model are first characterized and then input into the input layer of the evaluation model. Next, based on the fusion layer, the health profile, health management content, and health events are concatenated, weighted, and summed to capture the relationships between the data; a self-attention mechanism is also used to capture the temporal dependencies between the data. Then, based on the evaluation layer, the risk level of the user purchasing life insurance under their current health state is assessed according to the fused features. Specifically, logistic regression or support vector machines can be used to assess the risk level, and the risk level can be updated in real time when the decision-making model outputs different health management content and the prediction model outputs different health events.

[0098] In one exemplary implementation, the assessment result includes the user's current risk level in the current health state; step S104 is followed by step S105.

[0099] Step S105: Determine the insurance premium corresponding to the current risk level based on the user's current health status and the preset insurance premium table; the insurance premium table includes the correspondence between different risk levels and insurance premiums.

[0100] For example, the assessment results may include the user's current risk level under their current health condition. Generally, different risk levels correspond to different insurance premiums. This application embodiment can first preset an insurance premium table to establish the correspondence between each risk level and insurance premium. By mapping the user's current risk level under their current health condition to the preset insurance premium table, the insurance premium corresponding to the current risk level can be obtained. By establishing the correspondence between different risk levels and different insurance premiums, the insurance premiums for different users purchasing life insurance can be determined more accurately, thus avoiding over-insurance or under-insurance premiums that could negatively impact the insurance company.

[0101] Please see Figure 3 , Figure 3 This is a flowchart illustrating a training method for an insurance risk assessment model provided in an embodiment of this application. The insurance risk assessment model in this embodiment includes a feature extraction network, a profile construction model, and a risk assessment model. The training method includes steps S201 to S206.

[0102] Step S201: Obtain the training sample set, which includes the user's health status data and the corresponding evaluation result labels. The health status data includes various types of profile data.

[0103] Step S202: Extract features corresponding to health status data based on the feature extraction network.

[0104] Step S203: Based on the profile, build a model and fuse the features corresponding to the health status data to obtain the user's health profile.

[0105] Step S204: Based on the risk assessment model and the user's health profile, assess the risk of the user purchasing life insurance and obtain the assessment results.

[0106] It should be noted that the relevant discussions of steps S201 to S204 can be referred to the relevant embodiments of the aforementioned insurance risk assessment method, and will not be repeated here.

[0107] Step S205: Based on the preset loss function, determine the loss value according to the evaluation result label and the evaluation result.

[0108] Step S206: Based on the loss value, adjust at least the model parameters of the risk assessment model.

[0109] It is understandable that after each training sample set is input into the insurance risk assessment model, the corresponding user's assessment result can be obtained. After calculating the difference between the assessment result and the assessment result label based on a preset loss function, the model parameters of the risk assessment model within the insurance risk assessment model can be updated to reduce the difference between the assessment result and the assessment result label, making the predicted value output by the insurance risk assessment model closer to the true value, thereby achieving the learning objective and improving the accuracy of the model's output assessment results. Specifically, in this embodiment, the difference between the assessment result and the assessment result label can be calculated based on a preset loss function to obtain the loss value corresponding to the insurance risk assessment model. Then, the model parameters of the risk assessment model within the insurance risk assessment model can be updated based on this loss value. Alternatively, the model parameters of the profile building model within the insurance risk assessment model can also be updated, enabling the profile building model to construct a more accurate user health profile.

[0110] The training method for the insurance risk assessment model provided in the above embodiments includes: acquiring a training sample set, which includes user health status data and corresponding assessment result labels, wherein the health status data includes various types of archive data; extracting features corresponding to the health status data based on a feature extraction network; constructing a model based on the user profile, fusing the features corresponding to the health status data to obtain a user's health profile; assessing the risk of the user purchasing life insurance based on the user's health profile using a risk assessment model, and obtaining an assessment result; determining a loss value based on a preset loss function, according to the assessment result labels and the assessment result; and adjusting at least the model parameters of the risk assessment model based on the loss value. This embodiment of the application improves the inference ability of the insurance risk assessment model by training it multiple times, and adjusting the model parameters of the risk assessment model in each training session, thereby enabling the insurance risk assessment model to output more accurate assessment results.

[0111] Please see Figure 4 , Figure 4 This is a schematic block diagram of an insurance risk assessment device provided in an embodiment of this application. The insurance risk assessment device can be configured in a server or terminal device to perform the aforementioned insurance risk assessment method.

[0112] like Figure 4 As shown, the insurance risk assessment device includes: an acquisition module 110, an extraction module 120, a construction module 130, and an assessment module 140.

[0113] The acquisition module 110 is used to acquire the user's health status data, which includes various types of profile data.

[0114] The extraction module 120 is used to extract features corresponding to health status data based on a preset feature extraction network.

[0115] The construction module 130 is used to build a model based on a preset profile, and to fuse the features corresponding to the health status data to obtain the user's health profile.

[0116] The assessment module 140 is used to assess the risk of a user purchasing life insurance based on a preset risk assessment model and the user's health profile, and obtain the assessment results.

[0117] In one exemplary embodiment, the risk assessment model includes a decision model, a prediction model, and an evaluation model; the evaluation module 140 includes a decision submodule, a prediction submodule, and an evaluation submodule.

[0118] The decision-making submodule is used to obtain the user's corresponding health management content based on the decision-making model and the user's health profile. The health management content is used to indicate ways to improve the user's health status.

[0119] The prediction submodule is used to predict the user's corresponding health events based on the prediction model and the user's health profile.

[0120] The assessment submodule is used to assess the risk of a user purchasing life insurance based on the assessment model, the user's health profile, the user's corresponding health management content, and health events, and to obtain the assessment results.

[0121] In one exemplary implementation, the decision submodule is specifically used to determine the corresponding health management content for a user from a preset health knowledge graph based on the user's health profile and the decision model.

[0122] In one exemplary implementation, the evaluation model includes an input layer, a fusion layer, and an evaluation layer; the evaluation submodule includes an input submodule, a fusion submodule, and an output submodule.

[0123] The input submodule is used to input the user's health profile, corresponding health management content, and health events based on the input layer.

[0124] The fusion submodule is used to fuse health profiles, health management content, and health events based on the fusion layer to obtain fused features.

[0125] The output submodule is used to output the risk assessment results for users purchasing life insurance based on the evaluation layer and the fusion characteristics.

[0126] In one exemplary implementation, the assessment results include the user's current risk level in their current health state; the device also includes a cost determination module.

[0127] The cost determination module is used to determine the insurance cost corresponding to the current risk level based on the user's current health status and the preset insurance cost table; the insurance cost table includes the correspondence between different risk levels and insurance costs.

[0128] In one exemplary embodiment, the health status data includes medical images and gene sequencing reports; the feature extraction network includes a convolutional neural network and a bidirectional long short-term memory network; the extraction module includes a first extraction submodule, a second extraction submodule, and an alignment submodule.

[0129] The first extraction submodule is used to extract the first feature of medical images based on a convolutional neural network.

[0130] The second extraction submodule is used to extract the second feature of the gene sequencing report based on the bidirectional long short-term memory network.

[0131] The alignment submodule is used to map the first feature and the second feature to the same feature space for alignment, so as to obtain the features corresponding to the health status data.

[0132] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and its modules and units can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0133] The method of this application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0134] For example, the above-described methods and apparatus can be implemented as a computer program that can run on a computer device.

[0135] Please see Figure 5 , Figure 5 This is a schematic block diagram illustrating the structure of a computer device provided in an embodiment of this application. The computer device may be a server or a terminal device.

[0136] like Figure 5 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a storage medium and internal memory.

[0137] The storage medium may store an operating system and a computer program. The computer program includes program instructions that, when executed, cause the processor to perform steps of any insurance risk assessment method and steps of a training method for the insurance risk assessment model.

[0138] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0139] Internal memory provides an environment for the execution of computer programs stored in storage media. When these computer programs are executed by a processor, the processor can perform the steps of any insurance risk assessment method and the steps of training an insurance risk assessment model.

[0140] This network interface is used for network communication, such as sending assigned tasks.

[0141] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0142] It should be understood that a processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other convertible logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0143] In one embodiment, the processor is configured to execute a computer program and, when executing the computer program, may perform the following steps:

[0144] Acquire users' health status data, which includes various types of profile data;

[0145] Based on a pre-defined feature extraction network, features corresponding to health status data are extracted.

[0146] Based on a pre-defined profile building model, the features corresponding to health status data are fused to obtain the user's health profile.

[0147] Based on a pre-set risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment results are obtained.

[0148] Correspondingly, the processor is used to execute computer programs and, while executing computer programs, can also perform the following steps:

[0149] Obtain a training sample set, which includes users’ health status data and corresponding assessment result labels for the health status data. The health status data includes various types of profile data.

[0150] Based on a feature extraction network, features corresponding to health status data are extracted;

[0151] Based on the profile model, the features corresponding to the health status data are fused to obtain the user's health profile;

[0152] Based on the risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment results are obtained.

[0153] Based on a preset loss function, the loss value is determined according to the evaluation result label and the evaluation result;

[0154] Based on the loss value, at least the model parameters of the risk assessment model should be adjusted.

[0155] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of insurance risk assessment described above can be referred to the corresponding process in the embodiments of the aforementioned insurance risk assessment method, and the specific training process of the insurance risk assessment model described above can be referred to the corresponding process in the embodiments of the aforementioned insurance risk assessment model training method, and will not be repeated here.

[0156] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps:

[0157] Acquire users' health status data, which includes various types of profile data;

[0158] Based on a pre-defined feature extraction network, features corresponding to health status data are extracted.

[0159] Based on a pre-defined profile building model, the features corresponding to health status data are fused to obtain the user's health profile.

[0160] Based on a pre-set risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment results are obtained.

[0161] Correspondingly, when a computer program is executed by a processor, it can also perform the following steps:

[0162] Obtain a training sample set, which includes users’ health status data and corresponding assessment result labels for the health status data. The health status data includes various types of profile data.

[0163] Based on a feature extraction network, features corresponding to health status data are extracted;

[0164] Based on the profile model, the features corresponding to the health status data are fused to obtain the user's health profile;

[0165] Based on the risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment results are obtained.

[0166] Based on a preset loss function, the loss value is determined according to the evaluation result label and the evaluation result;

[0167] Based on the loss value, at least the model parameters of the risk assessment model should be adjusted.

[0168] The computer-readable storage medium can be an internal storage unit of the computer device described in the foregoing embodiments, such as a hard disk or memory of the computer device. Alternatively, it can be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the computer device.

[0169] It should be noted that the functions or steps that can be achieved by the computer-readable storage medium described above can be referred to in the embodiments of the aforementioned insurance risk assessment method and the embodiments of the training method for the insurance risk assessment model.

[0170] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0171] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0172] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An insurance risk assessment method, characterized in that, include: Acquire user health status data, which includes various types of profile data; Based on a preset feature extraction network, features corresponding to the health status data are extracted; Based on a preset profile building model, the features corresponding to the health status data are fused to obtain the user's health profile. Based on a preset risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed, and the assessment result is obtained.

2. The insurance risk assessment method according to claim 1, characterized in that, The risk assessment model includes a decision-making model, a prediction model, and an assessment model; based on the preset risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed to obtain the assessment results, including: Based on the decision model, according to the user's health profile, the corresponding health management content for the user is obtained, and the health management content is used to indicate ways to improve the user's health status; Based on the prediction model, and according to the user's health profile, predict the user's corresponding health events; Based on the assessment model, the risk of the user purchasing life insurance is assessed according to the user's health profile, the user's corresponding health management content, and health events, and the assessment result is obtained.

3. The insurance risk assessment method according to claim 2, characterized in that, Based on the decision model and the user's health profile, the health management content corresponding to the user is obtained, including: Based on the decision-making model, and according to the user's health profile, the corresponding health management content for the user is determined from a preset health knowledge graph.

4. The insurance risk assessment method according to claim 2, characterized in that, The evaluation model includes an input layer, a fusion layer, and an evaluation layer. Based on the evaluation model, and according to the user's health profile, the user's corresponding health management content, and health events, the risk of the user purchasing life insurance is assessed to obtain the evaluation results, including: Based on the input layer, the user's health profile, the user's corresponding health management content, and health events are input; Based on the fusion layer, the health profile, the health management content, and the health events are fused to obtain fused features; Based on the evaluation layer and according to the fusion features, the risk assessment result of the user's purchase of life insurance is output.

5. The insurance risk assessment method according to any one of claims 1 to 4, characterized in that, The assessment results include the user's current risk level in their current health state; Based on a preset risk assessment model and the user's health profile, the risk of the user purchasing life insurance is assessed. After obtaining the assessment results, the process further includes: Based on the user's current risk level under their current health condition and a preset insurance premium table, determine the insurance premium corresponding to the current risk level; The insurance cost table includes the correspondence between different risk levels and insurance costs.

6. The insurance risk assessment method according to any one of claims 1 to 4, characterized in that, The health status data includes medical images and gene sequencing reports; the feature extraction network includes a convolutional neural network and a bidirectional long short-term memory network; the feature extraction based on the preset feature extraction network extracts features corresponding to the health status data, including: Based on the convolutional neural network, the first feature of the medical image is extracted; Based on the bidirectional long short-term memory network, the second feature of the gene sequencing report is extracted; The first feature and the second feature are mapped to the same feature space for alignment to obtain the features corresponding to the health status data.

7. A training method for an insurance risk assessment model, characterized in that, The insurance risk assessment model includes a feature extraction network, a profile construction model, and a risk assessment model; the training method includes: Obtain a training sample set, which includes user health status data and corresponding evaluation result labels for the health status data. The health status data includes various types of profile data. Based on the feature extraction network, features corresponding to the health status data are extracted; Based on the profile, a model is constructed, and the features corresponding to the health status data are fused to obtain the user's health profile; Based on the risk assessment model, the risk of the user purchasing life insurance is assessed according to the user's health profile, and the assessment result is obtained. Based on a preset loss function, the loss value is determined according to the evaluation result label and the evaluation result; Based on the loss value, at least the model parameters of the risk assessment model should be adjusted.

8. An insurance risk assessment device, characterized in that, include: The acquisition module is used to acquire the user's health status data, which includes various types of profile data; The extraction module is used to extract features corresponding to the health status data based on a preset feature extraction network; The construction module is used to build a model based on a preset profile, and to fuse the features corresponding to the health status data to obtain the user's health profile. The assessment module is used to assess the risk of the user purchasing life insurance based on a preset risk assessment model and the user's health profile, and obtain the assessment result.

9. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement the insurance risk assessment method as described in any one of claims 1 to 6 and the training method for the insurance risk assessment model as described in claim 7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the insurance risk assessment method as described in any one of claims 1 to 6 and the steps of the training method for the insurance risk assessment model as described in claim 7.

Images