Insurance claim data processing method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses an insurance claim data processing method and device, computer equipment and a storage medium in the fields of financial science and technology and medical health. The method comprises the steps of obtaining to-be-processed data; performing image analysis on the medical image data through a preset image analysis model to obtain image feature information; performing data analysis on the wearable device monitoring data and the inquiry record data through a preset scoring model to obtain health scoring information; determining a basic value of a claim settlement proportion according to the image feature information, determining an adjustment value of the claim settlement proportion according to the health score information, and determining a target claim settlement amount according to the basic value and the adjustment value; performing data analysis on the image feature information, the health score information and the candidate service feature information through a preset health care recommendation model to obtain a recommended health care service; and generating a claim settlement processing result according to the target claim settlement amount and the recommended health-care service. According to the invention, the data processing efficiency is improved, and the integration of claim settlement and health care service is realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more particularly to an insurance claims data processing method, apparatus, computer equipment, and storage medium in the fields of fintech and healthcare. Background Technology

[0002] Health insurance is both an insurance product and involves the processing of health data, representing a typical application of the deep integration of fintech and healthcare. However, current health insurance operations face several challenges. In the healthcare industry, diverse types of medical data, such as electronic medical records, medical images, and examination data, present significant professional barriers. Health insurance claims often rely on manual review, reducing efficiency and increasing the risk of misjudgments due to varying levels of experience. Furthermore, health insurance remains largely a passive response model focused on disease treatment and reimbursement. Customers only file claims after an incident, resulting in a disconnect between health management and claims processing, and a lack of effective health intervention and reasonable health management services. Summary of the Invention

[0003] Therefore, it is necessary to provide an insurance claims data processing method, device, computer equipment, and storage medium to address the aforementioned technical problems, in order to solve the issues of low data processing efficiency and the disconnect between health management and claims in health insurance claims.

[0004] An insurance claims data processing method, comprising: Acquire data to be processed, including the target user's medical imaging data, wearable device monitoring data, and consultation record data; The medical image data is analyzed using a preset image analysis model to obtain image feature information; Health score information is obtained by analyzing the monitoring data and consultation record data of the wearable device through a preset scoring model; The base value of the claim ratio is determined based on the image feature information, the adjustment value of the claim ratio is determined based on the health score information, and the target claim amount is determined based on the base value and the adjustment value of the claim ratio. By using a preset health and wellness recommendation model, data analysis is performed on the image feature information, health score information, and candidate service feature information to obtain recommended health and wellness services. The claim processing result is generated based on the target claim amount and the recommended health and wellness services.

[0005] An insurance claims data processing device, comprising: The data acquisition module is used to acquire data to be processed, including medical imaging data of the target user, wearable device monitoring data, and consultation record data. The image analysis module is used to perform image analysis on the medical image data through a preset image analysis model to obtain image feature information; The health scoring module is used to analyze the monitoring data and consultation record data of the wearable device through a preset scoring model to obtain health scoring information. The claim amount determination module is used to determine a base value for the claim ratio based on the image feature information, determine an adjustment value for the claim ratio based on the health score information, and determine a target claim amount based on the base value and the adjustment value for the claim ratio. The health and wellness service recommendation module is used to perform data analysis on the image feature information, health score information and candidate service feature information through a preset health and wellness recommendation model to obtain recommended health and wellness services. The processing result generation module is used to generate a claims processing result based on the target claim amount and the recommended health and wellness services.

[0006] A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor implements the above-described insurance claims data processing method when executing the computer-readable instructions.

[0007] A computer-readable storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the insurance claims data processing method described above.

[0008] In the aforementioned insurance claims data processing method, apparatus, computer equipment, and storage medium, the insurance claims data processing method acquires the data to be processed; performs image analysis on medical image data using a preset image analysis model to obtain image feature information; performs data analysis on wearable device monitoring data and consultation record data using a preset scoring model to obtain health score information; determines a base value for the claim ratio based on the image feature information, determines an adjustment value for the claim ratio based on the health score information, and determines a target claim amount based on the base value and adjustment value for the claim ratio; performs data analysis on the image feature information, health score information, and candidate service feature information using a preset health and wellness recommendation model to obtain recommended health and wellness services; and generates a claims processing result based on the target claim amount and recommended health and wellness services. This invention combines multi-dimensional data from medical image data, wearable device monitoring data, and consultation record data for automated analysis, constructing a closed-loop process encompassing medical image verification, health behavior quantification, and health and wellness service recommendation, thereby improving data processing efficiency and achieving intelligent, dynamic, and precise health insurance claims processing. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an application environment for an insurance claims data processing method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an insurance claims data processing method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of an insurance claims data processing device according to an embodiment of the present invention; Figure 4 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0011] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0012] The insurance claims data processing method provided in this embodiment can be applied to, for example, Figure 1 In this application environment, the client communicates with the server. Clients include, but are not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0013] The insurance claims data processing method described in this embodiment can be applied to the health insurance industry, which deeply integrates the fintech and healthcare sectors. On one hand, the health insurance claims data processing flow involves the application of claims rules and the determination of claims amounts. Low claims efficiency and misjudgments of claims amounts can lead to decreased user satisfaction. On the other hand, the health insurance claims data processing flow involves the analysis of medical and health data and the recommendation of health services. For example, medical and health data includes multi-dimensional data such as medical images from medical examinations, heart rate and blood pressure parameters from health monitoring using smart wearable devices, and medical record data from consultations.

[0014] In one embodiment, such as Figure 2 As shown, an insurance claims data processing method is provided, including the following steps S10-S60: S10. Obtain data to be processed, including the target user's medical imaging data, wearable device monitoring data, and consultation record data.

[0015] Understandably, users can input or upload data to be processed through the client's display interface, and the client will then send this data to the server. Data to be processed refers to information that needs to be analyzed and processed in the health insurance claims process. The server receives the data to be processed and extracts the target user's medical imaging data, wearable device monitoring data, and consultation record data from it. The target user refers to the health insurance user who needs to file a claim. Medical imaging data refers to medical imaging examination images generated by the user during the current claims period, such as chest CT images, MRI images, angiography, and DICOM image storage data. Wearable device monitoring data refers to health data collected through smart wearable devices, such as heart rate, blood oxygen saturation, blood glucose, blood pressure, exercise intensity, and sleep quality. Consultation record data refers to records of hospital visits and medications taken by the user during the current claims period, including structured data (electronic medical records) and unstructured data (medication prescription records). Different users correspond to different medical imaging data, wearable device monitoring data, and consultation record data during different time periods.

[0016] S20. Perform image analysis on the medical image data using a preset image analysis model to obtain image feature information.

[0017] Understandably, the server analyzes medical image data using a pre-set image analysis model to obtain image feature information. The pre-set image analysis model is a pre-trained neural network model used to automatically analyze input medical images and extract information related to disease diagnosis and lesion severity. Image feature information refers to quantitative indicators extracted from medical images that are related to organ disease or health status, used to describe attributes such as the morphology, location, and severity of lesions. Image feature information is a crucial bridge connecting image analysis and claims decisions, and can specifically include morphological features, spatial location features, and temporal dynamic change features. For example, morphological features include lesion size (diameter / volume), shape (circular / irregular), boundary clarity, and density (CT value); spatial location features include the organ where the lesion is located and its relationship to surrounding tissues (e.g., whether it invades blood vessels or lymph nodes); and dynamic change features include the rate of change in lesion volume and density trends over multiple examinations.

[0018] S30. The wearable device monitoring data and consultation record data are analyzed using a preset scoring model to obtain health score information.

[0019] Understandably, in addition to performing intuitive image analysis on medical imaging data, the server also combines wearable device monitoring data and consultation record data for multimodal data analysis to ensure the accuracy of insurance claims. The server analyzes the wearable device monitoring data and consultation record data using a preset scoring model to obtain health score information. The preset scoring model refers to... The health score information refers to...

[0020] A pre-defined scoring model is used to perform multi-dimensional analysis of a user's non-image-based health data, such as wearable device monitoring data and medical records, to output a quantitative health score. This model can assess a user's current health status and potential risks, providing an objective basis for adjusting claim ratios. The pre-defined scoring model can be a neural network model based on machine learning or a mathematical algorithm model built based on statistical methods. The health score information is a comprehensive indicator value generated after quantitative analysis of the user's health data by the scoring model, reflecting the user's current health status. Specifically, the health score information can be a weighted value of different dimensional indicators. For example, physiological indicators include resting heart rate, blood pressure fluctuation, blood oxygen saturation, and blood glucose level, with a corresponding weight of 35%; behavioral indicators include sleep duration, exercise frequency, sedentary time, and medication adherence, with a corresponding weight of 25%.

[0021] S40. Determine the base value of the claim ratio based on the image feature information, determine the adjustment value of the claim ratio based on the health score information, and determine the target claim amount based on the base value and the adjustment value of the claim ratio.

[0022] Understandably, the server can determine the base claim ratio based on image feature information. The base claim ratio is a pre-agreed percentage in the insurance contract, determined based on the scope of insurance liability and the sum insured. Specifically, different percentages can be assigned based on the severity of organ lesions in the image feature information; for example, the base claim ratio for mild lesions is 30%, and for severe lesions, it is 80%. The server can determine the adjustment value for the claim ratio based on health score information. The adjustment value is an increment or decrement that dynamically adjusts the base value according to the insured's health behavior habits. Specifically, the adjustment value can be a value between -10% and +10% corresponding to different health scores; for example, an adjustment of +5% corresponds to a health score ≥80, and an adjustment of -10% corresponds to a health score <60. The adjustment value can guide users to improve their behavior and incentivize health management through a positive adjustment method that increases the ratio for higher health scores. The server determines the target claim amount based on the base claim ratio and the adjustment value. The target claim amount refers to the actual claim amount that the insured can receive after combining the base value and adjustment value of the claim ratio. Specifically, the target claim amount = contract base amount × (base value of claim ratio + adjustment value of claim ratio), where the contract base amount refers to the standardized payout amount stipulated in the insurance contract.

[0023] S50. The image feature information, health score information and candidate service feature information are analyzed by a preset health and wellness recommendation model to obtain recommended health and wellness services.

[0024] Understandably, in addition to determining the target claim amount based on the data to be processed, the server also recommends health and wellness services based on the data, realizing an integrated service of passive claims and proactive health management. The server analyzes image feature information, health score information, and candidate service feature information through a pre-set health and wellness recommendation model to obtain recommended health and wellness services. The pre-trained health and wellness recommendation model is a neural network model used to select suitable health and wellness services from candidate services based on the insured's health status. Image feature information and health score information reflect the insured's health status and behavioral characteristics; different health statuses and behavioral characteristics are suitable for different candidate health and wellness services. Candidate health and wellness services refer to a series of health management service options available to meet the health and wellness needs of different individuals. Each candidate health and wellness service corresponds to its own candidate service feature information, which refers to information used to characterize the service content and functional characteristics of the candidate health and wellness service. Recommended health and wellness services are those suitable for the target user selected from the candidate health and wellness services.

[0025] S60. Generate a claims processing result based on the target claim amount and the recommended health and wellness services.

[0026] Understandably, the server generates a claims processing result based on the target claim amount and recommended health and wellness services, and sends the result to the target user's client. The claims processing result refers to information indicating whether a claim is made, the amount of the claim, and the recommended health and wellness services.

[0027] This embodiment acquires data to be processed; performs image analysis on medical image data using a preset image analysis model to obtain image feature information; performs data analysis on wearable device monitoring data and consultation record data using a preset scoring model to obtain health score information; determines a base value for the claim ratio based on the image feature information, determines an adjustment value for the claim ratio based on the health score information, and determines the target claim amount based on the base value and adjustment value; performs data analysis on the image feature information, health score information, and candidate service feature information using a preset health and wellness recommendation model to obtain recommended health and wellness services; and generates a claim processing result based on the target claim amount and recommended health and wellness services. This embodiment combines multi-dimensional data from medical image data, wearable device monitoring data, and consultation record data for automated analysis, constructing a closed-loop process for medical image verification, health behavior quantification, and health and wellness service recommendation, improving data processing efficiency and achieving intelligent, dynamic, and precise health insurance claims.

[0028] In one embodiment, step S20, namely, performing image analysis on the medical image data using a preset image analysis model to obtain image feature information, includes: S201. Perform graph convolution analysis on the medical image data using a preset image analysis model to obtain the image spatial feature parameters of the target image object in the medical image data. S202. Obtain the historical spatial feature parameters of the target image object in the historical medical image data of the target user; S203. Determine the image fraud score based on the image spatial feature parameters and historical spatial feature parameters; S204. When the image fraud score is less than the preset image fraud threshold, determine the image feature information based on the image spatial feature parameters.

[0029] Understandably, when the server performs image analysis on medical image data, it first performs graph convolution analysis on the medical image data using a preset image analysis model to obtain the image spatial feature parameters of the target image object in the medical image data. The target image object refers to the organ object in the medical image image that needs to be analyzed for its health and disease status. The input to the preset image analysis model is medical image data (DICOM image file sequence), specifically in the format [batch, time_steps, h, w, d], where batch refers to the number of image samples processed in a batch; time_steps refers to the image file sequence at different time points (reflecting the "time" dimension); h, w, and d correspond to the height, width, and depth of a single medical image, respectively (reflecting the "three-dimensional space" dimension). Image spatial feature parameters refer to the image feature information, which are spatial location quantification indicators extracted from medical images that are related to the health or disease status of organs, such as the location of the diseased organ and its positional relationship with surrounding tissues (e.g., whether blood vessels are involved).

[0030] In one specific embodiment, the preset image analysis model is a neural network model based on 3D scene reconstruction and rendering (Neural Radiance Fields, NeRF) technology. A multilayer perceptron maps 3D coordinates and viewpoint direction to color and density, achieving high-precision spatial stereoscopic viewpoint synthesis. The formula for image convolution analysis is... ,in, The node feature matrix can be understood in medical imaging scenarios as a set of feature vectors (such as gray values, texture features, etc.) for different regions in the image (such as different parts of an organ or sub-regions of a lesion). Indicates the first Multilayer networks, graph convolutions are typically multi-layered structures, with superscripts... Feature matrices used to distinguish different network layers (e.g.) =0 represents the input layer feature. =1 represents the features after the first convolutional layer); This represents the adjacency matrix, used to describe the connectivity between different regions (nodes) in an image (such as whether the lesion area is adjacent to other areas and how far apart they are). The connectivity matrix (diagonal matrix) is used to describe the sum of connectivity strengths of each region (node) in an image, reflecting the correlation density between regions (such as the degree of connectivity between lesions and surrounding normal tissues). This represents the weight matrix, used to adjust the importance of different features in the convolution operation. It consists of parameters learned during model training (such as assigning higher weights to lesion edge features).

[0031] Next, the server obtains the historical spatial feature parameters of the target image object from the target user's historical medical image data, and determines the image fraud score based on the image spatial feature parameters and the historical spatial feature parameters. Historical medical image data refers to medical image examination images generated by the user within the past claim period. The historical spatial feature parameters of the target image object refer to the spatial location quantification indicators extracted from historical medical images that are related to the health or disease state of the same organ. The server performs a time-series dynamic change comparison based on the image spatial feature parameters and the historical spatial feature parameters, such as the rate of change in lesion volume and density trends across multiple examinations, and generates an image fraud score based on the comparison results. The image fraud score is a quantification value used to characterize the logical reasonableness of organ lesions in medical images from multiple examinations.

[0032] In one specific embodiment, the fraud scoring formula is "Fraud Score = 0.4×Image Conflict + 0.3×Spatiotemporal Anomaly + 0.3×Behavioral Deviation". By comprehensively evaluating and quantifying "Image Conflict" (such as logical contradictions between current images and historical images), "Spatiotemporal Anomaly" (such as unreasonable behavior of frequently visiting different medical institutions in a short period of time), and "Behavioral Deviation" (such as inconsistencies between health data and medical behavior), a fraud risk score is calculated to help identify false claims (such as AI forging DICOM medical image files).

[0033] Finally, the server compares the image fraud score with a preset image fraud threshold, and processes the comparison differently depending on the magnitude of the scores. The preset image fraud threshold is a pre-set threshold used to determine the authenticity of medical image data. When the image fraud score is greater than or equal to the preset threshold, it indicates that the target user's medical image data may be fake or synthetic, and a warning is issued without outputting image feature information. When the image fraud score is less than the preset threshold, it indicates that the target user's medical image data is genuine, and image feature information is directly determined based on image spatial feature parameters.

[0034] This embodiment performs graph convolution analysis and image fraud analysis on medical image data using a preset image analysis model. This ensures the availability of image spatial feature parameters, improves the accuracy and efficiency of claims verification in the image dimension, and provides quantitative basis for subsequent claims decisions and recommendations for health and wellness services based on image feature information.

[0035] In one embodiment, step S30, namely, analyzing the wearable device monitoring data and consultation record data using a preset scoring model to obtain health score information, includes: S301. Determine the exercise participation index value, risk control index value, and service use index value based on the wearable device monitoring data, and determine the consultation compliance index value based on the consultation record data; S302. The exercise participation index, risk control index, service usage index, and consultation compliance index are weighted and analyzed using a preset scoring model to obtain health score information.

[0036] Understandably, when the server analyzes wearable device monitoring data and consultation record data, it first determines the exercise participation index, risk control index, and service usage index based on the wearable device monitoring data, and determines the consultation adherence index based on the consultation record data. The exercise participation index is a quantitative indicator representing the user's exercise achievement rate (activity). The risk control index is a quantitative indicator representing the user's health risk control (risk_control). The service usage index is a quantitative indicator representing the user's participation in health services (service_use). The consultation adherence index is a quantitative indicator representing the user's medication adherence (medication).

[0037] Next, the server inputs data from four dimensions—exercise participation, risk control, service usage, and consultation adherence—into a pre-defined scoring model for weighted analysis, outputting a standardized health score from 0 to 100. Specifically, the pre-defined scoring model uses learnable weights to sum the data across the four dimensions (e.g., initial weights of 0.25, 0.2, 0.25, and 0.3), then passes the sum through a neural network (full-processed layer) to output a standardized result of 0-1, which is finally multiplied by 100 to convert it into a percentage health score. The corresponding weighted summation formula is: Health Score = 0.25 × Exercise Participation + 0.2 × Risk Control + 0.25 × Service Usage + 0.3 × Consultation Adherence. The weights for each of the exercise participation, risk control, service usage, and consultation adherence indicators can be adjusted as needed.

[0038] This embodiment quantifies scattered health behavior indicators (such as whether medication is taken regularly, whether exercise is up to standard, etc.) and integrates them into a single standardized score through weighted analysis using a preset scoring model. Based on the health score, a comprehensive assessment of the user's health status and behavior is achieved, providing a quantitative basis for subsequent claims decisions and recommendations for health and wellness services.

[0039] In one embodiment, step S40, namely determining the basic value of the claim ratio based on the image feature information, includes: S401. Determine the image feature index value based on the image feature information; S402. Determine the health characteristic index value based on the wearable device monitoring data and medical record data; S403. Perform attention analysis on the image feature index value and the health feature index value to obtain the basic value of the claim ratio.

[0040] Understandably, when determining the base value for the claim ratio, the server needs to consider whether the organ's disease state meets the claim conditions (e.g., whether the disease reaches the compensation standard stipulated in the contract), and also needs to provide an auxiliary assessment based on health behavior status to determine whether it meets the claim conditions. Furthermore, it needs to provide a corresponding base compensation ratio suggestion based on the degree of change in the organ's disease (e.g., the base ratio may be adjusted due to disease progression). On one hand, the server determines image feature index values ​​based on image feature information. Image feature index values ​​refer to quantitative indicator parameters used to assess whether organ disease meets the claim conditions from a medical imaging perspective. Image feature index values ​​can be 3D volume data of organs from medical images, such as features extracted through the SwinTransformer 3D encoder (3D lesion structure information from CT / MRI). On the other hand, the server determines health feature index values ​​based on wearable device monitoring data and consultation record data. Health feature index values ​​refer to quantitative indicator parameters used to assess whether health behavior status meets the claim conditions from wearable device monitoring data and consultation record data. Health feature index values ​​can include multiple health indicators (such as heart rate, blood oxygen, treatment adherence, etc., values ​​from wearable devices and consultation records). Finally, the server performs attention analysis on the image feature index values ​​and health feature index values ​​to obtain the base value for the claim ratio. Attention analysis quantifies the contribution of different features to claim conditions (such as disease diagnosis and health assessment) by dynamically assigning weights. Image features (such as CT values, texture heterogeneity, and lesion morphology) may reflect the physical characteristics of lesions, while health features (such as age, blood sugar levels, and sleep quality) may reflect an individual's physiological state or disease risk. After attention weighting, the two types of features are merged into a unified index to obtain the base value for the claim ratio corresponding to the unified index.

[0041] This embodiment uses a cross-modal attention mechanism to link and fuse the lesion state features of medical images (such as lesion size and location) with health behavior features (such as exercise achievement rate and blood glucose changes), integrating the spatial structural features of images with the dynamic indicator features of health data. This provides multi-dimensional comprehensive feature support for the decision on the base value of the claim ratio, ensuring the rationality of the base value of the claim ratio.

[0042] In one embodiment, step S40, namely determining the claim ratio adjustment value based on the health score information, includes: S404. Determine the claim reward score value based on the health score information; S405. The health score information and the claim reward score are processed according to the preset claim adjustment rules to obtain the claim ratio adjustment value.

[0043] Understandably, when determining the claim adjustment value, the server first introduces a reward function to determine the claim reward score value corresponding to the health score information. Then, it processes the health score information and the claim reward score value according to preset claim adjustment rules to obtain the claim adjustment value. The claim reward score value is used to quantify the impact of user health behaviors (such as treatment adherence and exercise target achievement rate) on the claim outcome. The preset claim adjustment rules refer to the pre-set claim adjustment rules that take into account the impact of user health behaviors on the claim outcome.

[0044] In one specific embodiment, the claim ratio adjustment value is obtained using the `update` method of `TD3Agent`. For example, the target is `target_Q = rewards + 0.99 × target_Q`, where `rewards` refers to the claim reward score, used to measure the immediate benefit of the current claim strategy. The claim ratio adjustment value is obtained by minimizing the error between the current `target_Q` value and the target `target_Q` value. The claim reward score is used to guide the claim strategy towards optimization towards "reducing claim costs and increasing users' enthusiasm for health management". The strategy update formula corresponding to the preset claim adjustment rule is as follows: ,in, The policy objective function, such as the loss function actor_loss = -self.critic(states, self.actor(states)).mean(), can be regarded as a simplified objective function; This represents policy parameters, such as weight parameters; The target Q-value is a key indicator for policy gradient updates. It indicates the state, such as the user's current health score, lesion imaging characteristics, and historical claims records; This indicates an action, such as an adjustment to the payout ratio (-10% to +10%). This represents expectation, used to calculate the expected value of long-term rewards. Specific preset claims adjustment rules are as follows: class TD3Agent: def __init__(self, state_dim, action_dim): self.actor = ActorNetwork(state_dim, action_dim) self.critic = CriticNetwork(state_dim, action_dim) self.target_actor = copy.deepcopy(self.actor) self.target_critic = copy.deepcopy(self.critic) def select_action(self, state): return self.actor(state) # Output the adjustment value of the compensation ratio (-0.1 to +0.1) def update(self, replay_buffer): states, actions, rewards, next_states = replay_buffer.sample() # Calculate the target Q value target_actions = self.target_actor(next_states) target_Q = self.target_critic(next_states, target_actions) target_Q = rewards + 0.99 * target_Q # Update Critic current_Q = self.critic(states, actions) critic_loss = F.mse_loss(current_Q, target_Q.detach()) # Delayed update of Actor if self.total_steps % 2 == 0: actor_loss = -self.critic(states, self.actor(states)).mean() self.actor_optimizer.zero_grad() actor_loss.backward() self.actor_optimizer.step() This embodiment introduces a reward function for claims reward scores and determines the claims ratio adjustment value according to preset claims adjustment rules. It can guide users to improve their behavior and incentivize health management by positively adjusting the claims ratio based on a high health score. This dynamically optimizes the claims ratio adjustment strategy.

[0045] In one embodiment, step S50, namely, performing data analysis on the image feature information, health score information, and candidate service feature information using a preset health and wellness recommendation model to obtain recommended health and wellness services, includes: S501. Determine the image monitoring matching degree based on the medical imaging data and wearable device monitoring data, and determine the relationship fraud score based on the consultation record data; S502. When the image monitoring matching degree reaches the preset matching degree threshold and the relationship fraud score is less than the preset relationship fraud threshold, the image feature information, health score information and candidate service feature information of the candidate health care services are analyzed by the preset health care recommendation model to obtain recommended health care services.

[0046] Understandably, before inputting image feature information and health score information into the preset health and wellness recommendation model, the server needs to perform a recommendation usability analysis on the medical imaging data, wearable device monitoring data, and consultation record data. First, the server determines the image monitoring matching degree based on the medical imaging data and wearable device monitoring data, and determines the relationship fraud score based on the consultation record data. The image monitoring matching degree is a quantitative value used to characterize the degree of matching between the degree of organ lesion in the medical imaging data and the monitoring index value of the same organ in the wearable device monitoring data. For example, if the medical imaging data shows a heart lesion, the corresponding wearable device monitoring data rate should show a certain degree of fluctuation. The relationship fraud score is a quantitative value used to characterize whether there are inconsistencies in the relationships between the insured, the medical institution, and the doctor extracted from the consultation record data.

[0047] Then, when the image monitoring matching degree reaches a preset matching degree threshold and the relationship fraud score is less than a preset relationship fraud threshold, the server analyzes the image feature information, health score information, and candidate service feature information of candidate health care services through a preset health care recommendation model to obtain recommended health care services. The preset matching degree threshold is a pre-set critical value used to assess the authenticity of the matching degree between medical image data and wearable device monitoring data. The preset relationship fraud threshold is a pre-set fraud score critical value used to assess the authenticity of consultation record data. When the image monitoring matching degree is less than or equal to the preset matching degree threshold and the relationship fraud score reaches the preset relationship fraud threshold, the insurance claim data processing process is stopped, an anomaly prompt message is generated, and sent to the terminal of the anomaly handling personnel to trigger manual investigation and processing.

[0048] In one specific embodiment, the preset health and wellness recommendation model uses federated learning to analyze data and output recommended health and wellness services. The homomorphic encrypted aggregation formula in federated learning is as follows: ,in, This indicates a homomorphic encryption function (such as the Paillier and CKKS algorithms) that supports multiplicative homomorphism from ciphertext multiplication to plaintext addition. Indicates the first Local data / model gradients of each participant, such as the training gradients of a hospital or clinic, or patient feature vectors; This represents the ciphertext of the summation of data / gradients from all participants. This represents the product of encrypted data from all participating parties. Leveraging multiplicative homomorphism, ciphertext multiplication is equivalent to plaintext addition. This enables "privacy-preserving distributed computing" in federated learning within healthcare institutions and multi-institutional data aggregation. The knowledge distillation loss formula in federated learning is... ,in, It represents the knowledge distillation loss, which measures the distributional difference between the student model and the teacher model, such as the loss function of the student model learning from the teacher model; This represents the Kullback-Leibler divergence, used to calculate the asymmetric difference between two probability distributions (information loss of the student model distribution relative to the teacher model distribution). This represents the output distribution of the student model. This represents the output distribution of the teacher model.

[0049] This embodiment, while ensuring the authenticity and usability of medical imaging data, wearable device monitoring data, and consultation record data, analyzes imaging feature information, health score information, and candidate service feature information through a preset health and wellness recommendation model to obtain recommended health and wellness services. This achieves a deep integration of claims and health services and improves the rationality of health and wellness service recommendations.

[0050] In one embodiment, step S501, namely determining the relationship fraud score based on the consultation record data, includes: S5011. Determine the consultation relationship feature data based on the consultation record data; S5012. The consultation relationship feature data is analyzed using a preset graph relationship model to obtain a relationship fraud score.

[0051] Understandably, when determining the relationship fraud score based on consultation record data, the server first determines the consultation relationship feature data based on the consultation record data. The consultation relationship feature data is a graph structure data extracted from the consultation record data to represent the relationship between the insured, doctor, medical institution, and medical examination equipment. It includes three parts: entity nodes, association edges, and attribute features. For example, the entities involved in this claim (such as patient, doctor, clinic, and imaging equipment) are extracted, and a local relationship graph (such as the association chain graph between "patient, doctor, hospital or clinic, and imaging equipment") is constructed.

[0052] Then, the server analyzes the consultation relationship feature data using a pre-set graph relationship model to obtain a relationship fraud score. The pre-trained graph relationship model is a neural network model used to identify and score abnormal relationship nodes in the relationship graph of the consultation relationship feature data. In one specific embodiment, the pre-set graph relationship model is a Fraud GNN-based neural network model that identifies abnormal behavior by analyzing graph structure data (such as user-doctor relationships, consultation relationship networks, etc.). The pre-set graph relationship model outputs the fraud probability of a node, such as "the clinic's fraud probability is 0.89" or "the image submitted by the doctor is suspicious 0.76," as well as the reason for the anomaly, such as "high clinic fraud risk" or "abnormal device association."

[0053] This embodiment analyzes the characteristic data of consultation relationships using a preset graph relationship model to obtain a relationship fraud score, which can accurately identify consultation relationships with fraud risks and ensure the authenticity and usability of consultation record data.

[0054] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0055] In one embodiment, an insurance claims data processing apparatus is provided, which corresponds one-to-one with the insurance claims data processing methods described in the above embodiments. For example... Figure 3 As shown, the insurance claims data processing device includes a data acquisition module 10, an image analysis module 20, a health scoring module 30, a claims amount determination module 40, a health and wellness service recommendation module 50, and a processing result generation module 60. Detailed descriptions of each functional module are as follows: Data acquisition module 10 is used to acquire data to be processed, including medical image data of the target user, wearable device monitoring data and consultation record data. Image analysis module 20 is used to perform image analysis on the medical image data through a preset image analysis model to obtain image feature information; The health scoring module 30 is used to perform data analysis on the wearable device monitoring data and consultation record data through a preset scoring model to obtain health scoring information. The claim amount determination module 40 is used to determine a base value for the claim ratio based on the image feature information, determine an adjustment value for the claim ratio based on the health score information, and determine a target claim amount based on the base value and the adjustment value for the claim ratio. The health and wellness service recommendation module 50 is used to perform data analysis on the image feature information, health score information and candidate service feature information through a preset health and wellness recommendation model to obtain recommended health and wellness services. The processing result generation module 60 is used to generate a claims processing result based on the target claim amount and the recommended health and wellness services.

[0056] In one embodiment, the image analysis module 20 includes: The graph convolution analysis unit is used to perform graph convolution analysis on the medical image data through a preset image analysis model to obtain the image spatial feature parameters of the target image object in the medical image data. The historical spatial feature parameter acquisition unit is used to acquire the historical spatial feature parameters of the target image object in the historical medical image data of the target user. An image fraud scoring unit is used to determine an image fraud scoring value based on the image spatial feature parameters and historical spatial feature parameters. The image feature information determination unit is used to determine image feature information based on the image spatial feature parameters when the image fraud score value is less than a preset image fraud threshold.

[0057] In one embodiment, the health scoring module 30 includes: The indicator value determination unit is used to determine the exercise participation indicator value, risk control indicator value, and service use indicator value based on the wearable device monitoring data, and to determine the consultation compliance indicator value based on the consultation record data. The indicator value weighted analysis unit is used to perform weighted analysis on the exercise participation indicator value, risk control indicator value, service use indicator value and consultation compliance indicator value through a preset scoring model to obtain health score information.

[0058] In one embodiment, the claim amount determination module 40 includes: An image feature index value determination unit is used to determine image feature index values ​​based on the image feature information. A health characteristic index value determination unit is used to determine health characteristic index values ​​based on the wearable device monitoring data and consultation record data. The attention analysis unit is used to perform attention analysis on the image feature index values ​​and the health feature index values ​​to obtain the basic value of the claim ratio.

[0059] In one embodiment, the claim amount determination module 40 further includes: The claims reward score determination unit is used to determine the claims reward score based on the health score information. The claims ratio adjustment value determination unit is used to process the health score information and the claims reward score value according to the preset claims adjustment rules to obtain the claims ratio adjustment value.

[0060] In one embodiment, the health and wellness service recommendation module 50 includes: The data preprocessing unit is used to determine the image monitoring matching degree based on the medical image data and wearable device monitoring data, and to determine the relationship fraud score based on the consultation record data; The recommended health and wellness service determination unit is used to analyze the image feature information, health score information, and candidate service feature information of candidate health and wellness services through a preset health and wellness recommendation model when the image monitoring matching degree reaches a preset matching degree threshold and the relationship fraud score is less than a preset relationship fraud threshold, so as to obtain recommended health and wellness services.

[0061] In one embodiment, the health and wellness service recommendation module 50 further includes: A consultation relationship feature data determination unit is used to determine consultation relationship feature data based on the consultation record data; The relationship fraud score determination unit is used to perform data analysis on the consultation relationship feature data through a preset graph relationship model to obtain a relationship fraud score.

[0062] Specific limitations regarding the insurance claims data processing device can be found in the limitations of the insurance claims data processing method described above, and will not be repeated here. Each module in the aforementioned insurance claims data processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0063] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a readable storage medium and internal memory. The readable storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The database stores data related to the insurance claims data processing method. The network interface communicates with external terminals via a network connection. When the computer-readable instructions are executed by the processor, they implement an insurance claims data processing method. The readable storage medium provided in this embodiment includes both non-volatile and volatile readable storage media.

[0064] In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor performs the following steps when executing the computer-readable instructions: Acquire data to be processed, including the target user's medical imaging data, wearable device monitoring data, and consultation record data; The medical image data is analyzed using a preset image analysis model to obtain image feature information; Health score information is obtained by analyzing the monitoring data and consultation record data of the wearable device through a preset scoring model; The base value of the claim ratio is determined based on the image feature information, the adjustment value of the claim ratio is determined based on the health score information, and the target claim amount is determined based on the base value and the adjustment value of the claim ratio. By using a preset health and wellness recommendation model, data analysis is performed on the image feature information, health score information, and candidate service feature information to obtain recommended health and wellness services. The claim processing result is generated based on the target claim amount and the recommended health and wellness services.

[0065] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. The readable storage media stores computer-readable instructions, which, when executed by one or more processors, perform the following steps: Acquire data to be processed, including the target user's medical imaging data, wearable device monitoring data, and consultation record data; The medical image data is analyzed using a preset image analysis model to obtain image feature information; Health score information is obtained by analyzing the monitoring data and consultation record data of the wearable device through a preset scoring model; The base value of the claim ratio is determined based on the image feature information, the adjustment value of the claim ratio is determined based on the health score information, and the target claim amount is determined based on the base value and the adjustment value of the claim ratio. By using a preset health and wellness recommendation model, data analysis is performed on the image feature information, health score information, and candidate service feature information to obtain recommended health and wellness services. The claim processing result is generated based on the target claim amount and the recommended health and wellness services.

[0066] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When executed, these computer-readable instructions 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 by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of 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 memory bus dynamic RAM (RDRAM), etc.

[0067] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by 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.

[0068] The software tools or components not belonging to this company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use. 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; and these 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. A method for processing insurance claims data, characterized in that, include: Acquire data to be processed, including the target user's medical imaging data, wearable device monitoring data, and consultation record data; The medical image data is analyzed using a preset image analysis model to obtain image feature information; Health score information is obtained by analyzing the monitoring data and consultation record data of the wearable device through a preset scoring model; The base value of the claim ratio is determined based on the image feature information, the adjustment value of the claim ratio is determined based on the health score information, and the target claim amount is determined based on the base value and the adjustment value of the claim ratio. By using a preset health and wellness recommendation model, data analysis is performed on the image feature information, health score information, and candidate service feature information to obtain recommended health and wellness services. The claim processing result is generated based on the target claim amount and the recommended health and wellness services.

2. The insurance claims data processing method as described in claim 1, characterized in that, The step of performing image analysis on the medical image data using a preset image analysis model to obtain image feature information includes: The medical image data is subjected to graph convolution analysis by a preset image analysis model to obtain the image spatial feature parameters of the target image object in the medical image data; Obtain the historical spatial feature parameters of the target image object from the historical medical image data of the target user; The image fraud score is determined based on the image spatial feature parameters and historical spatial feature parameters. When the image fraud score is less than the preset image fraud threshold, image feature information is determined based on the image spatial feature parameters.

3. The insurance claims data processing method as described in claim 1, characterized in that, The process of analyzing the wearable device monitoring data and consultation record data using a preset scoring model to obtain health score information includes: The exercise participation index, risk control index, and service usage index are determined based on the wearable device monitoring data, and the consultation compliance index is determined based on the consultation record data. The health score information is obtained by weighting and analyzing the exercise participation index, risk control index, service use index, and consultation compliance index using a preset scoring model.

4. The insurance claims data processing method as described in claim 1, characterized in that, The step of determining the basic value of the claim ratio based on the image feature information includes: Determine the image feature index value based on the image feature information; Health characteristic index values ​​are determined based on the wearable device monitoring data and medical record data. Attention analysis is performed on the image feature index values ​​and the health feature index values ​​to obtain the basic value of the claim ratio.

5. The insurance claims data processing method as described in claim 1, characterized in that, The step of determining the claim adjustment value based on the health score information includes: The claim reward score is determined based on the aforementioned health score information; The health score information and the claim reward score are processed according to the preset claim adjustment rules to obtain the claim ratio adjustment value.

6. The insurance claims data processing method as described in claim 1, characterized in that, The step involves analyzing the image feature information, health score information, and candidate service feature information using a preset health and wellness recommendation model to obtain recommended health and wellness services, including: The matching degree of image monitoring is determined based on the medical imaging data and wearable device monitoring data, and the relationship fraud score is determined based on the consultation record data. When the image monitoring matching degree reaches a preset matching degree threshold and the relationship fraud score is less than a preset relationship fraud threshold, the image feature information, health score information and candidate service feature information of the candidate health care services are analyzed by a preset health care recommendation model to obtain recommended health care services.

7. The insurance claims data processing method as described in claim 6, characterized in that, The step of determining the relationship fraud score based on the consultation record data includes: Based on the consultation record data, determine the consultation relationship characteristic data; Data analysis of the consultation relationship feature data is performed using a preset graph relationship model to obtain a relationship fraud score.

8. An insurance claims data processing device, characterized in that, include: The data acquisition module is used to acquire data to be processed, including medical imaging data of the target user, wearable device monitoring data, and consultation record data. The image analysis module is used to perform image analysis on the medical image data through a preset image analysis model to obtain image feature information; The health scoring module is used to analyze the monitoring data and consultation record data of the wearable device through a preset scoring model to obtain health scoring information. The claim amount determination module is used to determine a base value for the claim ratio based on the image feature information, determine an adjustment value for the claim ratio based on the health score information, and determine a target claim amount based on the base value and the adjustment value for the claim ratio. The health and wellness service recommendation module is used to perform data analysis on the image feature information, health score information and candidate service feature information through a preset health and wellness recommendation model to obtain recommended health and wellness services. The processing result generation module is used to generate a claims processing result based on the target claim amount and the recommended health and wellness services.

9. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the computer-readable instructions, it implements the insurance claims data processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors cause the insurance claims data processing method as described in any one of claims 1 to 7 to be performed.

Images