Business analysis method and device based on health data, equipment and medium

Abstract

The invention relates to the technical field of data analysis, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses a business analysis method, device, equipment and medium based on health data. The method comprises the steps of obtaining a normalized medical image and time sequence alignment data, carrying out convolution on the normalized medical image to obtain a health risk level, carrying out data trend analysis on the time sequence alignment data to obtain a time sequence risk score, fusing the health risk level and the time sequence risk score to obtain a health risk score, and screening insurance policy adjustment schemes. And obtaining a target adjustment strategy, collecting multi-dimensional behavior data of the target customer after the strategy service is pushed, and performing feedback optimization on the target adjustment strategy to obtain target insurance data. According to the method, the health data and the insurance data are efficiently fused and dynamically linked, so that the service response efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and in particular to a business analysis method and device based on health data, equipment and medium. BACKGROUND

[0002] The existing insurance business and health data management technology have significant defects, mainly manifested in that health data and insurance services are separated, dynamic health information is difficult to effectively integrate and apply, and risk management still relies on static health assessment, resulting in strategy lag and insufficient precision; in terms of technical implementation, decentralized modules are generally used, lacking a unified architecture, making it difficult to support real-time interaction and automated decision-making of multi-source heterogeneous data, poor scalability, and complex operation and maintenance; in terms of compliance management, especially in the processing of sensitive data such as medical images, there is a lack of reliable evidence and audit mechanism, making it difficult to meet regulatory requirements; at the same time, the existing system service triggering mechanism is passive, lacking intelligent shunting and personalized adaptation capability, and failing to build a health linkage system covering intelligent consultation, medical analysis, health care services and other scenarios, which overall restricts the health data-driven insurance optimization capability and customer experience improvement.

[0003] In the field of medical health, hospitals or health management platforms have deployed some intelligent consultation, medical image analysis and health monitoring equipment, but these systems are usually isolated from each other, lacking a unified architecture to support data integration and dynamic service optimization, making it difficult for real-time health data of patients to be efficiently transferred and applied in multiple links such as diagnosis and treatment, rehabilitation and chronic disease management. For example, intelligent consultation cannot be linked to historical image data to assist in diagnosis, medical images are difficult to automatically integrate into personalized health management plans, and there is a lack of compliance traceability mechanism for sensitive data such as images, which poses a risk of data misuse and restricts the improvement of overall service efficiency and data credibility of smart medical care.

[0004] In the field of financial technology business, although some insurance companies try to integrate health data into pricing and risk assessment systems, such as dynamically adjusting premiums using wearable device data, the overall system still remains at the level of single data source and static rules, lacking multi-dimensional integration and real-time decision-making capability for deep health data such as medical images and consultation records, and failing to achieve intelligent policy management and proactive service triggering based on customer health status. At the same time, due to the lack of a unified data processing and decision-making architecture, financial systems are difficult to seamlessly integrate new health data services, frequent system modification brings high cost and compliance risk, which seriously restricts the innovation and landing of health-driven financial products.

[0005] In summary, the existing technology has the problem of system fragmentation and data island in health data application, making it difficult to achieve efficient integration and dynamic linkage of multi-source health data, resulting in low service response efficiency. SUMMARY

[0006] The application provides a health data-based business analysis method, device, equipment and medium, and mainly aims to solve the problem of low service response efficiency caused by the difficulty in efficient fusion and dynamic linkage of multi-source health data.

[0007] In a first aspect, to achieve the above object, the application provides a health data-based business analysis method, which comprises the following steps:

[0008] Obtaining medical image data and physiological time series data of a target customer, and performing standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data;

[0009] Performing convolution on the normalized medical images to obtain lesion spatial features, and generating a health risk level by using the lesion spatial features;

[0010] Performing data trend analysis on the time series alignment data to obtain a time series risk score;

[0011] Performing weighted fusion on the health risk level and the time series risk score to obtain a health risk score of the target customer;

[0012] Obtaining a plurality of insurance policy adjustment schemes, screening the insurance policy adjustment schemes according to the health risk score to obtain a target adjustment strategy;

[0013] Pushing services to the target customer according to the target adjustment strategy, collecting multi-dimensional behavior data of the target customer after the service pushing, performing tracking analysis on the multi-dimensional behavior data to obtain a strategy execution effect;

[0014] Optimizing the target adjustment strategy by using the strategy execution effect to obtain target insurance data.

[0015] In a second aspect, the application further provides a health data-based business analysis device, which comprises the following modules:

[0016] A data standardization module is configured to obtain medical image data and physiological time series data of a target customer, and perform standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data;

[0017] A medical image convolution module is configured to perform convolution on the normalized medical images to obtain lesion spatial features, and generate a health risk level by using the lesion spatial features;

[0018] A data trend analysis module is configured to perform data trend analysis on the time series alignment data to obtain a time series risk score;

[0019] A data weighted fusion module is used to perform weighted fusion on the health risk level and the temporal risk score to obtain the health risk score of the target customer;

[0020] A policy plan screening module is used to obtain a number of policy adjustment plans, screen the policy adjustment plans according to the health risk score, and obtain a target adjustment strategy;

[0021] A data tracking and analysis module is used to push services to the target customers according to the target adjustment strategy, collect multi-dimensional behavior data of the target customers after the service push, track and analyze the multi-dimensional behavior data, and obtain the strategy execution effect;

[0022] The adjustment strategy optimization module is used to optimize the target adjustment strategy using the strategy execution effect to obtain target insurance data.

[0023] In a third aspect, the present invention further provides an electronic device, comprising:

[0024] at least one processor; and,

[0025] a memory communicatively connected to the at least one processor; wherein,

[0026] The memory stores a computer program that can be executed by the at least one processor. The computer program is executed by the at least one processor to enable the at least one processor to perform the above-mentioned business analysis method based on health data.

[0027] In a fourth aspect, the present invention also provides a computer-readable storage medium, in which at least one computer program is stored. The at least one computer program is executed by a processor in an electronic device to implement the above-mentioned business analysis method based on health data.

[0028] The application is helpful to remove noise in the image and highlight important features, so that doctors or algorithms can more accurately identify disease lesions, through the calculation of the time series distance matrix and the cumulative matrix, the data at different time points are accurately aligned, the deviation in time is eliminated, the normalized medical image is convolved to obtain the lesion spatial feature, and the health risk level is generated using the lesion spatial feature, and the weighted fusion of the lesion feature map using the attention mechanism can focus on the most important lesion area and ignore irrelevant parts, thereby improving the accuracy of diagnosis. After global average pooling, the feature vector of the lesion is simplified, but the important information of the disease is still retained, providing a stable foundation for subsequent lesion risk scoring, the time series risk score is obtained by performing data trend analysis on the time series alignment data, the long short-term memory (LSTM) layer is used to capture time dependence, which is helpful to extract key time features from historical data, thereby enhancing the sensitivity of the model to long-term trends and short-term fluctuations, the health risk score of the target customer is obtained by weighting and fusing the health risk level and the time series risk score, not only considering the current health status (such as disease risk), but also being able to predict the future health trend (such as potential chronic disease or acute disease), obtaining a plurality of insurance policy adjustment schemes, screening the insurance policy adjustment schemes according to the health risk score to obtain a target adjustment strategy, using the fitness function in the genetic algorithm, through multiple rounds of screening, crossover, disturbance and evaluation, gradually optimizing the insurance policy scheme, ensuring that the finally selected scheme not only meets the health needs of the customer, but also improves the customer's satisfaction, and brings the maximum benefit to the insurance company, according to the target adjustment strategy, the target customer is served, and multi-dimensional behavior data of the target customer after service pushing is collected, the multi-dimensional behavior data is tracked and analyzed to obtain the strategy execution effect, through the comparison of the simulation strategy and the annotated adjustment strategy, the effectiveness of the insurance service can be further verified, through the tracking analysis of the multi-dimensional behavior data, the system can monitor and evaluate the implementation effect of the strategy in real time, ensure that the service pushing is highly consistent with the customer's needs, use the strategy execution effect to optimize the target adjustment strategy to obtain target insurance data, efficiently fuse and dynamically link health data and insurance data, thereby improving the service response efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without any creative labor.

[0030] Figure 1 An application environment schematic diagram of a health data based business analysis method in an embodiment of the present application;

[0031] Figure 2 A flow schematic diagram of a health data based business analysis method provided in an embodiment of the present application;

[0032] Figure 3 A medical image convolution process flow schematic diagram in a health data based business analysis method provided in an embodiment of the present application;

[0033] Figure 4 A module schematic diagram of a health data based business analysis device provided in an embodiment of the present application;

[0034] Figure 5 A structure schematic diagram of an electronic device for implementing a health data based business analysis method provided in an embodiment of the present application;

[0035] Figure 6 Another structure schematic diagram of an electronic device for implementing a health data based business analysis method provided in an embodiment of the present application.

[0036] The object, function characteristics and advantages of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0037] In order to make the person in the art better understand the technical solutions of the present disclosure, and to fully understand and implement the implementation process of the present disclosure how to apply technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, not all embodiments. The embodiments of the present disclosure and each feature in the embodiments can be combined with each other without conflict, and the technical solutions formed thereby are all within the protection scope of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor should be within the protection scope of the present disclosure.

[0038] It should be noted that the terms "first", "second", and the like in the description and in the claims of the present disclosure and above-described drawings are intended to distinguish similar objects and are not necessarily used to describe a particular sequential or chronological order. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, device, product or apparatus including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or apparatus.

[0039] The embodiment of the present application provides a business analysis method based on health data. The execution subject of the business analysis method based on health data includes but is not limited to at least one of electronic devices capable of being configured to execute the device provided by the embodiment of the present application, such as a server and a terminal. In other words, the business analysis method based on health data can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. basic cloud computing services.

[0040] The embodiment of the present application provides a business analysis method based on health data, which can be applied to Figure 1In the application environment, the client communicates with the server through the network. The server can obtain the medical image data and physiological time series data of the target customer through the client, standardize the medical image data and the physiological time series data respectively, obtain normalized medical images and time series alignment data, which helps to remove noise in the image and highlight important features, so that doctors or algorithms can more accurately identify disease lesions, accurately align data at different time points through the calculation of time series distance matrix and cumulative matrix, eliminate the deviation in time, convolve the normalized medical images to obtain lesion spatial features, and generate health risk levels using the lesion spatial features. The weighted fusion of lesion feature maps using the attention mechanism can focus on the most important lesion area and ignore irrelevant parts, improving the accuracy of diagnosis. After global average pooling, the feature vector of the lesion is simplified, but the important information of the disease is still retained, providing a stable foundation for subsequent lesion risk scoring. The time series risk score is obtained by performing data trend analysis on the time series alignment data. The long short-term memory (LSTM) layer is used to capture time dependence, which helps to extract key time features from historical data and enhance the sensitivity of the model to long-term trends and short-term fluctuations. The health risk score of the target customer is obtained by weighting and fusing the health risk level and the time series risk score, which not only considers the current health status (such as disease risk), but also can predict the future health trend (such as potential chronic or acute diseases). Obtain a plurality of policy adjustment schemes, filter the policy adjustment schemes according to the health risk score to obtain a target adjustment strategy, use the fitness function in the genetic algorithm to gradually optimize the policy through multiple rounds of screening, crossover, disturbance and evaluation, ensure that the finally selected scheme not only meets the customer's health needs, but also improves customer satisfaction and brings maximum benefits to the insurance company. According to the target adjustment strategy, the target customer is pushed to the service, and the multi-dimensional behavior data of the target customer after the service is pushed is collected, and the multi-dimensional behavior data is tracked and analyzed to obtain the strategy execution effect. By comparing the simulation strategy with the annotated adjustment strategy, the effectiveness of the insurance service can be further verified. Through tracking analysis of multi-dimensional behavior data, the system can monitor and evaluate the implementation effect of the strategy in real time, ensure that the service push is highly consistent with the customer's needs, and optimize the target adjustment strategy using the strategy execution effect to obtain target insurance data. Efficiently integrate and dynamically link health data and insurance data, thereby improving service response efficiency. Finally, the target insurance data is output and fed back to the user client. The client can be, but is not limited to, various personal computers, notebook computers, smartphones, tablet computers and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers. The application will be described in detail below through specific embodiments.

[0041] The following is explained in the specification of the present application, which utilizes the fitness function in genetic algorithm to gradually optimize the insurance policy scheme through multiple rounds of screening, crossover, disturbance and evaluation, ensures that the finally selected scheme not only meets the health needs of the customer, but also improves the customer's satisfaction and brings the maximum benefit to the insurance company. Through this iterative process, not only the optimal insurance adjustment scheme can be found, but also the matching and profitability can be continuously optimized to ensure that every customer can obtain the most accurate and high-quality insurance protection, and the health data and insurance data are efficiently fused and dynamically linked to improve the service response efficiency.

[0042] Referring to Figure 2 Fig. 1 shows a flowchart of a business analysis method based on health data provided by an embodiment of the present application. In this embodiment, the business analysis method based on health data includes:

[0043] S1, obtain medical image data and physiological time series data of a target customer, and perform standardization processing on the medical image data and the physiological time series data to obtain normalized medical images and time series alignment data.

[0044] In the embodiment of the present application, the pixel values of the medical image data are normalized to obtain normalized medical images, two data are randomly selected from the physiological time series data as target time series data pairs, the optimal alignment path between the target time series data pairs is obtained by dynamic time warping (DTW) algorithm, and then the physiological time series data is aligned by using the optimal alignment path, and finally the time series alignment data is obtained.

[0045] In the specific scenario of medical health, it can be used for processing and analysis of medical image data and physiological time series data. By performing pixel normalization and image enhancement on the medical image data, the quality of the medical image can be effectively improved, making it clearer and more convenient for doctors to make accurate diagnosis. At the same time, by aligning the physiological time series data, the physiological data at different time points can be matched to achieve more accurate disease prediction and health management. For example, in the monitoring of heart disease, by aligning the electrocardiogram (ECG) time series data, doctors can accurately compare the electrocardio activities in different time periods, identify potential arrhythmia or other abnormal conditions, and then provide personalized treatment plans for patients.

[0046] In the specific scenario of financial technology, the process can be used in risk assessment and credit scoring. By standardizing historical financial data such as customer transaction records and loan history, financial institutions can eliminate data bias and ensure the fairness and accuracy of the analysis results. Alignment of time series data is also very important, especially in loan default prediction. By accurately aligning the time series data of the customer's past payment behavior, it can better analyze the customer's repayment habits and future risk trends, providing a more reliable basis for credit scoring systems and helping banks and financial institutions make more scientific credit decisions.

[0047] In the embodiments of the present application, the medical image data and the physiological time series data are respectively standardized to obtain normalized medical images and time series alignment data, which includes:

[0048] The medical image data is image enhanced to obtain a medical enhanced image;

[0049] The pixel value of the medical enhanced image is normalized to obtain a normalized medical image;

[0050] Randomly select two groups of physiological time series data as target time series data pairs;

[0051] Determine the distance between the target time series data pairs, and construct a time series distance matrix according to the distance;

[0052] Obtain an initial cumulative matrix, and determine the minimum cumulative distance of each matrix position using the initial cumulative matrix and the time series distance matrix;

[0053] Update the initial cumulative matrix according to the minimum cumulative distance to obtain a cumulative distance matrix;

[0054] Backtracking the cumulative distance matrix to obtain an optimal alignment path;

[0055] Align the physiological time series data using the optimal alignment path to obtain time series alignment data.

[0056] In detail, the pixel value in the medical image data is adjusted to a unified standard range, usually [0, 1] or [-1, 1]. Get each pixel value in the original medical image, scale it according to the minimum and maximum pixel values of the medical image data, so that all pixel values are in the new range. By subtracting the minimum pixel value and dividing by the maximum range of pixel values, the normalized pixel value is obtained, which helps to eliminate the pixel value difference between different medical images.

[0057] Common image enhancement techniques include histogram equalization, contrast adjustment, noise removal, etc. The initial medical image is filtered to remove noise and enhance the clarity of the image. By adjusting the brightness and contrast of the image, the darker or brighter areas in the image are more obvious, highlighting the key areas such as lesions or abnormal structures. After image enhancement, the details of the image are further optimized to ensure that the key features in the image are more prominent, thereby improving the accuracy of diagnosis, and ultimately obtaining the normalized medical image.

[0058] Two sets of data are randomly selected from all available physiological time series data, which typically represent physiological signals (such as blood pressure, heart rate, etc.) at different time points. After selecting the target time series data pair, it is necessary to determine the similarity or difference between the two sets of data in the target time series data pair. Common distance measurement methods include Euclidean distance, dynamic time warping (DTW) distance, etc., which can capture the shift and change between different time points of time series data. By calculating the distance between the time series data pairs, a numerical distance measurement result is obtained. According to these distance values, a time series distance matrix is constructed, where each element represents the distance relationship between different time series.

[0059] The initial cumulative matrix is the same size as the time series distance matrix, where each element represents the cumulative distance at a specific position. Initially, these values are usually set to infinity or other large values. Using the distance value at each position in the time series distance matrix, combined with the initial cumulative matrix, the minimum cumulative distance at each matrix position is calculated by recursion. For each position, the cumulative distance is determined by the minimum cumulative distance of the previous position and the distance value of the current time series data pair. Usually, the dynamic programming method is used to calculate step by step from the starting position, and the value of each position is updated by minimizing the sum of the cumulative distance of the previous position and the distance of the current time series data pair. The minimum cumulative distance path in all possible paths can be determined.

[0060] During the update process, the value of each matrix position will be calculated based on the minimum cumulative distance of the adjacent position and the distance of the current time series data pair. For each position, the minimum cumulative distance of the previous position (which can be the adjacent position above, left or upper left) is selected, and the distance of the current time series pair is added to the minimum cumulative distance. Through this process, the value of each position in the initial cumulative matrix gradually becomes the minimum cumulative distance from the starting point to that position. After multiple updates, the final cumulative distance matrix represents the minimum total distance of the optimal path from the starting position to each position.

[0061] From the end of the matrix (i.e. the last position of the target time series data pair), look at the adjacent previous position, choose the direction with the minimum cumulative distance (usually the upper, left or upper left adjacent position), and take this position as the next step of backtracking. By constantly backtracking to the starting position, a path is finally obtained, which represents the optimal alignment path from the starting point to the end point, i.e. the alignment scheme with the minimum cumulative distance among all possible paths.

[0062] By backtracking the information in the path, the time points of the physiological time series data are gradually adjusted so that the two time series data can be aligned on the time axis with the minimum distance. For each matched time point, the position of one of the time series data is adjusted so that the difference between the two data points is minimized. Each time point in the physiological time series data is accurately aligned, eliminating the deviation or misalignment in time, thereby obtaining the final time series alignment data.

[0063] By standardizing the medical image data and physiological time series data respectively, the consistency and comparability of the data can be significantly improved, thereby improving the accuracy and effectiveness of subsequent analysis. The process of normalizing medical images through pixel value normalization and image enhancement helps to remove noise in the image and highlight important features, so that doctors or algorithms can more accurately identify disease lesions. The alignment of physiological time series data accurately aligns data at different time points through the calculation of time series distance matrix and cumulative matrix, eliminating the deviation in time, so that physiological data from different sources can be directly compared and fused, providing more accurate health assessment, improving the quality and reliability of the data.

[0064] S2, convolve the normalized medical image to obtain lesion spatial features, and generate a health risk level using the lesion spatial features.

[0065] In the embodiments of the present application, the normalized medical image is subjected to convolution operation to extract lesion feature maps, which can reflect the spatial information of the lesion region. In order to highlight the key parts in the lesion region, the attention map of the lesion region is calculated, and the lesion feature maps are weighted and fused to obtain more accurate lesion spatial features. Through global average pooling and full connection operation, lesion risk score is generated according to the lesion spatial features, reflecting the health risk of the lesion region. Finally, according to the lesion risk score, the lesion spatial features are classified into grades, and finally the health risk level is obtained.

[0066] In the specific scenario of medical health, it can be applied to an automatic disease diagnosis system. For example, the convolution processing of medical image data (such as CT and MRI images) can help identify lesion areas (such as tumors and blood clots), highlight the most critical lesion areas, extract lesion spatial features, provide more accurate disease risk assessment for doctors, and automatically label potential risk areas in large-scale screening. Combined with global pooling and fully connected processing, the final health risk level can help doctors quickly assess the disease and develop personalized treatment plans. In addition, these technologies can also play a key role in early diagnosis and prediction of disease development, improving the accuracy and efficiency of medical services.

[0067] In the specific scenario of financial technology, it can be applied to fraud detection and customer risk assessment. For example, banks can identify abnormal transaction patterns or potential risk points by performing convolution analysis on customer transaction behavior data, and attention mechanisms can help the system focus on key features in transactions, such as large transactions and frequent transactions. Through convolution processing and weighted fusion of attention maps, the system can more accurately identify high-risk transactions and assess customer credit risk in real time. Ultimately, the risk score derived from comprehensive analysis not only helps optimize credit decisions, but also effectively reduces financial fraud and credit risk, improving the security and reliability of financial services.

[0068] Figure 3 A flowchart of a medical image convolution process in a business analysis method based on health data according to an embodiment of the present application is provided.

[0069] In the embodiment of the present application, the normalized medical image is convolved to obtain lesion spatial features, and the lesion spatial features are used to generate a health risk level, including:

[0070] The lesion area of the normalized medical image is divided to obtain a lesion area image;

[0071] The lesion area image is convolved to obtain a lesion feature image;

[0072] The attention score of each region in the lesion area image is determined, and the attention score is used to generate an attention map;

[0073] The lesion feature image is weighted and fused using the attention map to obtain lesion spatial features;

[0074] The lesion spatial features are globally averaged and pooled to obtain a pooled feature vector;

[0075] The pooled feature vector is fully connected to obtain a lesion risk score;

[0076] According to the lesion risk score, the lesion spatial features are graded to obtain a health risk level.

[0077] In detail, the lesion region in the medical image is separated from the background by image segmentation technology. Common segmentation methods include threshold-based segmentation, region growing, edge detection, or deep learning methods such as U-Net network, which can automatically detect and identify abnormal regions related to diseases in the image, such as tumors, nodules, or other lesion regions. The pixel values of the image are analyzed according to the pre-set standard to ensure that only the region where the lesion is located is extracted, and the normal tissue part is excluded. After segmentation processing, the lesion region image obtained contains the lesion region in the medical image.

[0078] The convolutional layer operates on the image by sliding one or more convolutional kernels (filters), gradually extracting local features in the lesion region image. These convolutional kernels identify important spatial features in the lesion region image, such as edges, corners, textures, and shapes. Through the operation of multiple convolutional layers, the network can gradually extract information from low-level features (such as edges and corners) to high-level features (such as the specific morphology of tumors or lesion regions). After processing by the convolutional layer, the lesion feature map obtained contains deep features of the lesion region.

[0079] Using the lesion feature map extracted by the convolutional neural network (CNN) and combining the attention mechanism network (such as self-attention mechanism or SE-Net), an attention score is assigned to each position, representing the importance of each position in the image to the diagnostic result. It is usually achieved by calculating the relationship between local features and global features, which can automatically focus on the most critical part of the lesion region and ignore irrelevant areas. According to these attention scores, an attention map is generated, where the value of each pixel position represents the attention weight of each position in the image. The attention map can clearly show the most important part of the lesion region.

[0080] The score of each position in the attention map is used to adjust the feature value of the corresponding position in the lesion feature map, thereby enhancing the features of important regions while suppressing the influence of irrelevant regions. The weighted fusion method can highlight the most critical part of the lesion region, ensuring that the network can pay more attention to the regions that are important to the diagnostic result. After weighted fusion, the lesion spatial features obtained are a refined feature map.

[0081] The global average pooling compresses the size of the feature map effectively, reduces the computational complexity, and retains the global information of each channel, thereby obtaining a smaller feature vector. After global average pooling, the pooled feature vector contains the global information of the lesion region.

[0082] The pooled feature vector is input into the fully connected layer. The fully connected layer performs linear transformation on the feature vector through a series of weight and bias parameters, and introduces nonlinearity through an activation function (such as ReLU or Sigmoid) to enhance the expression ability of the model. It can map the global features after pooling to a higher dimensional space, thereby capturing more complex patterns and relationships. After processing by the fully connected layer, the final output is a lesion risk score, reflecting the health risk level of the lesion region.

[0083] A plurality of risk level intervals (such as low risk, medium risk, high risk, etc.) are defined, and the risk level is determined according to the size of the lesion risk score. For example, if the lesion risk score is below a certain threshold, the lesion risk is classified as low risk; if the lesion risk score is higher, the lesion risk is classified as high risk; this can convert the continuous lesion risk score into discrete health risk levels, which not only helps to quickly identify high-risk patients, but also allows personalized treatment or monitoring strategies to be developed based on the patient's health status.

[0084] The division of the lesion region and the convolution operation help to extract key lesion features in the image, thereby effectively identifying and locating the lesion. The weighted fusion of the lesion feature map using the attention mechanism can focus on the most important lesion area and ignore irrelevant parts, improving the accuracy of diagnosis. After global average pooling, the feature vector of the lesion is simplified, but still retains important information about the disease, providing a stable foundation for subsequent lesion risk scoring. The lesion risk score obtained through full connection processing can be accurately classified according to the health risk level, helping doctors to identify high-risk patients in a timely manner and develop personalized treatment plans. This not only helps doctors improve their diagnostic efficiency, but also provides more accurate and personalized health management services for patients.

[0085] S3, performing data trend analysis on the time series alignment data to obtain a time series risk score.

[0086] In the embodiments of the present application, a preset long short-term memory (LSTM) network layer is used to capture the time dependence in the data, that is, to remember and learn the information at past time points, so as to understand the dynamic change trend of the data. The extracted time dependence is input into a multi-layer fully connected network for further processing to obtain a predicted value of the time series alignment data. By visualizing the time series alignment data and the predicted value, a data trend chart is generated to intuitively show the law and fluctuation of the data change over time. Finally, the data trend chart is analyzed for volatility to generate a time series risk score to evaluate the potential risks or abnormal fluctuations in the data.

[0087] In the medical health specific scenario, the time series data analysis technology, especially the time dependence capture based on the long short-term memory (LSTM) network, is of great significance for disease monitoring and early warning. For example, LSTM can analyze physiological data of patients (such as electrocardiogram, blood glucose level, blood pressure, etc.), identify the time-related trends therein, and predict the development trend of the disease. Through the volatility analysis of these time series data, doctors can evaluate the fluctuation range of the patient's health status, and then identify potential risks (such as sudden heart attack or acute onset of diabetes). In this way, hospitals or medical institutions can provide personalized risk assessment for patients, take preventive measures in advance, and improve the effect of prevention and treatment.

[0088] In the financial technology specific scenario, it can be applied in risk assessment and credit scoring systems. For example, banks and financial institutions use LSTM network to analyze time series data such as customers' transaction history, credit card consumption records, loan repayment situations, etc., to capture customers' financial behavior patterns and time dependence. These technologies help institutions predict customers' future credit risks and identify possible default behaviors. In terms of volatility analysis, by monitoring the fluctuations of customers' financial data, financial institutions can assess the credit stability of customers in real time, so as to adjust the credit limit or make risk control decisions in time. Such data trend analysis not only enhances the accuracy of financial decisions, but also helps to prevent financial risks and improve the safety and reliability of financial services.

[0089] In the embodiments of the present application, the data trend analysis on the time series alignment data to obtain a time series risk score comprises:

[0090] The time dependence of the time series alignment data is captured by using a preset long short-term memory layer;

[0091] The time dependence is multi-layer fully connected to obtain a data predicted value of the time series alignment data;

[0092] The time series alignment data and the data predicted value are visually analyzed to obtain a data trend chart;

[0093] Perform volatility analysis on the data trend chart, and generate a time series risk score based on the volatility analysis.

[0094] In detail, the time series alignment data is input into the LSTM network, and the LSTM can effectively process and store data dependencies over long time spans through its unique gating mechanism (including the forget gate, input gate, and output gate). By preserving and updating the hidden state and cell state at each time step, long-term and short-term dynamic changes in time series data are captured, enabling the LSTM to learn potential patterns in time series data, such as periodic changes and trend fluctuations. In particular, when there are long-term dependencies in the data, the LSTM layer can automatically identify key time points and trends in the time series data.

[0095] The time-dependent features extracted from the LSTM network are input into a multi-layer fully connected network. Each fully connected layer performs a linear transformation on the input data through a set of weights and biases, and applies an activation function to introduce non-linear relationships, further enhancing the model's ability to learn complex patterns. The multi-layer fully connected network can capture deeper relationships from the time series features extracted by the LSTM and map this information to the prediction space. Through continuous forward propagation, the network finally outputs the predicted values of the time series alignment data, which represent the possible trends or changes of the time series data at future time steps.

[0096] The original time series alignment data is combined with the future data values predicted by the fully connected layer to form a complete set of time series data. Visualization tools such as Matplotlib, Seaborn, etc. are used to plot these data as a curve or line chart with time as the horizontal coordinate and data values as the vertical coordinate. By displaying both actual observed data and predicted data, the trend line in the chart can visually show the rules of data changes over time and demonstrate the differences between actual data and predicted data.

[0097] By observing the fluctuation amplitude between actual data and predicted data in the data trend chart, the change frequency and volatility pattern of the data are identified. Volatility is usually quantified by calculating the difference between data points, standard deviation, or fluctuation amplitude, with particular attention to sudden or abnormal fluctuations in the data. These fluctuations may represent potential risks or instability factors. After analysis, the time series risk score is converted according to the size and trend of volatility. Larger volatility means higher risk, and smaller volatility indicates system stability. By mapping these volatility values to a pre-set risk level interval, the final time series risk score is generated.

[0098] By conducting data trend analysis on the time-aligned data, the prediction ability of future data trends and potential risks can be significantly improved. The use of Long Short-Term Memory (LSTM) layers to capture time dependencies helps to extract key temporal features from historical data, thereby enhancing the model's sensitivity to long-term trends and short-term fluctuations. Multi-layer fully connected further enhances the model's non-linear modeling capabilities, ensuring accurate capture of complex data patterns. By visualizing the data trend chart, the changing trend of the data can be observed directly, facilitating the timely discovery of abnormal fluctuations or potential risks. Volatility analysis provides an important evaluation of data stability, with high volatility generally indicating higher risk and low volatility generally indicating better data stability. By generating a time-series risk score, it helps to make more accurate and timely risk management and intervention measures.

[0099] S4, weighting and fusing the health risk level and the time-series risk score to obtain the health risk score of the target customer.

[0100] In the embodiments of the present application, the health risk level is based on medical image analysis and physiological time-series data, reflecting the overall health status of the customer. The time-series risk score is obtained by analyzing the future health data trend after trend analysis of the past 30 days of blood glucose and heart rate data through the LSTM model. A comprehensive health risk score is generated by weighting and fusing the two scores. Each score is assigned a weight, and the sum of the weights of the health risk level and the time-series risk score is 1. The proportion of the two can be adjusted according to actual needs (for example, the health risk level accounts for 40%, and the time-series risk score accounts for 60%). The two scores are fused using the weighted average formula to obtain the health risk score of the target customer, and the customer is divided into low, medium and high risk levels according to the score results, providing a basis for subsequent health management and intervention measures.

[0101] In the specific scenario of medical health, the health risk level obtained through medical image analysis can help doctors understand the current health status of patients, such as whether there are serious lesions or chronic diseases. The time-series risk score is based on physiological data such as blood glucose and heart rate, and can predict the trend of future health status. After weighting and fusing the two scores, the medical institution can accurately divide patients into low, medium and high risk levels and develop corresponding intervention measures. For example, high-risk patients may need more monitoring and more frequent treatment, while low-risk patients can maintain their health status through regular health checks.

[0102] In the specific scenario of financial technology, by analyzing the customer's medical images to obtain a health risk level, financial institutions can assess whether the customer has potential high-risk health problems. By analyzing time series data such as blood sugar and heart rate, an LSTM model can be used to predict future health trends, which can provide insurance companies with a more comprehensive health assessment. By weighting and fusing these two scores, insurance companies can adjust the premium or insurance policy of customers based on the comprehensive health risk score, ensuring the accuracy of risk pricing. For example, high-risk customers may face higher premiums, while low-risk customers enjoy more favorable insurance conditions.

[0103] The weighted fusion of health risk level and time series risk score can provide a more comprehensive and accurate health assessment. By combining medical image analysis and physiological time series data trend prediction, the health status of customers can be more comprehensively reflected, not only considering the current health status (such as disease risk), but also predicting future health trends (such as potential chronic diseases or sudden illnesses). This weighted fusion method combines the advantages of different data sources, improving the accuracy and reliability of risk assessment.

[0104] S5、Obtain several insurance policy adjustment schemes, and screen the insurance policy adjustment schemes according to the health risk score to obtain a target adjustment strategy.

[0105] In the embodiment of the present application, by analyzing the matching degree of the health risk score and the health status of the target customer, the customer health risk matching degree is obtained, and the fitness function is constructed combined with historical data, the population is initialized based on the genetic algorithm, and the fitness value of each scheme is calculated by using the fitness function, and the most potential parent scheme is screened out for cross processing, thereby generating new child insurance policy schemes. In the child scheme, a small perturbation is introduced to introduce diversity, and the health risk score matching analysis of the updated insurance policy scheme is performed. Through multiple iterations and optimization, the optimal insurance policy scheme is selected until the preset return times are met, and finally the target adjustment strategy most suitable for the customer is obtained.

[0106] In the specific scenario of medical health, it can be used to customize personalized medical insurance schemes or health management strategies according to the health risk score and needs of patients. By analyzing the health risk score of patients (such as the risk of disease occurrence, treatment needs, etc.), different insurance policy adjustment schemes can be screened and optimized. For example, some high-risk patients may need more health management services (such as rehabilitation plans or regular check-ups), while low-risk patients may only need basic health protection services. By optimizing the insurance policy scheme through genetic algorithm, it can be continuously adjusted and improved to ensure that each patient is provided with the best medical protection plan, while improving patient satisfaction and protection effect.

[0107] In the specific scenario of financial technology, it can be used for insurance product customization and risk assessment system. Based on the health risk score of the customer and the historical claim data, the system can automatically adjust and optimize the insurance product clauses. For example, for customers with high health risks, higher insurance plans can be recommended, covering more medical services or accidental risks. By continuously optimizing the insurance plan, insurance companies can provide personalized protection services while improving customer satisfaction and the profitability of insurance products. The application of genetic algorithm allows each adjustment to consider multiple factors, such as customer health matching, customer satisfaction, and insurance plan profitability, thereby creating more value for both insurance companies and customers.

[0108] In the embodiments of the present application, the target adjustment strategy is obtained by screening the insurance plan adjustment scheme according to the health risk score.

[0109] Obtain the health data of the target customer, and perform matching degree analysis on the health risk score and the health data to obtain the customer health risk matching degree.

[0110] Obtain the historical customer satisfaction and insurance plan profitability of the insurance plan adjustment scheme, and construct a fitness function using the customer health risk matching degree, the historical customer satisfaction, and the insurance plan profitability.

[0111] A number of insurance plan adjustment schemes are used as an initial population.

[0112] The fitness value of each insurance plan adjustment scheme in the initial population is determined using the fitness function.

[0113] According to the fitness value, the parent insurance plan scheme greater than the preset fitness threshold is screened out.

[0114] The parent insurance plan scheme is subjected to cross processing to obtain a child insurance plan scheme.

[0115] The child insurance plan scheme is subjected to slight disturbance to obtain an updated insurance plan scheme.

[0116] The updated insurance plan scheme and the health risk score are subjected to matching degree analysis, and the optimal insurance plan scheme is screened out according to the analysis result.

[0117] The optimal insurance plan scheme is returned to the step of determining the fitness value of each insurance plan adjustment scheme in the initial population using the fitness function, and the number of returns is counted.

[0118] When the number of returns reaches a preset number, stop returning, and take the final optimal insurance plan scheme as the target adjustment strategy.

[0119] In detail, by calculating the difference between the health risk score and the customer's health characteristics, a matching value is obtained, which represents the consistency of the health risk score with the customer's actual health status, helping to identify which customers' health risk assessment is most accurate and which customers' risk prediction may be biased. The customer health risk matching degree obtained will provide an important basis for subsequent policy adjustment and the development of personalized health management plans, ensuring that health management strategies are highly consistent with customer needs.

[0120] By analyzing historical customer satisfaction data, the acceptance and satisfaction of different policy adjustment schemes by customers are evaluated, and the economic benefits brought by each policy adjustment scheme to the insurance company are measured in combination with the profitability data of the policy. The customer health risk matching degree, historical customer satisfaction, and policy profitability are weighted and fused, and the overall performance of each policy adjustment scheme is quantified through the fitness function. The output value of the fitness function represents the overall pros and cons of each scheme in multiple dimensions, helping to identify the best policy adjustment scheme that can meet customer health needs, improve customer satisfaction, and bring considerable benefits to the company.

[0121] Each policy adjustment scheme is input into the fitness function as an individual, and each scheme is evaluated based on factors such as customer health risk matching degree, historical customer satisfaction, and policy profitability. The output value of the fitness function reflects the pros and cons of each policy adjustment scheme in multiple evaluation dimensions. A higher fitness value means better overall performance, which is more in line with customer needs, can improve customer satisfaction, and bring more benefits to the insurance company.

[0122] These parent schemes are subjected to crossover processing to generate child policy schemes. The crossover processing process simulates gene recombination in nature by randomly selecting some key features (such as health risk matching degree, customer satisfaction, and policy profitability) in the parent schemes and exchanging and combining them to generate new child policy schemes. These child schemes inherit the advantages of the parent schemes, while introducing new combinations through the crossover process, which may help to explore better policy adjustment strategies. Through multiple crossover operations, the generated child policy schemes gradually approach the optimal solution.

[0123] Some key parameters (such as policy clauses, coverage range, and customer matching degree) in each child scheme are subjected to slight random adjustments. These minor perturbations can increase the diversity of the population by changing the values of the parameters slightly or introducing small random changes in some features of the policy scheme. The purpose is to avoid the algorithm falling into a local optimal solution and increase the opportunity to explore new solutions, while ensuring that these changes do not cause large fluctuations, thereby affecting the stability of the scheme.

[0124] Based on the customer's health risk level (e.g., low, medium, high risk) and the updated policy terms, we calculate the matching value of each plan with the customer's health risk score. A highly matched plan means it can better meet the customer's health needs while also taking into account the customer's personalized requirements, allowing us to screen out the optimal plan.

[0125] Through continuous iterative optimization, the quality of the policy adjustment plan is improved with each iteration. The number of times the optimal solution is returned is counted. When the number of returns reaches the preset maximum, it indicates that the solution has converged or reached the optimal optimization result. The iteration process is stopped, and the final optimal policy solution is used as the target adjustment strategy. This iterative optimization method ensures that the policy adjustment plan, after multiple rounds of screening and optimization, can strike a balance between customer health risks, customer satisfaction, and policy profitability, thereby formulating the policy adjustment strategy that best meets customer needs.

[0126] By analyzing the compatibility between health risk scores and customer health conditions, we can accurately identify customers' health needs and ensure that insurance plans best meet these needs. Leveraging the fitness function within a genetic algorithm, through multiple rounds of screening, crossover, perturbation, and evaluation, we gradually optimize insurance plans, ensuring that the final plan not only meets customers' health needs, but also improves customer satisfaction and maximizes profits for the insurance company. This iterative process not only identifies the optimal policy adjustment plan, but also continuously optimizes its compatibility and profitability, ensuring that every customer receives the most accurate and high-quality insurance protection.

[0127] S6. Push services to the target customers according to the target adjustment strategy, collect multi-dimensional behavior data of the target customers after the service push, track and analyze the multi-dimensional behavior data, and obtain the strategy execution effect.

[0128] In this embodiment of the present invention, a simulated policy is generated by simulating the stored data of target customers. This policy is then compared with the annotated adjustment policy, and the insurance service for the target adjustment policy is obtained based on the comparison results. The insurance service is pushed to the preset target client, and the returned multi-dimensional behavioral data is collected. A customer profile is generated based on this multi-dimensional behavioral data. The push effect is analyzed based on the customer profile, thereby determining the policy execution effect.

[0129] In the medical health specific scenario, by comparing the simulated strategy with the annotated adjustment strategy, the medical institution can accurately formulate a personalized health management plan for the patient. For example, for high-risk patients (such as diabetic patients), the system can simulate different treatment and health management plans, compare the effects of these plans, and select the most appropriate plan for the patient based on the comparison results of the strategies. At the same time, by collecting multi-dimensional behavior data of the patient after the health management service (such as health behavior, treatment compliance, etc.), the system can construct a detailed customer portrait and analyze the behavior pattern. These data will provide real-time feedback on the treatment effect for medical providers, helping doctors adjust the treatment plan according to the feedback data of the patient, ensuring the effectiveness and timeliness of the treatment strategy.

[0130] In the financial technology specific scenario, by comparing the simulated strategy with the annotated adjustment strategy, the insurance company can formulate more refined insurance services. For example, the system can simulate different insurance plans (such as health insurance, life insurance, etc.), and compare these simulated strategies with the annotated adjustment strategies (based on customer health risk assessment). According to the comparison results, the insurance company can select the most suitable insurance product for promotion. By collecting multi-dimensional behavior data of the customer after receiving the insurance service (such as click rate, purchase decision, renewal intention, etc.), a precise customer portrait can be constructed to analyze the promotion effect. Based on these data analysis, the insurance company can adjust the marketing strategy and product pricing in real time, optimize the customer experience, and improve customer satisfaction.

[0131] In the embodiment of the present application, the service push to the target customer according to the target adjustment strategy, and the collection of multi-dimensional behavior data of the target customer after the service push, comprises:

[0132] Annotating the target adjustment strategy to obtain an annotated adjustment strategy;

[0133] Obtaining the storage data of the target customer, simulating the strategy based on the storage data to obtain a simulated strategy;

[0134] Comparing the simulated strategy with the annotated adjustment strategy, and determining whether the simulated strategy is consistent with the annotated adjustment strategy according to the comparison result;

[0135] If the simulated strategy is not consistent with the annotated adjustment strategy, the annotated adjustment strategy is optimized to obtain an optimized insurance service of the annotated adjustment strategy;

[0136] If the simulated strategy is consistent with the annotated adjustment strategy, the insurance service of the annotated adjustment strategy is obtained;

[0137] The insurance service is pushed to the preset target client, and the multi-dimensional behavior data of the target client returned by the preset target client is collected.

[0138] In detail, the existing target adjustment strategy is analyzed and annotated in detail, and additional annotation information is added at each key link and decision point of the strategy. These annotations can include the applicable conditions of the strategy, the execution steps, the expected targets, the potential risks and their countermeasures, and the adjustment methods to be taken in specific situations. By adding these annotations to the target adjustment strategy, not only can the execution logic of the strategy be better recorded and understood, but also the transparency of the strategy can be improved, ensuring that the implementation process can quickly adapt to changes in customer needs and optimize or adjust when necessary.

[0139] Information is extracted from the customer's historical health records, behavior data, transaction data, and other dimensions, and these data are analyzed to identify potential influencing factors. Data simulation algorithms such as Monte Carlo simulation, regression models, etc. are used to simulate health changes or customer behavior patterns under different conditions, generating a series of possible strategies. For example, for health management services, simulation strategies can be based on different treatment plans and intervention measures to predict the potential impact on customer health.

[0140] The key elements of the simulation strategy and the annotated adjustment strategy are compared one by one to analyze whether the predicted customer behavior or health changes in the simulation strategy are consistent with the preset targets and methods in the annotated adjustment strategy. The results generated by the simulation strategy reflect the customer's response under specific conditions, while the annotated adjustment strategy is based on expert-set rules and health management goals. Based on the comparison results, the matching degree of the core strategy, execution steps, and expected effects of the two is judged.

[0141] If the simulation strategy and the annotated adjustment strategy are inconsistent, it means that the existing annotated adjustment strategy does not fully match the simulation results. The annotated adjustment strategy is optimized, including re-examining the key parameters and decision points in the strategy, adjusting the content of the strategy based on the feedback of the simulation data, such as modifying the insurance policy clauses, adding additional protection, or adjusting the customer matching degree, etc., to improve the accuracy and adaptability of the strategy. The optimized annotated adjustment strategy will be used to develop new insurance services.

[0142] If the simulation strategy and the annotated adjustment strategy are consistent, it means that the existing annotated adjustment strategy performs well in meeting customer needs and expected effects, and the system will directly obtain the insurance service based on the consistent annotated adjustment strategy.

[0143] The optimized insurance service information is delivered to the customers in a timely manner through various channels such as application push, SMS, email, etc. The push content usually includes personalized insurance product recommendations, policy terms, preferential information, etc., aiming to guide the customers to take the next action. After receiving the push information, the customers interact according to their interests and needs, and the multi-dimensional behavior data of the customers such as click rate, browsing time, purchase behavior, consultation record, etc. are collected, reflecting the customers' reaction to the push content and their purchase decision-making tendency, providing an important basis for subsequent strategy optimization and customer analysis.

[0144] By adding annotations to the target adjustment strategy and conducting data simulation, the insurance service can be more accurately predicted and optimized, ensuring that the pushed content meets the customers' needs to the greatest extent. By comparing the simulation strategy with the annotated adjustment strategy, the effectiveness of the insurance service can be further verified. If they are inconsistent, the system will optimize the strategy to ensure that the final pushed service is highly consistent with the customers' needs. After the insurance service is pushed, the system collects multi-dimensional behavior data of the customers such as click, interaction, purchase, etc., and tracks the customers' reaction and demand changes in real time. This feedback mechanism not only improves customer satisfaction, but also enables the insurance company to more accurately optimize service content and improve service efficiency and effectiveness.

[0145] In the embodiment of the present application, the tracking analysis of the multi-dimensional behavior data to obtain the strategy execution effect comprises:

[0146] Standardizing the multi-dimensional behavior data to obtain standard behavior data;

[0147] Extracting key analysis data from the standard behavior data;

[0148] Generating a customer portrait of the target customer according to the key analysis data;

[0149] Analyzing the push effect of the key analysis data according to the customer portrait to determine the strategy execution effect.

[0150] In detail, the multi-dimensional behavior data of different sources and formats such as click rate, browsing time, purchase decision, interaction frequency, etc. are uniformly converted into the same dimension and scale, which is realized through normalization or standardization algorithm. The purpose is to eliminate the dimensional difference and extreme value interference of the data on the analysis result. For example, by mapping the maximum and minimum values of the data to the [0, 1] interval, or Z-score standardization, the mean value is 0 and the standard deviation is 1.

[0151] Identifying the most representative features or variables from standardized behavioral data can reflect customers' behavior patterns and needs. For example, it may include customers' click frequency, page dwell time, purchase conversion rate, browsing path, interaction frequency, etc. Through data analysis methods such as feature selection, principal component analysis (PCA) or correlation analysis, key data that have the greatest impact on customer behavior prediction and strategy optimization can be screened out, which can often reveal customers' preferences, interests, and potential purchase intentions, providing important basis for subsequent customer portrait construction and personalized service.

[0152] Integrating these key data (such as customers' interaction frequency, purchase history, browsing behavior, preferences, etc.) to build a multi-dimensional information file of customers. Through further analysis of these data, customers' core features such as customers' interest areas, consumption habits, product preferences, health conditions, etc. can be identified. Using machine learning algorithms (such as clustering analysis or classification models), customers can be classified or grouped according to their behavior and characteristics, forming customer portraits with high recognition. These customer portraits not only include customers' basic information, but also cover customers' behavior patterns, potential needs and risk prediction, providing strong support for personalized service push, precision marketing and strategy optimization.

[0153] By comparing the core features in the customer portrait with the pushed service content, it is evaluated whether the push meets the customer's interests, needs and behavior patterns, and the customer's response to the push content is tracked, such as click rate, purchase conversion rate, interaction frequency, etc. The difference between these behavior data and the expected response in the customer portrait is analyzed. Through this comparative analysis, it can be determined whether the push content successfully stimulates the customer's interest, whether it can promote the customer to take action, or whether there are parts that do not meet the customer's needs. Finally, through the evaluation of these data, a strategy execution effect report is generated.

[0154] Through tracking analysis of multi-dimensional behavior data, the system can monitor and evaluate the implementation effect of the strategy in real time, ensuring that the service push is highly consistent with the customer's needs. After extracting key analysis data, the core features of customer behavior can be identified, and accurate customer portraits can be further generated, which provides the basis for the formulation of personalized strategies. By comparing the customer portrait with the push effect, the accuracy and effectiveness of the push strategy can be evaluated, helping to identify which content has attracted the attention of customers and which strategies need to be optimized. Ultimately, this tracking analysis mechanism not only improves the efficiency and conversion rate of service push, but also provides data support for subsequent strategy optimization, enhances customer satisfaction, and improves business results.

[0155] S7, optimizing the target adjustment strategy using the strategy execution effect to obtain target insurance data.

[0156] In the embodiment of the present application, the performance of the current strategy in actual application is evaluated by analyzing the multi-dimensional behavior data after pushing and the effect of strategy implementation. For example, the customer's participation, satisfaction, conversion rate and other indicators will be the basis for optimization. If the existing strategy fails to achieve the expected effect, the key parameters in the target adjustment strategy will be adjusted based on the feedback data, such as the pricing, coverage, service terms of the insurance product, and the matching degree of the customer portrait and the strategy will be re-examined to ensure that the insurance plan better meets the needs of customers. Machine learning algorithms can be used for automatic adjustment to make the strategy more personalized and precise. The optimized target adjustment strategy will better meet the needs of customers and improve the execution effect of the strategy and the revenue of the business.

[0157] The target adjustment strategy is optimized using the effect of strategy implementation, which can effectively solve the problem of system fragmentation and data island in health data application. By integrating multi-dimensional customer behavior data and health data, efficient fusion and dynamic linkage of data from different sources can be achieved, eliminating the obstacles of data islands. The optimized target adjustment strategy can respond to customer needs in real time and dynamically adjust insurance services according to the latest health data and behavior feedback, improving the efficiency and accuracy of service response. This data-driven optimization approach ensures that various health data can work together to provide faster and more personalized health management solutions, improving customer experience and service quality.

[0158] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and 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 embodiment of the present application.

[0159] As shown in Figure 4 , it is a functional module diagram of a business analysis device based on health data provided by an embodiment of the present application.

[0160] In the embodiment of the present disclosure, a business analysis device based on health data is provided, which corresponds one-to-one with the above-mentioned embodiment of a business analysis method based on health data. As shown in Figure 4 , the business analysis device based on health data 100 can be installed in an electronic device, and according to the implemented functions, the business analysis device based on health data 100 includes a data standardization module 101, a medical image convolution module 102, a data trend analysis module 103, a data weighted fusion module 104, a policy scheme screening module 105, a data tracking analysis module 106, and an adjustment strategy optimization module 107. The detailed description of each functional module is as follows:

[0161] The data standardization module 101 is configured to obtain medical image data and physiological time series data of a target customer, and perform standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data.

[0162] The medical image convolution module 102 is configured to perform convolution on the normalized medical images to obtain lesion spatial features, and generate a health risk level by using the lesion spatial features.

[0163] The data trend analysis module 103 is configured to perform data trend analysis on the time series alignment data to obtain a time series risk score.

[0164] The data weighted fusion module 104 is configured to perform weighted fusion on the health risk level and the time series risk score to obtain a health risk score of the target customer.

[0165] The insurance policy scheme screening module 105 is configured to obtain a plurality of insurance policy adjustment schemes, screen the insurance policy adjustment schemes according to the health risk score, and obtain a target adjustment strategy.

[0166] The data tracking analysis module 106 is configured to push services to the target customer according to the target adjustment strategy, collect multi-dimensional behavior data of the target customer after the service push, perform tracking analysis on the multi-dimensional behavior data, and obtain a strategy execution effect.

[0167] The adjustment strategy optimization module 107 is configured to optimize the target adjustment strategy by using the strategy execution effect, and obtain target insurance data.

[0168] In an embodiment, the data standardization module 101 performs standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data, including:

[0169] Performing image enhancement on the medical image data to obtain medical enhanced images;

[0170] Performing normalization processing on pixel values of the medical enhanced images to obtain normalized medical images;

[0171] Randomly selecting two groups of the physiological time series data as target time series data pairs;

[0172] Determining distances between the target time series data pairs, and constructing a time series distance matrix according to the distances;

[0173] Obtaining an initial cumulative matrix, and determining a minimum cumulative distance of each matrix position by using the initial cumulative matrix and the time series distance matrix;

[0174] updating the initial accumulation matrix according to the minimum accumulated distance to obtain an accumulated distance matrix;

[0175] backtracking the accumulated distance matrix to obtain an optimal alignment path;

[0176] aligning the physiological time series data using the optimal alignment path to obtain time series aligned data.

[0177] In an embodiment, the medical image convolution module 102 performs convolution on the normalized medical image to obtain a lesion spatial feature, and generates a health risk level using the lesion spatial feature, including:

[0178] dividing the lesion area of the normalized medical image to obtain a lesion area image;

[0179] convolving the lesion area image to obtain a lesion feature image;

[0180] determining an attention score of each region in the lesion area image, and generating an attention map using the attention score;

[0181] weighting and fusing the lesion feature image using the attention map to obtain a lesion spatial feature;

[0182] performing global average pooling on the lesion spatial feature to obtain a pooling feature vector;

[0183] performing full connection processing on the pooling feature vector to obtain a lesion risk score;

[0184] dividing the lesion spatial feature into grades according to the lesion risk score to obtain a health risk level.

[0185] In an embodiment, the data trend analysis module 103 performs data trend analysis on the time series aligned data to obtain a time series risk score, including:

[0186] using a preset long short-term memory layer to capture the time dependence of the time series aligned data;

[0187] performing multi-layer full connection on the time dependence to obtain a data prediction value of the time series aligned data;

[0188] performing visual analysis on the time series aligned data and the data prediction value to obtain a data trend chart;

[0189] performing volatility analysis on the data trend chart, and generating a time series risk score according to the volatility obtained by the analysis.

[0190] In an embodiment, the insurance policy scheme screening module 105 screens the insurance policy adjustment scheme according to the health risk score to obtain a target adjustment strategy, including:

[0191] Obtaining health data of the target customer, and performing matching degree analysis on the health risk score and the health data to obtain a customer health risk matching degree;

[0192] Obtaining historical customer satisfaction and insurance policy profitability of the insurance policy adjustment scheme, and constructing an fitness function by using the customer health risk matching degree, the historical customer satisfaction and the insurance policy profitability;

[0193] Taking a plurality of the insurance policy adjustment schemes as an initial population;

[0194] Determining an fitness value of each insurance policy adjustment scheme in the initial population by using the fitness function;

[0195] Screening a parent insurance policy scheme greater than a preset fitness threshold according to the fitness value;

[0196] Performing cross processing on the parent insurance policy scheme to obtain a child insurance policy scheme;

[0197] Performing slight perturbation on the child insurance policy scheme to obtain an updated insurance policy scheme;

[0198] Performing matching degree analysis on the updated insurance policy scheme and the health risk score, and screening an optimal insurance policy scheme according to a scheme matching degree obtained by the analysis;

[0199] Returning the optimal insurance policy scheme to the step of determining the fitness value of each insurance policy adjustment scheme in the initial population by using the fitness function, and counting a return frequency;

[0200] When the return frequency reaches a preset frequency, stopping returning, and taking a final optimal insurance policy scheme as the target adjustment strategy.

[0201] In an embodiment, the data tracking analysis module 106 performs service pushing on the target customer according to the target adjustment strategy, and collects multi-dimensional behavior data of the target customer after the service pushing, including:

[0202] Adding an annotation to the target adjustment strategy to obtain an annotated adjustment strategy;

[0203] Obtaining storage data of the target customer, and performing strategy simulation on the storage data to obtain a simulated strategy;

[0204] Comparing the simulated strategy with the annotated adjustment strategy, and judging whether the simulated strategy is consistent with the annotated adjustment strategy according to a comparison result;

[0205] If the simulation strategy is inconsistent with the annotation adjustment strategy, the annotation adjustment strategy is optimized to obtain an insurance service of the optimized annotation adjustment strategy;

[0206] If the simulation strategy is consistent with the annotation adjustment strategy, an insurance service of the annotation adjustment strategy is obtained;

[0207] The insurance service is pushed to a preset target client, and multi-dimensional behavior data of the target customer returned by the preset target client is collected.

[0208] In an embodiment, the data tracking analysis module 106 performs tracking analysis on the multi-dimensional behavior data to obtain a strategy execution effect, including:

[0209] The multi-dimensional behavior data is standardized to obtain standard behavior data;

[0210] Key analysis data in the standard behavior data is extracted;

[0211] A customer portrait of the target customer is generated according to the key analysis data;

[0212] The push effect of the key analysis data is analyzed according to the customer portrait to determine the strategy execution effect.

[0213] In the present application, for a health data-based business analysis device, first, the present application obtains medical image data and physiological time series data of a target customer, standardizes the medical image data and the physiological time series data respectively, obtains normalized medical images and time series aligned data, which helps to remove noise in the images and highlight important features, so that doctors or algorithms can more accurately identify disease lesions, calculates the time series distance matrix and the cumulative matrix to accurately align data at different time points and eliminate time bias, convolves the normalized medical images to obtain lesion spatial features, and generates a health risk level using the lesion spatial features. The weighted fusion of lesion feature maps using the attention mechanism can focus on the most important lesion area and ignore irrelevant parts, improving the accuracy of diagnosis. After global average pooling, the feature vector of the lesion is simplified, but the important information of the disease is still retained, providing a stable foundation for subsequent lesion risk scoring. The time series aligned data is analyzed for data trends to obtain a time series risk score. The long short-term memory (LSTM) layer is used to capture time dependence, which helps to extract key time features from historical data and further enhance the model's sensitivity to long-term trends and short-term fluctuations. The health risk level and the time series risk score are weighted and fused to obtain the health risk score of the target customer, which not only considers the current health status (such as disease risk), but also predicts future health trends (such as potential chronic or acute diseases). Obtain several policy adjustment schemes, then filter the policy adjustment schemes according to the health risk score to obtain a target adjustment strategy, use the fitness function in the genetic algorithm to gradually optimize the policy through multiple rounds of screening, crossover, disturbance, and evaluation, ensure that the final selected scheme not only meets the customer's health needs, but also improves customer satisfaction and maximizes the benefits of the insurance company. According to the target adjustment strategy, the target customer is pushed to the service, and the multi-dimensional behavior data of the target customer after the service push is collected, the multi-dimensional behavior data is tracked and analyzed to obtain the strategy execution effect, and through the comparison of the simulation strategy and the annotated adjustment strategy, the effectiveness of the insurance service can be further verified. Through tracking analysis of multi-dimensional behavior data, the system can monitor and evaluate the implementation effect of the strategy in real time to ensure that the service push is highly consistent with customer needs. Finally, the target adjustment strategy is optimized using the strategy execution effect to obtain target insurance data, which efficiently integrates and dynamically links health data and insurance data, thereby improving service response efficiency. The specific limitations of a health data-based business analysis device can be found in the limitations of a health data-based business analysis method described above, which will not be repeated here. The various modules in the above health data-based business analysis device can be realized by software, hardware, and their combinations, in whole or in part.The above modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so that the processor calls and executes the operations corresponding to the above modules.

[0214] In an embodiment, a computer device is provided, which can be a server, and an internal structure diagram thereof can be as shown in Figure 5 The computer device includes a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is configured to communicate with an external client through a network connection. The computer program, when executed by the processor, implements the functions or steps of a health data-based business analysis method on the server side.

[0215] In an embodiment, a computer device is provided, which can be a client, and an internal structure diagram thereof can be as shown in Figure 6 The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is configured to communicate with an external server through a network connection. The computer program, when executed by the processor, implements the functions or steps of a health data-based business analysis method on the client side.

[0216] In an embodiment, a computer device is provided, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the following steps when executing the computer program:

[0217] Obtaining medical image data and physiological time series data of a target customer, and performing standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data;

[0218] Performing convolution on the normalized medical images to obtain lesion spatial features, and generating a health risk level using the lesion spatial features;

[0219] Performing data trend analysis on the time series alignment data to obtain a time series risk score;

[0220] The health risk grade and the time sequence risk score are weightedly fused to obtain a health risk score of the target customer;

[0221] A plurality of policy adjustment schemes are obtained, and the policy adjustment schemes are screened according to the health risk score to obtain a target adjustment strategy;

[0222] The target customer is served according to the target adjustment strategy, and multi-dimensional behavior data of the target customer after service pushing is collected, the multi-dimensional behavior data is tracked and analyzed to obtain a strategy execution effect;

[0223] The target adjustment strategy is optimized by using the strategy execution effect to obtain target insurance data.

[0224] In several embodiments provided in the present application, it should be understood that the disclosed devices and apparatuses can be implemented in other manners. For example, the above-described system embodiments are merely illustrative, and the division of the modules is merely a logical function division, and there can be another division manner in actual implementation.

[0225] In addition, each function module in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware, or in the form of hardware plus software function module.

[0226] Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0227] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0228] In some embodiments of the present embodiment, a computer readable storage medium is provided, and a computer program is stored on the computer readable storage medium, and the computer program is characterized in that the computer program is executed by a processor to implement the steps of the method described in the above embodiments.

[0229] The computer program is stored in the readable storage medium of the present application, and when the computer program is executed by the processor of the electronic device, the following can be achieved:

[0230] Acquire medical image data and physiological time series data of a target customer, and perform standardization processing on the medical image data and the physiological time series data respectively to obtain normalized medical images and time series alignment data;

[0231] Perform convolution on the normalized medical images to obtain lesion spatial features, and generate a health risk level using the lesion spatial features;

[0232] Perform data trend analysis on the time series alignment data to obtain a time series risk score;

[0233] Weighted fusion is performed on the health risk level and the time series risk score to obtain a health risk score of the target customer;

[0234] Obtain a plurality of policy adjustment schemes, and screen the policy adjustment schemes according to the health risk score to obtain a target adjustment strategy;

[0235] According to the target adjustment strategy, service push is performed on the target customer, and multi-dimensional behavior data of the target customer after service push is collected, and tracking analysis is performed on the multi-dimensional behavior data to obtain a strategy execution effect;

[0236] The strategy execution effect is used to optimize the target adjustment strategy to obtain target insurance data.

[0237] It should be noted that the functions or steps described above with respect to the computer readable storage medium or the computer device can correspond to the related descriptions of the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0238] The computer readable storage medium can also store at least one computer executable program / instruction, such as computer readable instructions. The computer readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory, etc. The computer readable storage medium may, for example, include read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer readable storage medium can be connected to a computing device such as a computer, and then when the computing device runs the computer readable instructions stored on the computer readable storage medium, the various methods described above can be performed.

[0239] In addition, the computer device can also include (but not limited to) a data bus, an input / output (I / O) bus, a display, and an input / output device (such as a keyboard, a mouse, a speaker, etc.), etc.

[0240] The processor can communicate with external devices through the I / O bus via wired or wireless network.

[0241] In one embodiment, the at least one computer-executable instruction can also be compiled or composed into a software product / computer program product, wherein the one or more computer-executable instructions are executed by the processor to perform the steps of the various functions and / or methods described in the embodiments of the present technology.

[0242] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by computer programs instructing related hardware, and the computer programs can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present 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. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0243] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the above-described functions.

[0244] In the embodiments provided by the present disclosure, it should be understood that the disclosed apparatus and method can also be implemented in other manners. The embodiments described above are merely exemplary for describing the present disclosure. For example, the flowcharts and block diagrams in the accompanying drawings show the possible implementation architectures, functions and operation of the apparatus, method and computer program product according to the embodiments of the present disclosure. In this regard, each block in the flowcharts and block diagrams can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logic function. It should also be noted that, in some alternative implementations, the functions noted in the blocks can occur in different orders from those noted in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a special-purpose hardware-based system for implementing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0245] It should be noted that, in the present disclosure, the terms "comprising", "containing" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, method, article or device. Without more limitations, the element limited by the statement "comprising a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0246] The above-described embodiments are merely used to illustrate the technical solutions of the present disclosure, rather than limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent replacements; and these modifications or replacements 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 disclosure, and should be included in the protection scope of the present disclosure.

[0247] It should be noted that, in the embodiments of the present disclosure, if non-company software tools or components appear, they are only used for example introduction, and do not represent actual use.

Claims

1. A business analysis method based on health data, characterized in that: The method comprises: Obtaining medical imaging data and physiological time series data of a target customer, and performing normalization processing on the medical imaging data and the physiological time series data to obtain normalized medical images and time series alignment data; performing convolution on the normalized medical image to obtain lesion spatial features, and generating a health risk level using the lesion spatial features; Performing data trend analysis on the time series alignment data to obtain a time series risk score; Performing weighted fusion on the health risk level and the temporal risk score to obtain a health risk score for the target customer; Obtaining several policy adjustment plans, screening the policy adjustment plans according to the health risk score, and obtaining a target adjustment strategy; Push services to the target customers according to the target adjustment strategy, collect multi-dimensional behavior data of the target customers after the service push, track and analyze the multi-dimensional behavior data, and obtain the strategy execution effect; The target adjustment strategy is optimized using the strategy execution effect to obtain target insurance data.

2. The business analysis method based on health data according to claim 1, characterized in that: The step of respectively normalizing the medical image data and the physiological time series data to obtain normalized medical images and time series aligned data includes: Performing image enhancement on the medical image data to obtain a medically enhanced image; Normalizing the pixel values ​​of the medical enhanced image to obtain a normalized medical image; Randomly selecting two groups of physiological time series data as target time series data pairs; Determine the distance between the target time series data pairs, and construct a time series distance matrix based on the distance; Obtaining an initial cumulative matrix, and determining the minimum cumulative distance of each matrix position using the initial cumulative matrix and the temporal distance matrix; The initial cumulative matrix is ​​updated according to the minimum cumulative distance to obtain a cumulative distance matrix; Backtracking the cumulative distance matrix to obtain an optimal alignment path; The physiological time series data is aligned using the optimal alignment path to obtain time series aligned data.

3. The business analysis method based on health data according to claim 1, characterized in that: The convolution of the normalized medical image to obtain lesion spatial features, and generating a health risk level using the lesion spatial features, includes: Dividing the lesion area of ​​the normalized medical image to obtain a lesion area image; performing convolution on the lesion area image to obtain a lesion feature image; Determining an attention score for each area in the lesion area image, and generating an attention map using the attention score; Using the attention map to perform weighted fusion on the lesion feature image to obtain lesion spatial features; Performing global average pooling on the lesion spatial features to obtain a pooled feature vector; Performing full connection processing on the pooled feature vector to obtain a lesion risk score; The lesion spatial characteristics are graded according to the lesion risk score to obtain a health risk grade.

4. The business analysis method based on health data according to claim 1, characterized in that: The performing of data trend analysis on the time series alignment data to obtain a time series risk score includes: Using a preset long short-term memory layer to capture the temporal dependency of the time series alignment data; Performing multi-layer full connection on the time dependency to obtain data prediction values ​​of the time series aligned data; Performing visual analysis on the time series alignment data and the data prediction value to obtain a data trend graph; A volatility analysis is performed on the data trend graph, and a time series risk score is generated based on the volatility obtained from the analysis.

5. The business analysis method based on health data according to claim 1, characterized in that: The screening of the policy adjustment plan according to the health risk score to obtain a target adjustment strategy includes: Obtaining the health data of the target customer, performing a matching analysis between the health risk score and the health data, and obtaining a matching degree of the customer's health risk; Obtaining historical customer satisfaction and policy profitability of the policy adjustment plan, and constructing a fitness function using the customer health risk matching degree, the historical customer satisfaction, and the policy profitability; Taking a number of the aforementioned policy adjustment plans as an initial population; Determining the fitness value of each policy adjustment plan in the initial population using the fitness function; Filter out the parent insurance policy plans with fitness values ​​greater than a preset fitness threshold according to the fitness values; Cross-process the parent policy plan to obtain the child policy plan; Performing a slight perturbation on the child policy plan to obtain an updated policy plan; Performing a matching analysis between the updated insurance plan and the health risk score, and selecting the optimal insurance plan based on the matching degree of the plans obtained from the analysis; Returning the optimal policy plan to the step of determining the fitness value of each policy adjustment plan in the initial population using the fitness function, and counting the number of returns; When the number of returns reaches the preset number, the return is stopped and the final optimal insurance policy plan is used as the target adjustment strategy.

6. The business analysis method based on health data according to claim 1, characterized in that: Pushing services to the target customers according to the target adjustment strategy and collecting multi-dimensional behavior data of the target customers after the service push includes: Annotating the target adjustment strategy to obtain an annotated adjustment strategy; Acquiring the stored data of the target customer, performing strategy simulation on the stored data, and obtaining a simulation strategy; Comparing the simulation strategy with the annotation adjustment strategy, and determining whether the simulation strategy is consistent with the annotation adjustment strategy based on the comparison result; If the simulation strategy is inconsistent with the annotation adjustment strategy, optimizing the annotation adjustment strategy to obtain insurance services of the optimized annotation adjustment strategy; If the simulation strategy is consistent with the annotation adjustment strategy, obtaining the insurance service of the annotation adjustment strategy; The insurance service is pushed to a preset target client, and the multi-dimensional behavior data of the target customer returned by the preset target client is collected.

7. The business analysis method based on health data according to claim 1, characterized in that: The tracking and analysis of the multi-dimensional behavior data to obtain the strategy execution effect includes: performing standardization processing on the multidimensional behavior data to obtain standard behavior data; Extracting key analytical data from the standard behavioral data; Generating a customer profile of the target customer based on the key analysis data; The push effect of the key analysis data is analyzed based on the customer portrait to determine the strategy execution effect.

8. A business analysis device based on health data, characterized in that: The device comprises: A data normalization module is used to obtain medical imaging data and physiological time series data of target customers, and perform normalization processing on the medical imaging data and the physiological time series data to obtain normalized medical images and time series alignment data; a medical image convolution module, configured to convolve the normalized medical image to obtain lesion spatial features, and generate a health risk level using the lesion spatial features; A data trend analysis module, configured to perform data trend analysis on the time series alignment data to obtain a time series risk score; A data weighted fusion module is used to perform weighted fusion on the health risk level and the temporal risk score to obtain the health risk score of the target customer; A policy plan screening module is used to obtain a number of policy adjustment plans, screen the policy adjustment plans according to the health risk score, and obtain a target adjustment strategy; A data tracking and analysis module is used to push services to the target customers according to the target adjustment strategy, collect multi-dimensional behavior data of the target customers after the service push, track and analyze the multi-dimensional behavior data, and obtain the strategy execution effect; The adjustment strategy optimization module is used to optimize the target adjustment strategy using the strategy execution effect to obtain target insurance data.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the business analysis method based on health data according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for business analysis based on health data as described in any one of claims 1 to 7 is implemented.

Images