Intelligent medical health management system based on data mining

Abstract

The invention provides an intelligent medical health management system based on data mining, and relates to the field of intelligent health management. User physiological monitoring index data such as a blood pressure value, a blood oxygen value and a blood glucose value of a user at multiple time points and user behavior monitoring index data such as a motion three-way acceleration value and a motion three-way angular velocity value of the user are analyzed to know the physical condition of the user, and a personalized health management scheme is generated based on the physical condition of the user. Therefore, under the condition of fully considering the change of the body condition of the user, a more personalized and real-time health management scheme can be generated, so that the body health management effect of the user is ensured.

Description

Technical Field

[0001] The present application relates to the field of intelligent health management, and more specifically, to an intelligent medical health management system based on data mining. Background Art

[0002] With the popularization of medical knowledge and the enhancement of health awareness, people pay more and more attention to the role of health management in maintaining long-term health and improving the quality of life. In modern society, due to factors such as work pressure, environmental pollution, and unhealthy eating habits, the incidence of chronic diseases such as hypertension, diabetes, and heart disease has increased year by year. Traditional health management programs often adopt standardized methods, such as recommending fixed diet plans and exercise patterns, but this ignores individual differences. Everyone's physical condition and living habits are different. The "one-size-fits-all" approach adopted by traditional health management programs cannot meet the specific needs of different individuals, and many health management plans lack real-time feedback and cannot be adjusted in time to adapt to individual changes.

[0003] Therefore, an optimized smart medical and health management solution is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, this application provides a smart medical health management system based on data mining.

[0005] According to one aspect of the present application, a smart medical health management system based on data mining is provided, which includes:

[0006] The user monitoring data acquisition module is used to obtain the user's blood pressure value, blood oxygen value, blood sugar value, three-dimensional acceleration value of human body movement, and three-dimensional angular velocity value of human body movement at multiple predetermined time points within a predetermined time period;

[0007] A user monitoring data encoding module, for performing time series analysis on the blood pressure values, blood oxygen values, blood sugar values, three-dimensional acceleration values ​​of human motion, and three-dimensional angular velocity values ​​of human motion at the plurality of predetermined time points to obtain a user monitoring time series feature vector;

[0008] A user monitoring feature enhancement module, configured to perform multi-scale feature association processing on the user monitoring time series feature vector to obtain a multi-scale user monitoring association time series feature emphasis vector;

[0009] A health management plan generating module is used to generate a health management plan for the user based on the multi-scale user monitoring associated time series feature emphasis vector.

[0010] This application has significant technical effects due to the adoption of the above technical solutions:

[0011] The smart medical health management system based on data mining provided by the present application analyzes the user's physiological monitoring index data such as blood pressure value, blood oxygen value, blood sugar value and the user's three-dimensional acceleration value, three-dimensional angular velocity value and other user behavior monitoring index data at multiple time points to understand the user's physical condition, and generates a personalized health management plan based on the user's physical condition. In this way, a more personalized and real-time health management plan can be generated while fully considering the changes in the user's physical condition, thereby ensuring the user's physical health management effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0013] Figure 1 This is a system block diagram of a smart medical health management system based on data mining according to an embodiment of the present application.

[0014] Figure 2 It is a block diagram of a user monitoring data encoding module in a smart medical health management system based on data mining according to an embodiment of the present application.

[0015] Figure 3 It is a block diagram of a user physiological monitoring time series feature extraction unit in a smart medical health management system based on data mining according to an embodiment of the present application.

[0016] Figure 4 It is a block diagram of a user behavior monitoring time series feature extraction unit in a smart medical health management system based on data mining according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0018] In today's society, with the rapid development of information technology and the continuous improvement of public education, medical knowledge has become more widely popularized and people's health awareness has been strengthened. In this context, the important role of health management in maintaining people's long-term health and improving their quality of life has been increasingly valued.

[0019] In today's modern society, people are facing many challenges. Work pressure is always there, and high-intensity work tasks and competitive environment make people often in a state of tension and mental tension for a long time. At the same time, environmental pollution should not be underestimated. Air pollution and water pollution pose a potential threat to human health. In addition, unhealthy eating habits are more common in fast-paced life. High-fat, over-processed foods are everywhere on people's tables, and the intake of fruits and vegetables is insufficient. Under the combined effect of these factors, the incidence of chronic diseases such as hypertension, diabetes and heart disease has shown an upward trend year by year.

[0020] Traditional health management programs usually adopt a standardized model. For example, people are recommended a fixed diet plan, such as a certain amount of various nutrients to be consumed every day, and a specific exercise pattern, such as aerobic exercise and strength training several times a week. However, this approach ignores the crucial individual differences. Everyone's physical condition is unique. Some people may have a specific history of disease or genetic factors, while others have different levels of physical function. In addition, people's living habits vary greatly, and the nature of their work, work and rest time, etc. are different. The "one-size-fits-all" approach of traditional health management programs obviously cannot meet the specific needs of different individuals. In addition, many health management plans lack a real-time feedback mechanism and cannot be adjusted in time according to individual changes. Therefore, a smart medical health management program based on data mining is expected, which conducts in-depth mining and analysis of the user's body monitoring data to understand the user's physical condition, and generates a personalized health management program that meets the user's physical condition based on the user's physical condition.

[0021] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for smart medical and health management.

[0022] Figure 1 : is a system block diagram of a smart medical health management system based on data mining according to an embodiment of the present application. Figure 1As shown, in the smart medical health management system 100 based on data mining, it includes: a user monitoring data acquisition module 110, which is used to obtain the user's blood pressure value, blood oxygen value, blood sugar value, three-dimensional acceleration value of human movement, and three-dimensional angular velocity value of human movement at multiple predetermined time points within a predetermined time period; a user monitoring data encoding module 120, which is used to perform time series analysis on the blood pressure value, blood oxygen value, blood sugar value, three-dimensional acceleration value of human movement, and three-dimensional angular velocity value of human movement at the multiple predetermined time points to obtain a user monitoring time series feature vector; a user monitoring feature enhancement module 130, which is used to perform multi-scale feature association processing on the user monitoring time series feature vector to obtain a multi-scale user monitoring association time series feature emphasis vector; a health management plan generation module 140, which is used to generate a health management plan for the user based on the multi-scale user monitoring association time series feature emphasis vector.

[0023] In the embodiment of the present application, the user monitoring data acquisition module 110 is used to obtain the user's blood pressure value, blood oxygen value, blood sugar value, human body motion three-dimensional acceleration value, and human body motion three-dimensional angular velocity value at multiple predetermined time points within a predetermined time period. It should be understood that blood pressure is one of the important indicators for measuring the health of the cardiovascular system. Hypertension is a major risk factor for many cardiovascular diseases, including heart disease, stroke, and kidney disease. By monitoring blood pressure values, the user's blood pressure condition can be timely understood, the cardiovascular health risk can be assessed, and the health management plan can be adjusted when necessary, such as adjusting diet, increasing exercise or drug treatment. Blood oxygen saturation is an indicator of the oxygen content in the blood. Low blood oxygen saturation may be associated with a variety of health problems, including respiratory diseases and heart diseases. Monitoring blood oxygen values ​​helps to assess the user's respiratory function and circulatory efficiency. Blood sugar level is a key indicator for assessing diabetes risk and controlling diabetes. High blood sugar is the main feature of diabetes, while low blood sugar may lead to serious health problems. By monitoring blood sugar values, abnormal blood sugar can be detected in time, and diet, exercise and drug treatment can be adjusted to maintain blood sugar within a normal range and prevent the occurrence of diabetic complications. The three-dimensional acceleration values ​​of human motion represent the acceleration values ​​of the human body in the X-axis, Y-axis and Z-axis respectively, and the three-dimensional acceleration values ​​of motion reflect the user's motion pattern and activity intensity. By monitoring the three-dimensional acceleration values ​​of the human body, we can understand the user's daily activity, exercise habits and possible unhealthy behaviors such as sedentary behavior. This information is conducive to specifying personalized exercise plans to promote physical health. The three-dimensional angular velocity values ​​of the human body represent the angular velocity values ​​in the three directions of the X-axis, Y-axis and Z-axis respectively, providing information about the rotation or rotation speed of the user's body. Although its role in directly assessing health status may not be as obvious as other indicators, it can be used to more finely analyze the user's motion pattern and posture changes, help discover potential movement disorders or bad posture habits, and thus provide additional reference information for health management programs. In general, blood pressure values, blood oxygen values, blood sugar values, three-dimensional acceleration values ​​of human motion, and three-dimensional angular velocity values ​​of human motion together constitute a multidimensional portrait of the user's health status. By comprehensively analyzing the changing trends and interrelationships of these indicators, the user's health status and risk level can be more accurately assessed, thereby generating a more targeted and effective health management program. In particular, in a specific implementable embodiment of the present application, blood pressure values, blood oxygen values, blood sugar values, three-dimensional acceleration values ​​of human body movement, and three-dimensional angular velocity values ​​of human body movement data can be acquired respectively through a blood pressure sensor, a blood oxygen sensor, a blood sugar sensor, a three-dimensional accelerometer and a gyroscope.

[0024] In the embodiment of the present application, the user monitoring data encoding module 120 is used to perform time series analysis on the blood pressure values, blood oxygen values, blood sugar values, human body motion three-dimensional acceleration values, and human body motion three-dimensional angular velocity values ​​at the plurality of predetermined time points to obtain the user monitoring time series feature vector. Specifically, Figure 2 FIG. 1 is a block diagram of a user monitoring data encoding module in a smart medical health management system based on data mining according to an embodiment of the present application. Figure 2 As shown, the user monitoring data encoding module 120 includes: a user physiological monitoring time series feature extraction unit 121, which is used to extract time series features of the blood pressure values, blood oxygen values, and blood sugar values ​​at the multiple predetermined time points to obtain a user physiological monitoring time series feature vector; a user behavior monitoring time series feature extraction unit 122, which is used to extract time series features of the three-dimensional acceleration values ​​of human body motion and the three-dimensional angular velocity values ​​of human body motion at the multiple predetermined time points to obtain a user behavior monitoring time series feature vector; a user physiological monitoring time series feature optimization unit 123, which is used to perform feature distribution difference shift correction based on detail changes on the user physiological monitoring time series feature vector based on the user behavior monitoring time series feature vector to obtain an optimized user physiological monitoring time series feature vector; a user multi-source monitoring feature fusion unit 124, which is used to fuse the optimized user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector to obtain the user monitoring time series feature vector.

[0025] More specifically, Figure 3 FIG. 1 is a block diagram of a user physiological monitoring time series feature extraction unit in a smart medical health management system based on data mining according to an embodiment of the present application. Figure 3 As shown, the user physiological monitoring timing feature extraction unit 121 includes: a user physiological monitoring indicator data structuring subunit 121-1, which is used to arrange the blood pressure values, blood oxygen values, and blood glucose values ​​at the multiple predetermined time points according to the time dimension to obtain a user physiological monitoring indicator timing input matrix; a user physiological monitoring indicator data encoding subunit 121-2, which is used to pass the user physiological monitoring indicator timing input matrix through a user physiological monitoring indicator timing encoder to obtain a user physiological monitoring timing feature matrix; a user physiological monitoring timing feature dimensionality reduction subunit 121-3, which is used to expand the user physiological monitoring timing feature matrix to obtain the user physiological monitoring timing feature vector.

[0026] Considering that the physiological state of the human body changes dynamically over time, the physiological indicators at different time points can often reflect different health conditions, and in order to realize dynamic monitoring of users, update the physiological data of users in real time, and facilitate the generated health management plan to be more suitable for the current physical condition of users, therefore, in the technical solution of the present application, it is necessary to arrange the blood pressure values, blood oxygen values, and blood sugar values ​​at the multiple predetermined time points according to the time dimension to obtain the user's physiological monitoring indicator timing input matrix. By arranging the blood pressure data, blood oxygen data, and blood sugar data according to the time dimension, the timing characteristics of the user's physiological data can be retained, so as to facilitate a more accurate analysis of the user's physiological state change trend. Integrating multiple physiological monitoring indicator data at different time points into one matrix can unify the data structure, which is convenient for subsequent data processing and analysis. This unified data structure helps to reduce the complexity and error rate of data processing and improve the efficiency and accuracy of data processing.

[0027] The user physiological monitoring index timing input matrix contains the timing information of multiple physiological indicators, which can fully reflect the dynamic changes of the user's physical condition. In order to capture the key information such as the trend, periodicity, mutation point, etc. of the user's physiological indicators changing over time in this input matrix, so as to help timely discover the user's potential health problems, it is necessary to pass the user physiological monitoring index timing input matrix through the user physiological monitoring index timing encoder to obtain the user physiological monitoring timing feature matrix. Through the processing of the user physiological monitoring index timing encoder, abnormal patterns that are inconsistent with the normal physiological state can be identified, which can provide clues for early warning and intervention, and then provide personalized health advice, such as adjustments to diet. In this application, the user physiological monitoring index timing encoder is a convolutional neural network model in which adjacent layers use mutually transposed convolution kernels. Adjacent layers use mutually transposed convolution kernels, which helps to maintain the symmetry of features during feature extraction. This symmetry helps to reduce information loss in the user physiological monitoring index timing input matrix and improve the accuracy and completeness of feature extraction. Moreover, this structure is conducive to enhancing the feature learning ability of convolutional neural networks for complex data, which can better capture the complex correlations and changing trends between physiological indicators, and provide strong support for the generation of accurate health management plans.

[0028] Specifically, in an embodiment of the present application, a possible implementation method of passing the user physiological monitoring indicator timing input matrix through a user physiological monitoring indicator timing encoder to obtain a user physiological monitoring timing feature matrix may be as follows: each layer of the user physiological monitoring indicator timing encoder performs the following on the input data in the forward transmission of the layer: performing convolution processing based on a first two-dimensional convolution kernel, pooling processing along the channel dimension, and nonlinear processing on the input data to obtain a first activation feature map; and performing convolution processing based on a second two-dimensional convolution kernel, pooling processing along the channel dimension, and nonlinear activation processing on the first activation feature map to obtain a second activation feature map, wherein the first two-dimensional convolution kernel and the second two-dimensional convolution kernel are transposed of each other; wherein the input of the first layer of the user physiological monitoring indicator timing encoder is the user physiological monitoring indicator timing input matrix, and the output of the last layer of the user physiological monitoring indicator timing encoder is the user physiological monitoring timing feature matrix.

[0029] The user physiological monitoring timing feature matrix usually contains data of multiple dimensions (i.e., features) and multiple time points (i.e., samples). Although matrix operations are powerful, they may not be as efficient as vector operations in some cases, especially when processing large-scale data sets. In order to reduce the amount of calculation and improve the computational efficiency of the model, the user physiological monitoring timing feature matrix needs to be expanded in the technical solution of the present application to obtain the user physiological monitoring timing feature vector.

[0030] More specifically, Figure 4 FIG. 1 is a block diagram of a user behavior monitoring time series feature extraction unit in a smart medical health management system based on data mining according to an embodiment of the present application. Figure 4 As shown, the user behavior monitoring time series feature extraction unit 122 includes: a user behavior monitoring indicator data structuring subunit 122-1, which is used to arrange the three-dimensional acceleration values ​​of the human body motion and the three-dimensional angular velocity values ​​of the human body motion at the multiple predetermined time points according to the time dimension to obtain a user behavior monitoring indicator time series input matrix; a user behavior monitoring indicator data encoding subunit 122-2, which is used to pass the user behavior monitoring indicator time series input matrix through a user behavior monitoring indicator time series encoder to obtain a user behavior monitoring time series feature matrix; a user behavior monitoring time series feature matrix dimensionality reduction subunit 122-3, which is used to expand the user behavior monitoring time series feature matrix to obtain the user behavior monitoring time series feature vector.

[0031] Considering that human motion data is highly dynamic and changes over time. The changes in acceleration and angular velocity values ​​at different time points can reflect the user's motion pattern, activity intensity, and possible changes in health status. In order to maintain the time series of user behavior data and ensure that subsequent data extraction can capture these dynamic changes, in the technical solution of the present application, the three-dimensional acceleration values ​​of human motion and the three-dimensional angular velocity values ​​of human motion at the multiple predetermined time points are arranged according to the time dimension to obtain a user behavior monitoring indicator time series input matrix. By arranging the user behavior monitoring data according to the time dimension, time series analysis techniques can be used to identify the user's motion pattern, predict future motion trends, and even detect abnormal behaviors, such as sudden falls or abnormal static states. This information can evaluate the user's athletic ability, activity level, and potential health risks. For example, a long period of low activity level may increase the risk of cardiovascular disease, while a sudden fall may be a warning signal for stroke or fall risk. In the arranged user behavior monitoring indicator time series input matrix, each row is the user behavior monitoring data of a sample at a time point, and each column is a specific behavior monitoring indicator, such as acceleration in the X-axis direction, acceleration in the Y-axis direction, etc.

[0032] User behavior data changes over time and has obvious time-series dependence. In order to understand the long-term and short-term patterns of user behavior information in the user behavior monitoring indicator time-series input matrix, so as to more accurately analyze the user's behavior characteristics, so as to more accurately evaluate the user's health status, it is necessary to pass the user behavior monitoring indicator time-series input matrix through the user behavior monitoring indicator time-series encoder to obtain the user behavior monitoring time-series feature matrix. The key information extracted by the user behavior monitoring indicator time-series encoder includes but is not limited to: the activity intensity and behavior habits of the user in different time periods. In this application, the user behavior monitoring indicator time-series encoder is a convolutional neural network model in which adjacent layers use mutually transposed convolution kernels. Convolutional neural networks have unique advantages in processing time-series data. The local features and time-series relationships in the data in the user behavior monitoring indicator time-series input matrix can be captured through convolution operations, and the mutually transposed convolution kernel design further enhances the model's processing ability and flexibility for time-series data.

[0033] Considering that the user behavior monitoring time series feature matrix has a high dimension, in order to facilitate the subsequent processing and analysis of user behavior data, it is necessary to expand the user behavior monitoring time series feature matrix to obtain the user behavior monitoring time series feature vector. By expanding, the two-dimensional user behavior monitoring feature data is converted into one-dimensional data, which reduces the feature dimension and does not lose key information during the feature expansion process, that is, it will not affect the final result of generating the health management plan.

[0034] In particular, considering that different physiological monitoring devices (such as blood pressure sensors, blood oxygen sensors, blood glucose sensors) and behavioral monitoring devices (such as accelerometers, gyroscopes) may adopt different measurement principles and algorithms, resulting in differences in data collection methods and accuracy. For example, some devices may be unstable under certain conditions (such as temperature, humidity, etc.), thereby affecting the reliability of the data. Moreover, the data of physiological monitoring and behavioral monitoring may not be completely synchronized in time. For example, physiological data may be measured every few minutes, while behavioral data may be recorded continuously. These factors will lead to the fine-grained inconsistency of feature distribution between the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector obtained by the final encoding process. Due to the fine-grained inconsistency of feature distribution between the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector, this will cause the fused feature vector to have abnormal stacking on the boundary of the feature manifold. Based on this, in the technical solution of the present application, before fusing the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector to obtain the user monitoring feature vector, the user physiological monitoring time series feature vector is subjected to feature distribution difference shift correction based on detail changes based on the user behavior monitoring time series feature vector.

[0035] Specifically, in an embodiment of the present application, based on the user behavior monitoring timing feature vector, the user physiological monitoring timing feature vector is subjected to a feature distribution difference shift correction based on detail changes to obtain an optimized user physiological monitoring timing feature vector, including: calculating the positional difference between the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector to obtain a user monitoring shift positional difference representation vector; after cascading the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector into a user monitoring joint feature vector, the user monitoring joint feature vector is input into a one-dimensional convolutional layer to obtain a user monitoring shift convolutional coding representation vector; after linearly embedding and modulating the user physiological monitoring timing feature vector using a first embedding matrix to obtain a user physiological monitoring timing feature linear modulation vector, the user physiological monitoring timing feature vector is calculated. The user physiological monitoring timing feature linear modulation vector and the user behavior monitoring timing feature vector are differentiated by position to obtain a user monitoring shift modulation differential representation vector; the user monitoring shift position differential representation vector is used as a value vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector are used as a query vector and a key vector, and the user monitoring shift position differential representation vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector are input into a shift corrector based on an analog converter structure to obtain a user monitoring multi-scale hierarchical fusion shift modulation weight vector; the position point multiplication between the user monitoring multi-scale hierarchical fusion shift modulation weight vector and the user physiological monitoring timing feature vector is calculated to obtain the optimized user physiological monitoring timing feature vector.

[0036] In an embodiment of the present application, another possible implementation method of performing a feature distribution difference shift correction based on detail changes on the user physiological monitoring time series feature vector based on the user behavior monitoring time series feature vector to obtain an optimized user physiological monitoring time series feature vector may be: the user physiological monitoring time series feature vector is processed by the following formula to obtain an optimized user physiological monitoring time series feature vector; wherein the formula is:

[0037]

[0038] V 2d =Conv 1D ([V1; V2])

[0039]

[0040] V1'=V1⊙V d

[0041] Wherein, V1 represents the time series feature vector of the user's physiological monitoring, V2 represents the time series feature vector of the user's behavior monitoring, Indicates point-by-point subtraction, V 1d Indicates that the user monitoring shift is represented by position difference vector, [;] indicates cascade operation, Conv 1D (·) represents one-dimensional convolutional coding, V 2d represents the user monitoring shift convolutional coding representation vector, W1 represents the first embedding matrix, b1 represents the first modulation offset vector, V 3d represents the user monitoring shift modulation differential representation vector, represents vector multiplication, Softmax represents the normalized exponential function, T represents the transposed vector of the feature vector, d represents the length of the feature vector, Relu represents the activation function, V d represents the user monitoring multi-scale hierarchical fusion shift modulation weight vector, ⊙ represents the position point multiplication, and V1' represents the optimized user physiological monitoring time series feature vector.

[0042] In the technical solution of the present application, in the process of fusing the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector, a feature distribution difference shift correction method based on detail changes is proposed to address the problem of fine-grained inconsistency in feature distribution that arises in the process of fusing the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector. The method aims to reduce the abnormal stacking phenomenon of the fused feature vector on the boundary of the feature manifold, so as to improve the robustness and classification performance of the model.

[0043] Specifically, the method first calculates the element-by-element difference between the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector to obtain a user monitoring shift position differential representation vector, thereby capturing the difference between the two feature vectors. Then, the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector are cascaded into a joint feature vector and input into a one-dimensional convolution layer to generate a user monitoring shift convolution coding representation vector. That is, the deep-level shift implicit information between the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector is captured by one-dimensional convolution coding. At the same time, the user physiological monitoring timing feature vector is linearly embedded and modulated using a first embedding matrix to be encoded into a user physiological monitoring timing feature linear modulation vector, and the element-by-element difference between the user physiological monitoring timing feature linear modulation vector and the user behavior monitoring timing feature vector is calculated to obtain a user monitoring shift modulation differential representation vector. The user monitoring shift modulation differential representation vector is used to represent the position expression of the shift information of the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector in a linear space. That is, in the technical solution of the present application, the user monitoring shift position differential representation vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the user physiological monitoring timing feature vector and the user behavior monitoring timing feature vector in different feature spaces.

[0044] Next, the user monitoring shift position differential representation vector is used as the value vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector are used as the query vector and the key vector respectively, and these vectors are input into the shift corrector based on the simulated transformer structure to generate the user monitoring multi-scale hierarchical fusion shift modulation weight vector. In this way, the importance of features at different scales is evaluated by the attention mechanism, that is, the attention mechanism is used to simulate the hierarchical precipitation expression of the fine-grained difference feature distribution between the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector in different feature spaces. Finally, the element-by-element product of the user monitoring multi-scale hierarchical fusion shift modulation weight vector and the user physiological monitoring time series feature vector is calculated to obtain the optimized user physiological monitoring time series feature vector, that is, the weight is assigned to each element in the user physiological monitoring time series feature vector through the weighted operation, which reduces the influence of local disturbances and improves the quality of the feature vector.

[0045] In summary, in the method, the user monitoring shift position differential representation vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector respectively represent the explicit expression of the fine-grained distribution difference shift information between the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector in different feature spaces. The generation of the user monitoring multi-scale hierarchical fusion shift modulation weight vector utilizes the attention mechanism to evaluate the importance of features at different scales and provide weight information for feature shift correction. Ultimately, the generation of the optimized user physiological monitoring time series feature vector reduces the impact of disturbances, improves the robustness of the model to local disturbances in the input data, enhances the model's adaptability to noisy or variant data sets, improves the fusion quality of the feature vector, and improves the overall generation performance and training efficiency of the model.

[0046] The user's physiological data and behavioral data reflect the user's internal physiological state and external behavioral performance, respectively. Physiological monitoring data (such as heart rate, blood pressure, blood oxygen saturation, etc.) provides direct information about the user's physical function and health status; while behavioral monitoring data (such as exercise patterns, activity intensity, daily habits, etc.) reveals the user's lifestyle, behavioral habits and these habits. In order to form a more comprehensive and in-depth understanding of the user's health status, the optimized user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector are integrated in the technical solution of the present application to obtain a user monitoring time series feature vector. It should be understood that the use of physiological or behavioral monitoring data alone may have limitations, such as data noise, individual differences, etc. By fusing the two feature vectors, the information of both can be comprehensively utilized to reduce the errors and uncertainties caused by a single data source and improve the accuracy and reliability of the health management plan.

[0047] In an embodiment of the present application, the user monitoring feature enhancement module 130 is used to perform multi-scale feature association processing on the user monitoring time series feature vector to obtain a multi-scale user monitoring association time series feature emphasis vector. Specifically, in an embodiment of the present application, the user monitoring feature enhancement module is used to: pass the user monitoring time series feature vector through a multi-scale user monitoring time series feature association understander to obtain the multi-scale user monitoring association time series feature emphasis vector. Considering that the fused user monitoring time series feature vector may contain redundant information and noise, this information may not be necessary for subsequent health management decisions, and different features may have different importance and hierarchy when expressing the user's health status. Directly using the original feature vector may not be able to fully capture the hierarchical relationship and interaction between these features, so it is necessary to pass the user monitoring time series feature vector through a multi-scale user monitoring time series feature association understander to extract more representative features, thereby more accurately characterizing the user's health status. The multi-scale user monitoring time series feature association understander described in the present application is a convolutional neural network model including a first convolution layer and a second convolution layer, wherein the first convolution layer uses a first-scale one-dimensional convolution kernel, and the second convolution layer uses a second-scale one-dimensional convolution kernel, and the first scale is different from the second scale. Convolution kernels of different scales can capture data features at different time scales. The first-scale convolution kernel may focus more on short-term data changes in the user monitoring time series feature vector, while the second-scale convolution kernel can capture longer-term data trends in the user monitoring time series feature vector. This multi-scale feature extraction capability helps to more comprehensively understand the user's health status and helps to improve the accuracy and effectiveness of health management programs.

[0048] In an embodiment of the present application, the health management solution generation module 140 is used to generate a health management solution for the user based on the multi-scale user monitoring associated time series feature emphasis vector. Specifically, in an embodiment of the present application, the health management solution generation module is used to: pass the multi-scale user monitoring associated time series feature emphasis vector through a health management solution generator to obtain a generation result, and the generation result is the health management solution for the user. It should be understood that the health management solution generator is a machine learning model that can generate scientific, reasonable and feasible health management suggestions and guidance based on the health status of the user reflected in the multi-scale user monitoring associated time series feature emphasis vector. The health management solution specifically includes personalized dietary recommendations, including food types, intake, eating time, etc., and also includes an exercise plan, mainly including exercise type, intensity, frequency and duration, etc.

[0049] In summary, the data mining-based smart medical health management system 100 based on the embodiment of the present application is explained, which analyzes the user's physiological monitoring index data such as blood pressure value, blood oxygen value, blood sugar value and the user's three-dimensional acceleration value, three-dimensional angular velocity value and other user behavior monitoring index data at multiple time points to understand the user's physical condition, and generates a personalized health management plan based on the user's physical condition. In this way, a more personalized and real-time health management plan can be generated while fully considering the changes in the user's physical condition, thereby ensuring the user's physical health management effect.

[0050] As described above, the smart medical health management system 100 based on data mining according to the embodiment of the present application can be implemented in various terminal devices, such as a server for smart medical health management based on data mining. In one example, the smart medical health management system 100 based on data mining according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the smart medical health management system 100 based on data mining can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the smart medical health management system 100 based on data mining can also be one of the many hardware modules of the terminal device.

[0051] Alternatively, in another example, the data mining-based smart medical health management system 100 and the terminal device may also be separate devices, and the data mining-based smart medical health management system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0052] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A smart medical health management system based on data mining, characterized in that: include: The user monitoring data acquisition module is used to obtain the user's blood pressure value, blood oxygen value, blood sugar value, three-dimensional acceleration value of human body movement, and three-dimensional angular velocity value of human body movement at multiple predetermined time points within a predetermined time period; A user monitoring data encoding module, for performing time series analysis on the blood pressure values, blood oxygen values, blood sugar values, three-dimensional acceleration values ​​of human motion, and three-dimensional angular velocity values ​​of human motion at the plurality of predetermined time points to obtain a user monitoring time series feature vector; A user monitoring feature enhancement module, configured to perform multi-scale feature association processing on the user monitoring time series feature vector to obtain a multi-scale user monitoring association time series feature emphasis vector; A health management plan generating module is used to generate a health management plan for the user based on the multi-scale user monitoring associated time series feature emphasis vector.

2. The intelligent medical health management system based on data mining according to claim 1 is characterized in that: The user monitoring data encoding module includes: A user physiological monitoring time series feature extraction unit, used for extracting time series features of the blood pressure values, blood oxygen values, and blood sugar values ​​at the plurality of predetermined time points to obtain a user physiological monitoring time series feature vector; A user behavior monitoring time series feature extraction unit, used for extracting time series features of the three-dimensional acceleration values ​​and the three-dimensional angular velocity values ​​of the human body motion at the plurality of predetermined time points to obtain a user behavior monitoring time series feature vector; A user physiological monitoring time series feature optimization unit, configured to perform a feature distribution difference shift correction based on detail changes on the user physiological monitoring time series feature vector based on the user behavior monitoring time series feature vector to obtain an optimized user physiological monitoring time series feature vector; The user multi-source monitoring feature fusion unit is used to fuse the optimized user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector to obtain the user monitoring time series feature vector.

3. The intelligent medical health management system based on data mining according to claim 2 is characterized in that: The user physiological monitoring time series feature extraction unit comprises: A user physiological monitoring index data structuring subunit, for arranging the blood pressure values, blood oxygen values, and blood glucose values ​​at the plurality of predetermined time points according to the time dimension to obtain a user physiological monitoring index time series input matrix; A user physiological monitoring indicator data encoding subunit, used for passing the user physiological monitoring indicator timing input matrix through a user physiological monitoring indicator timing encoder to obtain a user physiological monitoring timing characteristic matrix; The user physiological monitoring time series feature dimensionality reduction subunit is used to expand the user physiological monitoring time series feature matrix to obtain the user physiological monitoring time series feature vector.

4. The intelligent medical health management system based on data mining according to claim 3 is characterized in that: The user behavior monitoring time series feature extraction unit comprises: A user behavior monitoring indicator data structuring subunit is used to arrange the three-dimensional acceleration values ​​of human body motion and the three-dimensional angular velocity values ​​of human body motion at the plurality of predetermined time points according to the time dimension to obtain a user behavior monitoring indicator time series input matrix; A user behavior monitoring indicator data encoding subunit, used for passing the user behavior monitoring indicator time series input matrix through a user behavior monitoring indicator time series encoder to obtain a user behavior monitoring time series feature matrix; The user behavior monitoring time series feature matrix dimension reduction subunit is used to expand the user behavior monitoring time series feature matrix to obtain the user behavior monitoring time series feature vector.

5. The intelligent medical health management system based on data mining according to claim 4 is characterized in that: The user physiological monitoring timing feature optimization unit is used to: Calculating the position difference between the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector to obtain a user monitoring shift position difference representation vector; After concatenating the user physiological monitoring time series feature vector and the user behavior monitoring time series feature vector into a user monitoring joint feature vector, the user monitoring joint feature vector is input into a one-dimensional convolutional layer to obtain a user monitoring shift convolutional coding representation vector; After performing linear embedding modulation encoding on the user physiological monitoring time series feature vector using the first embedding matrix to obtain a user physiological monitoring time series feature linear modulation vector, a position difference between the user physiological monitoring time series feature linear modulation vector and the user behavior monitoring time series feature vector is calculated to obtain a user monitoring shift modulation difference representation vector; Taking the user monitoring shift position differential representation vector as a value vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector as a query vector and a key vector, the user monitoring shift position differential representation vector, the user monitoring shift convolutional coding representation vector and the user monitoring shift modulation differential representation vector are input into a shift corrector based on an emulated transformer structure to obtain a user monitoring multi-scale hierarchical fusion shift modulation weight vector; The optimized user physiological monitoring time series feature vector is obtained by calculating the point-by-point multiplication between the user monitoring multi-scale hierarchical fusion shift modulation weight vector and the user physiological monitoring time series feature vector.

6. The intelligent medical health management system based on data mining according to claim 5 is characterized in that: The user monitoring feature enhancement module is used to: pass the user monitoring time series feature vector through a multi-scale user monitoring time series feature association understander to obtain the multi-scale user monitoring association time series feature emphasis vector.

7. The intelligent medical health management system based on data mining according to claim 6 is characterized in that: The user physiological monitoring indicator timing encoder and the user behavior monitoring indicator timing encoder are convolutional neural network models in which adjacent layers use mutually transposed convolution kernels, and the multi-scale user monitoring timing feature association understander is a convolutional neural network model including a first convolution layer and a second convolution layer, wherein the first convolution layer uses a one-dimensional convolution kernel of a first scale, and the second convolution layer uses a one-dimensional convolution kernel of a second scale, and the first scale is different from the second scale.

8. The intelligent medical health management system based on data mining according to claim 7 is characterized in that: The health management solution generation module is used to: pass the multi-scale user monitoring associated time series feature emphasis vector through a health management solution generator to obtain a generation result, and the generation result is the health management solution for the user.

Images