A health monitoring management method based on big data

Abstract

The application discloses a health monitoring management method based on big data, relates to the technical field of health monitoring, and adopts an adaptive partition optimization algorithm to optimize the user portrait model, acquires the user portrait model after optimization, can effectively improve the model optimization efficiency and optimization effect, so that user portrait recognition is more accurate, then the user real-time portrait big data is identified by the user portrait model after optimization deployed on the scheduling server, highly personalized and forward-looking health suggestions can be pushed, and the efficiency and accuracy of health monitoring management are effectively improved.

Description

Technical Field

[0001] This application relates to the field of health monitoring technology, and more specifically, to a health monitoring and management method based on big data. Background Technology

[0002] With increasing health awareness and the widespread adoption of wearable devices, collecting personal health data through smart devices has become possible. This data includes multi-dimensional information such as heart rate, blood pressure, steps, and sleep quality, constituting massive amounts of user health big data. How to effectively utilize this big data to provide users with personalized and accurate health advice is an important research direction in the fields of smart healthcare and health management.

[0003] In existing technologies, some health management systems provide health advice through simple threshold judgments or rule-based engines. For example, the system will issue an alert when the heart rate exceeds a certain preset value. However, these methods have significant limitations: First, they cannot comprehensively consider the complex correlations between multi-dimensional data, making it difficult to capture the overall dynamic changes in a user's health; second, fixed thresholds and rules ignore individual differences, failing to provide truly personalized guidance for users with different physical conditions and lifestyles; finally, these systems lack the ability to self-learn and optimize, making it difficult to continuously improve the accuracy and effectiveness of their recommendation strategies.

[0004] In recent years, deep learning technology has achieved great success in fields such as image recognition and natural language processing, and has also provided new ideas for health data analysis. By building deep learning models, users' health patterns can be learned from massive amounts of historical data, and accurate profile identification can be performed. However, the optimization process of deep learning models is complex, especially in high-dimensional, multi-temporal health big data scenarios, where models are prone to getting stuck in local optima, resulting in slow convergence speed and low recognition accuracy. Traditional optimization algorithms (such as gradient descent) are sensitive to initial values ​​when dealing with such complex models and are prone to getting lost in complex loss function terrains, making it difficult to find the global optimum. Summary of the Invention

[0005] This application aims to provide a health monitoring and management method based on big data, which addresses the technical problems of existing health management systems, such as inaccurate suggested strategies, low personalization, and low optimization efficiency of deep learning models, which are prone to getting trapped in local optima.

[0006] This application provides a health monitoring and management method based on big data, including:

[0007] Collect pre-stored big data learning tasks or input through human-computer interaction; wherein, the big data learning tasks include user historical profile big data and their corresponding health advice strategy tags; each health advice strategy tag corresponds to a health advice strategy stored in the database;

[0008] A user profile recognition model is constructed using a deep learning model. Based on the big data learning task, the user profile model is optimized using an adaptive partitioning optimization algorithm to obtain the optimized user profile model, and the optimized user profile model is deployed on a server.

[0009] Collect real-time user profile big data corresponding to users, upload the real-time user profile big data to the server, and schedule the optimized user profile model deployed on the server to identify the real-time user profile big data to obtain target health suggestion strategies;

[0010] The target health recommendation strategy is fed back to the user's designated device, thus completing the health monitoring and management method based on big data.

[0011] In one possible implementation, the user historical profile big data includes user vital sign data, user behavior data, and environmental data at multiple consecutive sampling time points;

[0012] The user's vital signs data include at least one of heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, electrocardiogram, and blood glucose.

[0013] The user behavior data includes at least one of the following: steps, activity level, sleep duration, and calorie consumption;

[0014] The environmental data includes at least one of the following: air quality index, temperature, humidity, and noise.

[0015] In one possible implementation, a user profile recognition model is constructed using a deep learning model, including: constructing a user profile recognition model using a CNN-LSTM model.

[0016] In one possible implementation, the user profile model is optimized based on the big data learning task and using an adaptive partitioning optimization algorithm to obtain the optimized user profile model, including:

[0017] Initialize the population; wherein the population includes multiple individuals, and each individual includes the parameters to be optimized in the user profile model;

[0018] Based on the big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual;

[0019] An adaptive position perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in an individual with optimized initial neighborhood.

[0020] Based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined.

[0021] A group information fusion mechanism is used to optimize the individuals after guidance optimization, resulting in individuals after group information fusion optimization.

[0022] The fitness-influenced mutation mechanism is used to perform global optimization on the individuals after the population information fusion optimization, resulting in globally optimized individuals;

[0023] Determine whether the current number of optimizations is greater than or equal to the preset maximum number of optimizations. If so, re-obtain the optimal individual based on the globally optimized individuals, and use the parameters of the re-obtained optimal individual as the final parameters of the user profile model to obtain the optimized user profile model.

[0024] In one possible implementation, based on the big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual, including:

[0025] For any individual, the parameters contained in the individual are applied to the user profile model to obtain the user profile model after applying the parameters.

[0026] The user's historical profile big data is used as the input to the user profile model after application parameters, and the corresponding health advice strategy label is used as the expected output of the user profile model after application parameters to obtain the cross-entropy loss function value or the root mean square loss function value.

[0027] Add the obtained cross-entropy loss function value or root mean square loss function value to the non-zero positive constant term to obtain the non-zero sum result; where the non-zero positive constant term is <0.00001;

[0028] The reciprocal of the sum of the non-zero values ​​is used to obtain the fitness of each individual, and the individual with the highest fitness is determined as the optimal individual.

[0029] In one possible implementation, an adaptive positional perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in an individual with optimized initial neighborhood, including:

[0030]

[0031] in, Let j represent the j-th individual in the t-th optimization process, where j = 1, 2, ..., NP, and NP represents the total number of individuals. Let j represent the individual after optimization of the initial neighborhood. Represents the natural constant. Represents the first random number between (0,1). Represents pi (π). represents the adaptive perturbation factor, and T represents the preset maximum number of optimization attempts.

[0032] In one possible implementation, based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined, including:

[0033]

[0034] in, Let m represent the individual after optimization of the m-th initial neighborhood during the t-th optimization process, where m = 1, 2, ..., NP. This represents the m-th individual after guided optimization. This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number uniformly distributed between [0,1]. Represents the optimal individual. Indicates the adaptive guiding factor. This represents the maximum value of the adaptive guidance factor, and is set to 2; This represents the minimum value of the adaptive guidance factor, and is set to 0.0001.

[0035] In one possible implementation, a group information fusion mechanism is used to perform group information fusion optimization on the individuals after guided optimization, resulting in individuals after group information fusion optimization, including:

[0036]

[0037] in, Let d be the parameter of the individual after the nth guided optimization in the t-th optimization process, where n = 1, 2, ..., NP. Let d represent the d-th dimension parameter of the individual after the information fusion and optimization of the nth group. Represents the group information fusion factor. Let represent the d-th dimension parameter of the individual after the h-th guided optimization in the t-th optimization process, where h=1,2,…,NP, d=1,2,…,D, and D represents the total dimension of the individual's parameters. This represents the upper limit of the d-th dimension parameter. This represents the lower bound of the d-th dimension parameter. This represents the maximum value of the group information fusion factor, and is set to 0.4; This represents the minimum value of the group information fusion factor, and is set to 0.01; This represents the information collection factor for a single individual. This represents the information fusion coefficient, and is set to 0.01. Let represent the Euclidean distance between the individual after the nth guided optimization and the individual after the hth guided optimization in the t-th optimization process. Let e ​​represent the fifth random number between (0,1), and let e represent the natural constant.

[0038] In one possible implementation, a fitness-influenced mutation mechanism is used to globally optimize the individuals after the population information fusion optimization, resulting in globally optimized individuals, including:

[0039]

[0040] in, This represents the d-th dimension parameter of the individual after the information fusion optimization of the k-th group during the t-th optimization process. Let d represent the d-th dimension parameter of the individual after the k-th global optimization. This represents the sixth random number between (0,1). This represents the variation range factor for the (t+1)th iteration. This represents the variation range factor for the t-th variation. This represents the ratio of the difference between the upper and lower limits of the d-th dimension parameter to the maximum number of optimization iterations. Indicates the population optimization progress factor. Indicates the first threshold. Indicates the second threshold, and Greater than , Let represent the sum of the fitness of individuals after the fusion and optimization of information from all groups during the t-th optimization process. Let represent the sum of the fitness of individuals after the fusion and optimization of all population information during the (t-1)th optimization process, and NP represent the total number of individuals.

[0041] In one possible implementation, real-time user profile big data is collected for each user, uploaded to a server, and an optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data to obtain target health recommendation strategies, including:

[0042] Collect real-time user profile big data corresponding to each user; wherein, the real-time user profile big data has the same data structure as the historical user profile big data;

[0043] The real-time user profile big data is uploaded to the server, and the optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data to obtain the probability distribution output by the optimized user profile model.

[0044] Based on the probability distribution, the health advice strategy label with the highest probability is taken as the target label, and the health advice strategy corresponding to the target label is obtained from the database, and the obtained health advice strategy is taken as the target health advice strategy.

[0045] Beneficial effects:

[0046] This application provides a health monitoring and management method based on big data. It uses an adaptive partitioning optimization algorithm to optimize the user profile model, thereby obtaining an optimized user profile model. This effectively improves the model optimization efficiency and effect, making user profile recognition more accurate. Then, the optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data, which can push highly personalized and forward-looking health suggestions, effectively improving the efficiency and accuracy of health monitoring and management. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a health monitoring and management method based on big data proposed in an embodiment of this application. Detailed Implementation

[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0050] like Figure 1 As shown in the figure, this application provides a health monitoring and management method based on big data, including:

[0051] S101. Collect pre-stored big data learning tasks or input through human-computer interaction; wherein, the big data learning tasks include user historical profile big data and their corresponding health advice strategy tags; each health advice strategy tag corresponds to a health advice strategy stored in the database.

[0052] User historical profile big data refers to multi-dimensional data collected continuously or discretely from users over a period of time. According to a preferred embodiment of this application, this data includes user vital sign data, user behavior data, and environmental data at multiple consecutive sampling time points.

[0053] User vital signs data reflect a user's physiological state, such as at least one of the following: heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, electrocardiogram, and blood glucose. This data can be collected through devices such as smart bracelets, smartwatches, blood pressure monitors, and blood glucose meters.

[0054] User behavior data reflects a user's daily activities and habits, such as at least one of the following: steps, activity level, sleep duration, and calorie consumption.

[0055] Environmental data: reflects the user's external environment, such as at least one of the following: Air Quality Index (AQI), temperature, humidity, and noise.

[0056] Health advice strategy tags are category tags pre-labeled by medical experts or health consultants, corresponding to users' historical profile big data. For example, data from a certain period might be tagged as "Good cardiovascular health, maintain current lifestyle habits; insufficient exercise, increase aerobic exercise; poor sleep quality, adjust sleep schedule," etc. Each tag uniquely corresponds to a specific, actionable health advice strategy stored in the database.

[0057] By collecting pre-stored big data learning tasks or inputting them through human-computer interaction, we can effectively learn expert-customized health monitoring strategies, which can not only improve the efficiency of health monitoring and management, but also achieve 24 / 7 health monitoring and management.

[0058] S102. A user profile recognition model is constructed using a deep learning model. Based on the big data learning task, the user profile model is optimized using an adaptive partitioning optimization algorithm to obtain the optimized user profile model, and the optimized user profile model is deployed on the server.

[0059] Existing technologies typically use the Adam optimizer to optimize deep learning models, which is prone to getting trapped in local optima. Therefore, this application uses an adaptive partitioning optimization algorithm to optimize the user profile model. This not only ensures the optimization efficiency of the algorithm and solves the problem of getting trapped in local optima, but also removes the sensitivity of the initial value. Even with random initialization, optimization can be effectively achieved.

[0060] The user profile model optimized by this adaptive partitioning optimization algorithm can more accurately learn the data relationships in big data learning tasks, enabling the optimized user profile model to more accurately identify user profiles and improve the accuracy of health monitoring and management.

[0061] S103. Collect real-time user profile big data corresponding to the user, upload the real-time user profile big data to the server, and schedule the optimized user profile model deployed on the server to identify the real-time user profile big data to obtain target health suggestion strategies.

[0062] By offloading the complex model optimization process to the server, user terminal devices only need to handle data collection and uploading, significantly reducing the computational demands on them. The optimized model responds quickly to real-time user data, providing near real-time health feedback. Furthermore, the adaptive partitioning optimization algorithm is insensitive to initial values ​​and possesses powerful global search capabilities, making the model optimization process more stable and reliable. This avoids optimization failures caused by poor random initialization, enhancing the robustness of the entire health management system.

[0063] S104. Feed back the target health recommendation strategy to the device specified by the user to complete the health monitoring and management method based on big data.

[0064] This application constructs a complete closed loop from data acquisition, model optimization, strategy generation to feedback application. The optimized user profile model is deployed on a server and continuously processes real-time user data. As new data accumulates, the system can periodically initiate a new round of big data learning tasks, using adaptive partitioning optimization algorithms to further optimize and update the model. This enables the health management system to adapt to changes in users' health conditions (such as aging and changes in lifestyle), achieving self-evolution of the model and continuous optimization of suggested strategies, providing users with long-term, dynamic, and reliable intelligent health services.

[0065] In one possible implementation, a user profile recognition model is constructed using a deep learning model, including: constructing a user profile recognition model using a CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory Network) model.

[0066] This application collects multi-dimensional historical user profile big data including vital signs, behavior, and environment, and optimizes it using a CNN-LSTM deep learning model. CNN can effectively extract spatial features from the data (such as the correlation between different vital signs at the same time point), while LSTM excels at capturing long-term dependencies in time series (such as the trend of heart rate changes with activity level). This combined model can deeply understand the user's complex and dynamic health patterns, far exceeding traditional threshold judgments or simple rules. Ultimately, based on the accurate identification of real-time user profile big data, the system can push highly personalized and forward-looking health suggestions. For example, if it detects that your recent sleep quality has declined and your daytime activity has decreased, it suggests moderate aerobic exercise to improve sleep, rather than simply recommending sleep deprivation.

[0067] In one possible implementation, the user profile model is optimized based on the big data learning task and using an adaptive partitioning optimization algorithm to obtain the optimized user profile model, including:

[0068] Initialize the population. The population consists of multiple individuals, and each individual includes the parameters to be optimized in the user profile model (such as connection weights, thresholds, etc.).

[0069] Optionally, random initialization or chaotic mapping initialization can be used to obtain the population. Each individual in the population should be encoded in vector form to facilitate continuous optimization.

[0070] Based on the big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual.

[0071] An adaptive position perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in an individual with optimized initial neighborhood.

[0072] Based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined.

[0073] A group information fusion mechanism is used to optimize the individuals after guidance optimization, resulting in individuals with optimized group information fusion.

[0074] The fitness-influenced mutation mechanism is used to perform global optimization on individuals after the population information fusion optimization, resulting in globally optimized individuals.

[0075] Determine whether the current number of optimizations is greater than or equal to the preset maximum number of optimizations. If so, re-obtain the optimal individual based on the globally optimized individuals, and use the parameters of the re-obtained optimal individual as the final parameters of the user profile model to obtain the optimized user profile model.

[0076] Existing technologies typically use the Adam optimizer to optimize deep learning models, which is prone to getting trapped in local optima. Therefore, this application uses an adaptive partitioning optimization algorithm to optimize the user profile model. This not only ensures the optimization efficiency of the algorithm and solves the problem of getting trapped in local optima, but also removes the sensitivity of the initial value. Even with random initialization, optimization can be effectively achieved.

[0077] Optionally, after each change, the individual can be processed to ensure the effectiveness of the algorithm.

[0078] In one possible implementation, based on the big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual, including:

[0079] For any individual, the parameters contained in the individual are applied to the user profile model to obtain the user profile model after applying the parameters.

[0080] The user's historical profile big data is used as the input to the user profile model after application parameters, and the corresponding health advice strategy label is used as the expected output of the user profile model after application parameters to obtain the cross-entropy loss function value or the root mean square loss function value.

[0081] Add the obtained cross-entropy loss function value or root mean square loss function value to the non-zero positive constant term to obtain the non-zero sum result; where the non-zero positive constant term is <0.00001;

[0082] The reciprocal of the sum of the non-zero values ​​is used to obtain the fitness of each individual, and the individual with the highest fitness is determined as the optimal individual.

[0083] In one possible implementation, an adaptive positional perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in an individual with optimized initial neighborhood, including:

[0084]

[0085] in, Let j represent the j-th individual in the t-th optimization process, where j = 1, 2, ..., NP, and NP represents the total number of individuals. Let j represent the individual after optimization of the initial neighborhood. Represents the natural constant. Represents the first random number between (0,1). Represents pi (π). represents the adaptive perturbation factor, and T represents the preset maximum number of optimization attempts.

[0086] In the early stages of optimization, this mechanism uses an exponentially decaying perturbation factor to allow individuals to explore randomly over a large range, enhancing the algorithm's global search capability and preventing premature convergence to local optima. In the mid-stages of the algorithm, the perturbation of individuals alternates between increasing and decreasing, further ensuring optimization diversity. In the later stages, individuals only explore a small area near their current location, effectively guaranteeing the algorithm's optimization accuracy.

[0087] In one possible implementation, based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined, including:

[0088]

[0089] in, Let m represent the individual after optimization of the m-th initial neighborhood during the t-th optimization process, where m = 1, 2, ..., NP. This represents the m-th individual after guided optimization. This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number uniformly distributed between [0,1]. Represents the optimal individual. Indicates the adaptive guiding factor. This represents the maximum value of the adaptive guidance factor, and is set to 2; This represents the minimum value of the adaptive guidance factor, and is set to 0.0001.

[0090] This mechanism uses the current best individual as a guide and adaptively adjusts the guiding factor to enable the population of individuals to quickly converge toward the optimal solution region during the optimization process, with increasingly higher accuracy. This ensures both the breadth of the search and the speed and accuracy of convergence.

[0091] In one possible implementation, a group information fusion mechanism is used to perform group information fusion optimization on the individuals after guided optimization, resulting in individuals after group information fusion optimization, including:

[0092]

[0093] in, Let d be the parameter of the individual after the nth guided optimization in the t-th optimization process, where n = 1, 2, ..., NP. Let d represent the d-th dimension parameter of the individual after the information fusion and optimization of the nth group. Represents the group information fusion factor. Let represent the d-th dimension parameter of the individual after the h-th guided optimization in the t-th optimization process, where h=1,2,…,NP, d=1,2,…,D, and D represents the total dimension of the individual's parameters. This represents the upper limit of the d-th dimension parameter. This represents the lower bound of the d-th dimension parameter. This represents the maximum value of the group information fusion factor, and is set to 0.4; This represents the minimum value of the group information fusion factor, and is set to 0.01; This represents the information collection factor for a single individual. This represents the information fusion coefficient, and is set to 0.01. Let represent the Euclidean distance between the individual after the nth guided optimization and the individual after the hth guided optimization in the t-th optimization process. Let e ​​represent the fifth random number between (0,1), and let e represent the natural constant.

[0094] By introducing Euclidean distance between individuals, this mechanism enables each individual to learn from all individuals, promoting the exchange and fusion of high-quality information within the population. This helps to escape the trap of local optima and find better solutions. Furthermore, in the later stages of the algorithm, the degree of population information fusion can be adaptively adjusted to ensure the algorithm's convergence.

[0095] In one possible implementation, a fitness-influenced mutation mechanism is used to globally optimize the individuals after the population information fusion optimization, resulting in globally optimized individuals, including:

[0096]

[0097] in, This represents the d-th dimension parameter of the individual after the information fusion optimization of the k-th group during the t-th optimization process. Let d represent the d-th dimension parameter of the individual after the k-th global optimization. This represents the sixth random number between (0,1). This represents the variation range factor for the (t+1)th iteration. This represents the variation range factor for the t-th variation. This represents the ratio of the difference between the upper and lower limits of the d-th dimension parameter to the maximum number of optimization iterations. Indicates the population optimization progress factor. This represents the first threshold (e.g., it can be set to 1 / 4 * minimum fitness value). This represents the second threshold (e.g., it can be set to 1 / 6 * minimum fitness value), and Greater than , Let represent the sum of the fitness of individuals after the fusion and optimization of information from all groups during the t-th optimization process. Let represent the sum of the fitness of individuals after the fusion and optimization of all population information during the (t-1)th optimization process, and NP represent the total number of individuals.

[0098] The mutation range is determined by the changes in the overall fitness of the population. The mutation range can be adaptively adjusted according to the convergence state of the population. When the algorithm has not converged, a larger optimization step size can be given, which can effectively avoid the algorithm getting stuck in local optima. As the algorithm iterates and the population converges, the step size is forced to gradually shrink, which is converted into a fine search and improves the search accuracy.

[0099] By combining the above mechanisms, the model can be effectively prevented from getting stuck in local optima during the optimization process, enabling the optimized user profile model to more accurately identify user profiles.

[0100] In one possible implementation, real-time user profile big data is collected for each user, uploaded to a server, and an optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data to obtain target health recommendation strategies, including:

[0101] Collect real-time user profile big data corresponding to each user. The data structure of the real-time user profile big data is the same as that of the historical user profile big data. For example, if the historical user profile big data consists of data from M consecutive sampling time points, then the real-time user profile big data should also consist of data from M consecutive sampling time points, except that the real-time user profile big data is the most recently collected data.

[0102] The real-time user profile big data is uploaded to the server, and the optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data to obtain the probability distribution output by the optimized user profile model.

[0103] Based on the probability distribution, the health advice strategy label with the highest probability is taken as the target label, and the health advice strategy corresponding to the target label is obtained from the database, and the obtained health advice strategy is taken as the target health advice strategy.

[0104] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0105] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, electronic devices, and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0106] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0107] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0108] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0109] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0110] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A health monitoring and management method based on big data, characterized in that, include: Collect pre-stored big data learning tasks or input through human-computer interaction; wherein, the big data learning tasks include user historical profile big data and their corresponding health advice strategy tags; each health advice strategy tag corresponds to a health advice strategy stored in the database; A user profile recognition model is constructed using a deep learning model. Based on the big data learning task, the user profile model is optimized using an adaptive partitioning optimization algorithm to obtain the optimized user profile model, and the optimized user profile model is deployed on a server. Collect real-time user profile big data corresponding to users, upload the real-time user profile big data to the server, and schedule the optimized user profile model deployed on the server to identify the real-time user profile big data to obtain target health suggestion strategies; The target health recommendation strategy is fed back to the user's designated device, thus completing the health monitoring and management method based on big data. Based on the aforementioned big data learning task, and using an adaptive partitioning optimization algorithm to optimize the user profile model, an optimized user profile model is obtained, including: Initialize the population; wherein the population includes multiple individuals, and each individual includes the parameters to be optimized in the user profile model; Based on the big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual; An adaptive position perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in an individual with optimized initial neighborhood. Based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined. A group information fusion mechanism is used to optimize the individuals after guidance optimization, resulting in individuals after group information fusion optimization. The fitness-influenced mutation mechanism is used to perform global optimization on the individuals after the population information fusion optimization, resulting in globally optimized individuals; Determine whether the current number of optimizations is greater than or equal to the preset maximum number of optimizations. If so, re-obtain the optimal individual based on the globally optimized individuals, and use the parameters in the re-obtained optimal individual as the final parameters of the user profile model to obtain the optimized user profile model. Based on the aforementioned big data learning task, the fitness of each individual is obtained, and the individual with the highest fitness is determined as the optimal individual, including: For any individual, the parameters contained in the individual are applied to the user profile model to obtain the user profile model after applying the parameters. The user's historical profile big data is used as the input to the user profile model after application parameters, and the corresponding health advice strategy label is used as the expected output of the user profile model after application parameters to obtain the cross-entropy loss function value or the root mean square loss function value. Add the obtained cross-entropy loss function value or root mean square loss function value to the non-zero positive constant term to obtain the non-zero sum result; where the non-zero positive constant term is <0.00001; Take the reciprocal of the sum of the non-zero values ​​to obtain the fitness of each individual, and determine the individual with the highest fitness as the optimal individual; An adaptive position perturbation mechanism is used to optimize the initial neighborhood of an individual, resulting in individuals with optimized initial neighborhoods, including: ； ； in, Let j represent the j-th individual in the t-th optimization process, where j = 1, 2, ..., NP, and NP represents the total number of individuals. Let j represent the individual after optimization of the initial neighborhood. Represents the natural constant. Represents the first random number between (0,1). Represents pi (π). This represents the adaptive perturbation factor, and T represents the preset maximum number of optimization attempts; Based on the optimal individual, an adaptive learning mechanism is used to guide the optimization of individuals after the initial neighborhood optimization, and the individuals after guided optimization are determined, including: ； ； in, Let m represent the individual after optimization of the m-th initial neighborhood during the t-th optimization process, where m = 1, 2, ..., NP. This represents the m-th individual after guided optimization. This represents the second random number between (0,1). This represents a third random number between (0,1). This represents the fourth random number uniformly distributed between [0,1]. Represents the optimal individual. Indicates the adaptive guiding factor. This represents the maximum value of the adaptive guidance factor, and is set to 2; This represents the minimum value of the adaptive guidance factor, and is set to 0.0001; A group information fusion mechanism is used to further optimize the individuals after guided optimization, resulting in individuals optimized by group information fusion, including: ； ； ； in, Let d be the parameter of the individual after the nth guided optimization in the t-th optimization process, where n = 1, 2, ..., NP. Let d represent the d-th dimension parameter of the individual after the information fusion and optimization of the nth group. Indicates the group information fusion factor. Let represent the d-th dimension parameter of the individual after the h-th guided optimization in the t-th optimization process, where h=1,2,…,NP, d=1,2,…,D, and D represents the total dimension of the individual's parameters. This represents the upper limit of the d-th dimension parameter. This represents the lower bound of the d-th dimension parameter. This represents the maximum value of the group information fusion factor, and is set to 0.4; This represents the minimum value of the group information fusion factor, and is set to 0.01; This represents the information collection factor for a single individual. This represents the information fusion coefficient, and is set to 0.

01. Let represent the Euclidean distance between the individual after the nth guided optimization and the individual after the hth guided optimization in the t-th optimization process. Let represent the fifth random number between (0,1), and e represent the natural constant; The fitness-influenced mutation mechanism is used to globally optimize the individuals after the population information fusion optimization, resulting in globally optimized individuals, including: ； ； ； in, This represents the d-th dimension parameter of the individual after the information fusion optimization of the k-th group during the t-th optimization process. Let d represent the d-th dimension parameter of the individual after the k-th global optimization. This represents the sixth random number between (0,1). This represents the variation range factor for the (t+1)th iteration. This represents the variation range factor for the t-th variation. This represents the ratio of the difference between the upper and lower limits of the d-th dimension parameter to the maximum number of optimization iterations. Indicates the population optimization progress factor. Indicates the first threshold. Indicates the second threshold, and Greater than , Let represent the sum of the fitness of individuals after the fusion and optimization of information from all groups during the t-th optimization process. Let represent the sum of the fitness of individuals after the fusion and optimization of all population information during the (t-1)th optimization process, and NP represent the total number of individuals.

2. The health monitoring and management method based on big data according to claim 1, characterized in that, The user historical profile big data includes user vital signs data, user behavior data, and environmental data from multiple consecutive sampling time points; The user's vital signs data include at least one of heart rate, blood pressure, blood oxygen saturation, body temperature, respiratory rate, electrocardiogram, and blood glucose. The user behavior data includes at least one of the following: steps, activity level, sleep duration, and calorie consumption; The environmental data includes at least one of the following: air quality index, temperature, humidity, and noise.

3. The health monitoring and management method based on big data according to claim 1, characterized in that, A user profile recognition model is constructed using a deep learning model, including: constructing a user profile recognition model using a CNN-LSTM model.

4. The health monitoring and management method based on big data according to claim 1, characterized in that, Collect real-time user profile big data corresponding to each user, upload the real-time user profile big data to the server, and schedule the optimized user profile model deployed on the server to identify the real-time user profile big data to obtain target health suggestion strategies, including: Collect real-time user profile big data corresponding to each user; wherein, the real-time user profile big data has the same data structure as the historical user profile big data; The real-time user profile big data is uploaded to the server, and the optimized user profile model deployed on the server is scheduled to identify the real-time user profile big data to obtain the probability distribution output by the optimized user profile model. Based on the probability distribution, the health advice strategy label with the highest probability is taken as the target label, and the health advice strategy corresponding to the target label is obtained from the database, and the obtained health advice strategy is taken as the target health advice strategy.

Images