Child allergy management system

Abstract

The invention relates to the field of intelligent medical treatment, in particular to a child allergy management system. Comprising an acute allergy early warning module which is used for carrying out risk early warning in acute allergy time; the data acquisition module is used for acquiring basic data of a to-be-tested person, and the basic data comprises physiological index data, environment data and medical record data; the risk assessment module is used for transmitting the basic data to a risk assessment model for assessment to obtain a risk result; the construction process of the risk assessment model comprises the following steps: obtaining a basic data set of a to-be-tested person and a label of a risk category; outputting the data set and the label of the risk category to a neural network for training to obtain a risk assessment model; wherein the neural network carries out parameter initialization based on the oscillation period, and initialization parameters are calculated through a weight adjustment factor, the oscillation period, a sine item or a cosine item of the oscillation period and a product of correlation parameters of the neural network. The application has good clinical value.

Description

Technical Field

[0001] This application relates to the field of intelligent healthcare, and particularly to a management system, device, program product, and computer-readable storage medium for childhood allergies. Background Art

[0002] As a globally prevalent health problem, the occurrence risk of allergies is affected by multiple complex factors, including the patient's physiological indicators, environmental conditions, medical history, and genetic factors. With the exacerbation of environmental pollution and climate change, the incidence of allergy-related diseases shows an upward trend, and the pediatric population has become the main group with synchronous increases in both incidence and prevalence. Since allergic reactions can rapidly develop into serious health threats, especially triggering acute conditions such as anaphylactic shock, it is crucial to timely and accurately assess and manage the risk of allergy occurrence and provide effective early warnings. However, the management of childhood allergic diseases faces multiple challenges, such as untimely patient follow-up and medical treatment, single methods of allergy awareness education, and poor treatment compliance; at the physician level, due to the imperfect training mechanism for allergy specialists, there are problems such as insufficient diagnostic capabilities of physicians and incomplete mastery of standardized management processes. Through the continuous efforts of several generations of pediatricians in China, the standardized diagnosis and treatment level of childhood allergic diseases has been significantly improved, but the diagnosis rate and control rate are still not ideal, and many challenges still remain. Secondly, in the task of allergy occurrence risk early warning, existing artificial intelligence methods rely too much on local optima during model training and lack effective global optimization strategies, resulting in a decrease in the accuracy of the model when dealing with complex features. In addition, the single weight initialization method limits the exploration ability in the initial stage and easily causes difficulties in the training convergence of the model in complex medical tasks. Summary of the Invention

[0003] In view of the above problems, the present invention provides a management system for childhood allergies, which specifically includes: An allergy risk early warning module: used for early warning of the risk of acute allergy occurrence; the allergy risk early warning module includes: A data collection module: used to obtain the basic data of the person to be tested, including physiological index data, environmental data, and medical record data; A risk assessment module: used to input the basic data into a risk assessment model for evaluation to obtain a risk result, and the risk result includes one or more of the following: no allergy risk, low allergy risk, medium allergy risk, high allergy risk; The construction process of the risk assessment model is as follows: Obtain the basic data set of the person to be tested and the labels of risk categories; Input the data set and the labels of risk categories into a neural network for training to obtain a risk assessment model; Among them, the neural network is initialized based on the oscillation period, and the initialization parameters are calculated through the product of the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the associated parameters of the neural network.

[0004] During the construction of the risk assessment model, the oscillation period adaptively adjusts and optimizes the iterative process of the neural network. When the loss of the loss function fluctuates violently during the iteration process, the oscillation amplitude is adjusted and then the iteration is carried out.

[0005] During the construction of the risk assessment model, random perturbations are added after the iterative parameter update to optimize the parameter update range of the neural network, and the magnitude of the random perturbation is dynamically adjusted according to the number of training iterations.

[0006] The construction of the evaluation model also includes data augmentation; the dataset and the labels of the risk categories are augmented to obtain augmented data, and the augmented data is input into the neural network model for training to obtain a risk assessment model; the data augmentation is performed through a generative adversarial network, and the loss function calculation of the generative adversarial network includes adversarial loss and feature contrast loss. The calculation of the adversarial loss includes random noise data, feature data of allergic medical data, and the calculation of the feature contrast loss includes allergic medical data.

[0007] The feature data of the allergic medical data includes one or more of the following: statistical features, semantic features, image features, time series features, specific medical index features; the allergic medical data includes one or more of the following: heart rate, blood pressure, air quality index, pollen concentration, allergy history, type of medication used, age, gender, temperature and humidity index, family allergy history.

[0008] The generative adversarial network also includes a distribution adaptation evaluation mechanism, including diversity loss and quality evaluation loss. The quality evaluation loss measures the quality difference by comparing the mean and variance between the generated allergic medical data and the real allergic medical data; the diversity loss is calculated through the diversity loss factor and the generated samples.

[0009] The system also includes one or more of the following: Early allergy symptom identification module: used for early identification and screening of symptoms related to allergic diseases; Allergy remission monitoring module: used for recording detection data during the allergy remission period; Allergy test intelligent reading module: used for intelligent interpretation of allergen detection reports; Allergy emergency plan module: used to push the treatment plans for acute allergic events that have occurred or allergic events that are at risk of occurring; Allergy external health management module: used for health management of allergic children related to allergic and non-allergic diseases.

[0010] The object of the present invention is to provide a method for risk assessment of allergy occurrence, including: Obtaining the basic data of the person to be tested, including physiological index data, environmental data, and medical record data; Inputting the basic data into a risk assessment model for evaluation to obtain a risk result, where the risk result includes one or more of the following: no allergy risk, low allergy risk, medium allergy risk, high allergy risk; The construction process of the risk assessment model is as follows: Obtaining the basic data set of the person to be tested and the label of the risk category; Inputting the data set and the label of the risk category into a neural network for training to obtain a risk assessment model; Among them, the neural network is initialized with parameters based on the oscillation period, and the initial parameters are calculated by the product of the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the correlation parameters of the neural network.

[0011] During the construction process of the risk assessment model, the oscillation period adaptively adjusts and optimizes the iterative process of the neural network. When the loss of the loss function fluctuates violently during the iteration process, the oscillation amplitude is adjusted and then the iteration is carried out.

[0012] During the construction process of the risk model, random perturbation is added after the iterative parameter update to optimize the parameter update range of the neural network, and the size of the random perturbation is dynamically adjusted by the number of training iterations.

[0013] The construction of the evaluation model also includes data augmentation; after data augmentation of the data set and the label of the risk category, the augmented data is obtained, and the augmented data is input into the neural network model for training to obtain a risk assessment model; the data augmentation is carried out through a generative adversarial network, and the loss function calculation of the generative adversarial network includes adversarial loss and feature contrast loss, and the adversarial loss calculation includes random noise data, feature data of allergy medical data, and the feature contrast loss calculation includes allergy medical data.

[0014] The feature data of the allergy medical data includes one or more of the following: statistical features, semantic features, image features, time series features, specific medical index features; the allergy medical data includes one or more of the following: heart rate, blood pressure, air quality index, pollen concentration, allergy history, type of drug used, age, gender, temperature and humidity index, family allergy history.

[0015] The generative adversarial network further includes a distribution adaptation evaluation mechanism, including a diversity loss and a quality evaluation loss. The quality evaluation loss measures the quality difference by comparing the mean and variance between the generated allergy medical data and the real allergy medical data. The diversity loss is obtained by calculating the diversity loss factor and the generated samples.

[0016] The object of the present invention is to provide a computer device, which includes a memory, a processor, and a computer program or instruction stored on the memory. The computer program or instruction is executed by the processor to implement the above-mentioned risk assessment method for allergy occurrence.

[0017] The object of the present invention is to provide a computer-readable storage medium, on which a computer program or instruction is stored. The computer program or instruction is executed by the processor to implement the above-mentioned risk assessment method for allergy occurrence.

[0018] Advantages of the present invention: 1. In the allergy occurrence risk warning task, a neural network optimized by an oscillation period is adopted. The parameter initialization uses the oscillation period and random perturbation to optimize the initial distribution of the neural network weights and biases, and increase the model's exploration ability of the feature space. Moreover, the oscillation period is combined with a dynamic adjustment mechanism to adapt the period amplitude according to the fluctuation of the loss function, improving the global search ability during training and avoiding falling into local optima. In addition, a dynamic adjustment of local bias is added during the parameter optimization process, so that more depth feature interaction information is incorporated into the neural network weight update.

[0019] 2. In the allergy occurrence risk warning task, a weight update combining random perturbation and dynamic feedback is adopted. After each weight update, random perturbation is used and gradually decays as the training progresses, ensuring a wider exploration of the parameter space in the initial stage and maintaining stability during later convergence. At the same time, the dynamic feedback factor is used to optimize the weight update step size according to the importance difference of different features, making the model more adaptable to the complexity and diversity of allergy medical data.

[0020] 3. In the allergy occurrence risk warning task, an enhancement of the non-linear feature adaptation ability is adopted. In the error backpropagation, the local bias of the oscillation period and the gradient information are coupled to improve the neural network's learning ability of the non-linear features of allergy medical data. In addition, the model's adaptation ability to complex feature interaction patterns is significantly enhanced, and it can capture subtle allergy factor associations, further improving the prediction accuracy.

[0021] 4. In the task of allergy occurrence risk early warning, the medical data generated by existing data augmentation methods has low quality, and the generated samples lack semantic consistency, which easily leads to poor model training effects. In addition, the diversity among samples is insufficient, and the problem of being unable to effectively cover diverse types of allergic reactions. In the allergy occurrence risk early warning task, a generative adversarial network is used to augment medical data. By using the mechanism of mutual confrontation between the generator and the discriminator, medical data is generated by combining random noise with feature embedding, thus solving the problem of insufficient samples of allergy-related medical data. In addition, the generator adopts a feature contrast loss to make the generated data consistent with the real data in the high-level semantic space. The feature contrast is optimized based on medical data feature embeddings (such as heart rate, air quality index, etc.) to ensure the medical reliability of the generated data. At the same time, the discriminator combines a diversity loss. By evaluating the differences among samples, it ensures the diversity of the generated data and effectively avoids the problem of single patterns. Moreover, the discriminator adopts a quality evaluation loss to evaluate the data quality based on the mean and variance differences between the generated data and the real data, improving the numerical consistency of the generated data.

[0022] 5. In the allergy occurrence risk early warning task, a dynamic feature control mechanism is adopted. The generator expands the coverage of the generated samples by adaptively adjusting the noise distribution and feature embedding method, enabling it to reflect more diverse types of allergic reactions and symptoms. The discriminator guides the generator to generate high-quality medical data that is superior to traditional methods in both data feature distribution and semantic space through a distribution adaptive evaluation mechanism.

[0023] 6. The present invention integrates the situations in each stage of allergy and proposes a children's allergy management system, including early allergy detection, allergy remission monitoring, intelligent allergy test reading, allergy risk early warning, allergy emergency plan, and external allergy health management, to enhance personalized treatment for the unique needs of each child, enabling the whole process effective management of the child patients, and realizing the closed-loop management of the whole process information service for the dynamic monitoring, precise diagnosis, intelligent evaluation, intelligent screening, standardized treatment, and follow-up plan of children's allergic diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0025] Figure 1 It is a schematic diagram of the children's allergy management system provided by the embodiment of the present invention; Figure 2 It is a schematic diagram of the architecture of the children's allergy management system provided by the embodiment of the present invention; Figure 3The application interface of the "client" application provided by the embodiments of the present invention; Figure 4 The application interface of the "physician side" application provided by the embodiments of the present invention; Figure 5 The schematic flow diagram of the risk assessment method for allergic reactions provided by the embodiments of the present invention; Figure 6 The schematic diagram of the risk assessment device for allergic reactions provided by the embodiments of the present invention. Detailed implementation manners

[0026] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] In some processes described in the specification and claims of the present invention and the above-mentioned drawings, a plurality of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The serial numbers of the operations, such as S101, S102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0028] Figure 1 The schematic diagram of the management of childhood allergies provided by the embodiments of the present invention specifically includes: The allergy risk warning module: used for warning of the risk of acute allergic reactions; the allergy risk warning module includes: The data acquisition module: obtains the basic data of the person to be tested, including physiological index data, environmental data, and medical record data; In a specific embodiment, the model training data for allergic reaction risk warning of the present invention is collected from the acquisition of multi-source data, including but not limited to the following categories: 1) Patient physiological index data: such as heart rate, blood pressure, respiratory rate, etc.; 2) Environmental data: such as air quality index (PM2.5, PM10), temperature and humidity, allergen concentration, etc.; 3) Medical record data: such as allergy history, drug use records, hospitalization records, etc.; 4) Demographic information: such as age, gender, place of residence, etc.

[0029] In one embodiment, the attributes of the data include: Ra1 is the heart rate (bpm), Ra2 is the blood pressure (mmHg), Da1 is the air quality index (AQI), Da2 is the pollen concentration (ppm), Pa1 is the allergy history (classified as none, mild, moderate, severe), Pa2 is the type of medication used (classified as antihistamines, steroids, etc.), Xa1 is the age (years), Xa2 is the gender (classified as male, female), Da3 is the temperature-humidity index (combined feature), and Pa3 is the family allergy history (classified as none, present).

[0030] It should be noted that this embodiment is only to illustrate a data format and type of the present invention. In actual applications, the attributes of the data are usually more than 10 attributes, and the number of data attributes may reach dozens or even hundreds.

[0031] Furthermore, the collected data is labeled. The labeling method of the present invention is manual labeling. In one embodiment, the labeling categories include: no allergy risk, low allergy risk, medium allergy risk, and high allergy risk, a total of 4 categories.

[0032] In this embodiment, five example data are as follows:

[0033]

[0034] The present invention uses the Word2Vec algorithm to vectorize the text. The Word2Vec algorithm is a commonly used vectorization algorithm in the art. It scans these texts to be vectorized according to a preset large-scale corpus, and represents each word as a one-hot encoded vector. The dimension of the one-hot encoded vector is equal to the size of the vocabulary in the corpus.

[0035] In one embodiment, the construction of the evaluation model further includes data augmentation; after data augmentation of the dataset and the labels of the risk categories, the augmented data is input into the neural network model for training to obtain a risk assessment model; the data augmentation is performed through a generative adversarial network. The loss function calculation of the generative adversarial network includes adversarial loss and feature contrast loss. The adversarial loss calculation includes random noise data, feature data of allergy medical data, and the feature contrast loss calculation includes allergy medical data.

[0036] In one embodiment, the feature data of the allergy medical data includes, but is not limited to, one or more of the following: statistical features, semantic features, image features, time series features, specific medical index features; the allergy medical data includes, but is not limited to, one or more of the following: heart rate, blood pressure, air quality index, pollen concentration, allergy history, type of medication used, age, gender, temperature-humidity index, family allergy history.

[0037] In one embodiment, the generative adversarial network further includes a distribution adaptation evaluation mechanism, including a diversity loss and a quality evaluation loss. The quality evaluation loss measures the quality difference by comparing the mean and variance between the generated allergy medical data and the real allergy medical data. The diversity loss is calculated by multiplying a diversity loss factor with the generated samples.

[0038] In a specific embodiment, in the task of the present invention, the acquisition, annotation, and preprocessing of allergy-related medical training data are time-consuming and laborious, and insufficient training samples easily lead to poor generalization ability of the model and affect the accuracy of the model at the same time.

[0039] To solve the problem of limited generalization ability of the model caused by insufficient original allergy-related medical data, the generative adversarial network is used to augment the collected and annotated allergy-related medical training data, and a feature contrast loss and a diversity loss are adopted during the training process of the generative adversarial network to improve the quality and diversity of the generated allergy-related medical data.

[0040] Specifically, the training process of the generative adversarial network is as follows:

[0041] 1) Initialize the parameters of the generator and discriminator of the generative adversarial network. The generator uses random noise and the feature information of the original allergy-related medical data to generate preliminary synthetic data, and the discriminator distinguishes between the real allergy-related medical data and the synthetic allergy-related medical data. At the same time, through the way of adversarial training, the generator is prompted to gradually improve the quality and diversity of the synthetic allergy-related medical data. Let be the initial weight of the generator, be the initial weight of the discriminator. The initialization method is random initialization, and the initialized parameters follow a normal distribution with a mean of 0 and a variance of the identity matrix.

[0042] 2) The generator generates allergy-related medical data by combining random noise and a feature embedding vector, and inputs the synthetic allergy-related medical data into the discriminator, aiming to minimize the discrimination error of the discriminator and enhance the diversity and authenticity of the allergy-related medical data through an adaptive mechanism, so as to solve the problem of lack of diversity in the generated allergy-related medical data caused by a single noise input. The loss function of the generator includes an adversarial loss and a feature contrast loss. The feature contrast loss performs a high-level semantic comparison between the generated allergy-related medical data and the real allergy-related medical data in the form of a perceptual loss, and keeps consistent with the real allergy-related medical data at the high-level semantic level. The calculation method of the adversarial loss of the generator is expressed as:

[0043]

[0044] In the formula, is the generator adversarial loss; is a random noise vector. By changing the distribution of the input noise or adjusting the generation strategy, such as using different noise vectors or feature embeddings, the generated data can cover a wider range of allergic reaction types and symptoms, thereby increasing the coverage of the dataset; is the feature embedding extracted from real allergy-related medical data, such as the high-dimensional representation obtained from the original medical data through feature extraction methods, including statistical features (such as mean, variance, etc.), semantic features (case text embeddings extracted by language models such as BioBERT), image features (medical image feature vectors extracted by convolutional neural networks such as ResNet), time series features (trends or periodic features reflecting the dynamic changes of allergic reactions), and specific medical indicators (such as IgE antibody levels, skin prick test results, etc.); is the generator network function, is the discriminator network function, denotes expectation. Preferably, can be set as a random vector with a uniform distribution or a normal distribution.

[0045] Moreover, the calculation method of the feature contrast loss of the generator is expressed as:

[0046]

[0047] In the formula, is the feature contrast loss of the generator, is the weight coefficient for adjusting this loss, is the output of the feature extraction network (such as a multi-layer convolutional neural network), is the real allergy-related medical data. For example, the attributes include Ra1 as heart rate (bpm), Ra2 as blood pressure (mmHg), Da1 as air quality index (AQI), Da2 as pollen concentration (ppm), Pa1 as allergy history (classified as none, mild, moderate, severe), Pa2 as type of medication used (classified as antihistamines, steroids, etc.), Xa1 as age (years), Xa2 as gender (classified as male, female), Da3 as temperature-humidity index (combined feature), Pa3 as family allergy history (classified as none, present); is the L2 norm, which is used here to measure the difference between the generated allergy-related medical data and the real allergy-related medical data in the feature space. Preferably, is set to 0.3.

[0048] Furthermore, in order to achieve dynamic adaptive control of the weight coefficient for adjusting this loss, an adjustment factor based on the root mean square error is adopted. A greater penalty is imposed on larger distances at the initial stage of training, and it is gradually weakened as the generation quality improves. The calculation method is expressed as:

[0049]

[0050] Wherein, is the initial weight factor, is the distance metric between the generated allergy-related medical data and the real allergy-related medical data, is the maximum distance that appears during the training process. Preferably, is set to 0.3.

[0051] Based on this, the feature contrast loss function helps to ensure the medical reliability and practicality of the generated data by comparing the high-level semantic similarity between the generated allergy-related medical data and the real data in the feature space. The generated data can better simulate the key attributes of the real data, thus more accurately reflecting the allergy phenomenon in practical applications; the quality assessment loss function enables the discriminator to evaluate the consistency of the data by comparing the statistical characteristics (such as mean and variance) of the generated data and the real data, ensuring that the generated data is numerically consistent with the real data, and helping to improve the accuracy and reliability of the model when dealing with real-world data.

[0052] 3) While the discriminator performs binary classification on whether the input allergy-related medical data is real, it also analyzes the differences between the generated allergy-related medical data and the real allergy-related medical data based on statistical distribution. By adopting the distribution adaptation evaluation mechanism, it can effectively evaluate and guide the improvement of the generated allergy-related medical data in terms of quality and diversity, thus solving the problem of single pattern or overfitting of the generated allergy-related medical data. The diversity loss of the discriminator is combined with the quality assessment loss, where the quality assessment loss measures the difference between distributions by comparing the means and variances of the generated allergy-related medical data and the real allergy-related medical data;

[0053] The calculation method of the quality assessment loss of the discriminator is expressed as:

[0054]

[0055] Wherein, is the quality assessment loss of the discriminator, is the weight coefficient of the quality assessment loss of the discriminator, and respectively represent the mean and variance of the generated allergy-related medical data, and respectively represent the mean and variance of the real allergy-related medical data; is the L2 norm, which is used here to measure the numerical difference between the mean and variance. Preferably, is set to 0.3.

[0056] Based on this, the diversity loss ensures a certain difference between the generated samples, not only avoiding overfitting of the model but also enhancing the generalization ability of the model. In the task of allergic-related medical data augmentation, considering the individual differences between cases helps to enhance the diversity of the augmented samples.

[0057] Furthermore, in order to encourage the differences between the generated allergic-related medical data in different samples, the discriminator imposes additional constraints on the generated allergic-related medical data with the help of the diversity loss, guiding the generator to output diverse samples as much as possible while generating high-quality allergic-related medical data. The calculation method of the diversity loss is expressed as:

[0058]

[0059] In the formula, is the diversity loss, is the diversity loss weight factor, represents the th generated sample, represents the th generated sample, is the number of generated allergic-related medical data. Preferably, is set to 0.3.

[0060] 4) Iterative training of the generator and the discriminator. By alternately updating the weights and continuously improving their respective capabilities, the generator gradually learns to generate higher-quality and more diverse synthetic allergic-related medical data in the confrontation, and the discriminator gradually learns to more accurately determine the true and false in the confrontation, thus solving the problem that pure one-way training is likely to cause the network to be difficult to converge or fall into mode collapse. The calculation method of the weight update of the generator and the discriminator is expressed as:

[0061]

[0062]

[0063] In the formula, is the parameter update operation, is the learning rate of the generative adversarial network. Preferably, is set to 0.001.

[0064] 5) Repeat the above steps iteratively until the preset stop iteration condition is met, which means the model training is completed. In one embodiment, the preset stop iteration condition is to reach the preset maximum number of iterations. Preferably, the preset maximum number of iterations is set to 1000 times.

[0065] After the training of the allergy-related medical data augmentation model is completed, the trained allergy-related medical data augmentation model is used to increase the number of samples. In one embodiment, assume that the original collected samples are 800, and the allergy-related medical data augmentation model generates 200 samples through augmentation. Then, the augmented allergy-related medical data set contains 1000 samples.

[0066] Risk assessment module: Input the basic data into the risk assessment model for evaluation to obtain a risk result, where the risk result includes one or more of the following: no allergy risk, low allergy risk, medium allergy risk, high allergy risk; The construction process of the risk assessment model is as follows: Obtain the basic data set of the subject to be tested and the label of the risk category; Input the data set and the label of the risk category into the neural network for training to obtain a risk assessment model; Among them, the neural network is initialized with parameters based on the oscillation period, and the initial parameters are calculated through the product of the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the associated parameters of the neural network.

[0067] In one embodiment, the parameters include weights and biases. The initialization of the weights is calculated by the product of the weight adjustment factor, the oscillation period, the sine term of the oscillation period, and the associated parameters of the neural network to obtain the initial weights, and the initialization of the biases is calculated by the product of the weight adjustment factor, the oscillation period, the cosine term of the oscillation period, and the associated parameters of the neural network to obtain the initial biases.

[0068] In one embodiment, during the construction process of the risk assessment model, the oscillation period adaptively adjusts to optimize the iterative process of the neural network. When the loss of the loss function fluctuates violently during the iteration process, the oscillation amplitude is adjusted and then the iteration is carried out.

[0069] In one embodiment, during the construction process of the risk model, random perturbations are added after the iterative parameter update to optimize the parameter update range of the neural network, and the magnitude of the random perturbations is dynamically adjusted according to the number of training iterations.

[0070] In a specific embodiment, the present invention uses a neural network algorithm optimized based on the oscillation period as the allergy-related medical data feature extraction model to extract features from the allergy-related medical data. Aiming at the technical problems such as being prone to falling into local optimal solutions and being difficult to efficiently explore the parameter space in the optimization of the neural network, the weights and biases of the neural network are updated by using the dynamic oscillation method, and the global search is realized by combining the random perturbation and the adaptive adjustment method, so that the neural network can get rid of the dependence on the local gradient of the gradient descent method during the training process and achieve the global optimum or approximate global optimum.

[0071] Specifically, the training process of the neural network algorithm based on oscillation period optimization is as follows: 1) Initialize the parameters of the neural network. Different from the conventional random initialization method, the present invention adopts an initialization method based on oscillation theory, sets the oscillation period and perturbation amplitude, improves the exploration ability in the initial stage and the subsequent optimization efficiency. In the feature extraction task of allergy-related medical data, the data may have highly nonlinear and complex patterns. By adopting the parameter initialization method based on oscillation theory and using the oscillation period and random perturbation amplitude, not only can the diversity of the initial weight distribution be improved, but also the ability of the neural network to explore the feature space can be enhanced. For the complex correlations in allergy-related medical data, it can effectively avoid the convergence difficulties that may be brought by the traditional initialization method and improve the efficiency and effect of optimization. The initialization method is expressed as:

[0072]

[0073] In the formula, is the initial weight of the neural network, is the initial weight adjustment factor of the neural network, is a random number uniformly distributed from -1 to 1, is the sine term obtained according to the initial oscillation period of the neural network, is the initial oscillation period of the neural network; is the correlation parameter of the number of connections of the neuron, is the number of connections of the neuron. Preferably, set , is a random number uniformly distributed from 0 to 1, that is, the oscillation period is randomly distributed between 0 and ; set , .

[0074] Furthermore, the bias initialization of the neural network not only considers random factors, but also is related to the weighted average of the input pattern of the neuron, and is expressed as:

[0075]

[0076] In the formula, is the initial bias of the neural network, is the bias initialization adjustment factor of the neural network, is the cosine term obtained according to the initial oscillation period of the neural network, is the correlation parameter of the sum of the bias and the initial weight of the neural network, is the current neuron and the th input connection corresponding to the sum of all initial weights, is the current neuron and the The initial weights corresponding to the input connections. Preferably, set , .

[0077] 2) During the iterative process of neural network training, adaptively adjust the oscillation period according to the change of the loss function value and the complexity of the neural network. Moderately increase the oscillation amplitude when the loss function fluctuates violently, so as to accelerate the search process and avoid premature convergence to the local minimum. During the training process, dynamically adjust the oscillation period according to the loss function fluctuation. Increasing the oscillation amplitude when the loss fluctuates violently helps to jump out of the local optimal solution. Especially in the optimization difficulties caused by uneven sample distribution in allergy-related medical data, the adjustment mechanism of the oscillation period improves the global search ability of the model, enabling it to better capture the global characteristics of allergy-related medical data while avoiding premature convergence to the sub-optimal solution, expressed as:

[0078]

[0079] In the formula, is the change amount of the oscillation period of the neural network in the th iteration, is the first oscillation change adjustment parameter, is the second oscillation change adjustment parameter, is the loss function value of the neural network in the th iteration, is the loss function value of the neural network in the th iteration, is the dynamic adjustment factor of the neural network, is the initial weight corresponding to the current neuron and the th input connection. Preferably, set , .

[0080] Furthermore, update the oscillation period according to the change amount of the oscillation period of the neural network, expressed as:

[0081]

[0082] In the formula, is the oscillation period of the neural network in the th iteration, is the oscillation period of the neural network in the th iteration, is the change amount of the oscillation period of the neural network in the th iteration.

[0083] 3) During the process of error backpropagation, calculate the local deviation of the neural network. The local deviation takes into account both the gradient of the loss function and the coupling effect of the oscillation period and the deviation of the adjacent layer, incorporates more dynamic information during parameter update, improves the adaptability of the neural network to non-linear and deep structures, combines the oscillation period with the gradient information to calculate the local deviation, makes the parameter update contain more dynamic information, thereby improving the adaptability of the neural network to the non-linear features in allergy-related medical data. Especially for the complex feature interaction relationships in allergy-related medical data, enhances the network's learning ability for deep feature patterns. The calculation method of the local deviation is expressed as:

[0084]

[0085] In the formula, is the local deviation of the neural network at the -th iteration, is the partial derivative of the loss function of the neuron with respect to its current weight at the -th iteration, is the weight of the neural network at the -th iteration, is the weight corresponding to the connection between the current neuron and the -th input at the -th iteration of the neural network, is the adjustment factor for deviation transfer, is the correlation parameter of the sum of the bias and the initial weight of the neural network, is the sum of all weights corresponding to the connection between the neuron and the -th input at the -th iteration. Preferably, set .

[0086] 4) Update the weights and biases, and comprehensively use the local deviation to achieve more flexible dynamic adjustment of the neural network parameters, thereby enhancing the fitting ability for complex allergy-related medical data. By using the oscillation period influence and feedback adjustment factor for the update of weights and biases, the parameter optimization can adapt to the importance differences of different features in allergy data, ensure the robustness of the neural network in a complex medical data environment, and improve the accuracy of feature extraction, especially the recognition ability for the subtle associations between difficult-to-capture allergy factors, which is expressed as:

[0087]

[0088] In the formula, is the weight of the neural network at the -th iteration, is the weight of the neural network at the -th iteration, is the step size for the update of the neural network weight, is the local deviation of the neural network at the -th iteration, is the oscillation period of the neural network at the -th iteration, is the feedback regulation factor, is the sum of all weights corresponding to the connection between the neuron at the -th iteration and the -th input, is the exponent of the feedback regulation. Preferably, is set to 0.2, is set to 0.5, is set to 2.

[0089] Furthermore, the update of the bias is also affected by the oscillation period, and the calculation method is expressed as:

[0090]

[0091] In the formula, is the bias of the neural network at the -th iteration, is the bias of the neural network at the -th iteration, is the step size of the bias update, is the regulation factor in the bias update. Preferably, is set to 0.0001, is set to 0.05.

[0092] 5) After each iteration completes the update of the weights and biases, in order to enhance the ability to search for the global minimum, a random perturbation is added to the updated weights of the neural network. The perturbation magnitude changes with the number of training rounds, so as to maintain a large search range in the initial stage and reduce the fluctuations when converging in the later stage. When dealing with allergy-related medical data, it can alleviate the local extreme value problem in the feature space, improve the generalization performance of the neural network, and ensure that the extracted medical features are more representative. The calculation method is expressed as:

[0093]

[0094] In the formula, is the parameter update operation, is the perturbation amplitude of the neural network weights at the -th iteration, is a random number uniformly distributed from -1 to 1.

[0095] Furthermore, the perturbation amplitude gradually decreases as the training progresses, and the calculation method is expressed as:

[0096]

[0097] In the formula, is the initial perturbation amplitude, is the perturbation attenuation adjustment factor, is the current iteration number. Preferably, is set to 0.001, is set to 0.4.

[0098] 6) Determine whether the training of the neural network converges. Observe the change amount of the loss function after consecutive multiple rounds of iteration. If the change of the loss function is less than a certain threshold within the preset number of iteration rounds, terminate the training. The calculation method of the change amount of the loss function of the neural network is expressed as:

[0099]

[0100] In the formula, is the change amount of the loss function of the neuron in the th iteration, is the loss function value of the neural network in the th iteration, is the loss function value of the neural network in the th iteration.

[0101] If is lower than the preset threshold in multiple rounds of iteration, it is determined that the training of the neural network converges, and the iteration is stopped. Preferably, the preset threshold for stopping iteration is 0.005.

[0102] After the neural network training is completed, use the trained neural network to extract features from the allergy-related medical data. Further, input the data after feature extraction into the preset Softmax function to calculate the class probability, and take the class with the largest class probability as the class for allergy occurrence risk warning. For example, the classes include no allergy risk, low allergy risk, medium allergy risk, and high allergy risk, a total of 4 classes.

[0103] In one embodiment, the system performs risk warning based on the risk assessment result and gives a treatment plan.

[0104] In one embodiment, the allergy risk warning module further includes one or more of the following: a data augmentation module, a warning notification module. The data augmentation module inputs the data in the acquisition module into the data augmentation module to obtain augmented data, and inputs the augmented data into the risk assessment module to perform risk assessment to obtain a risk result; the warning notification module performs warning notification based on the risk result of the risk assessment module.

[0105] In one embodiment, the system further includes a data storage module for hierarchically storing the data in the data acquisition module. The hierarchical storage includes hot data storage and cold data storage. Hot data is data for real-time access, and cold data is data for long-term preservation. In one embodiment, the system further includes a user interaction module, which includes a user side and a doctor side. The user can view the real-time health status and risk assessment results, obtain personalized health management suggestions or popular science content, record allergic attacks or allergic medications, interpret allergic results intelligently, conduct daily tests for allergic relief, issue early warnings for the risk of allergic occurrence, formulate emergency plans for acute allergic attacks, and manage the health of children with allergies through the user side; the doctor can view the risk assessment results, formulate personalized health management plans, push a list of high-risk patients, construct health records, conduct intelligent screening, perform precise allergic assessments, conduct follow-up management, and carry out hierarchical diagnosis and treatment through the doctor side.

[0106] In one embodiment, the system further includes a security and privacy protection module for restricting access rights through encryption.

[0107] In one embodiment, the system further includes a system integration and interface module for real-time acquisition of patients' medical data, real-time meteorological data, real-time air quality index, and real-time pollen concentration.

[0108] In one embodiment, the early allergy screening module: is used for the early differential screening of symptoms related to allergic diseases; an early screening model is obtained through training an early screening model with children's clinical data. Allergic asthma is a common asthma phenotype in children, accounting for more than 80% of all asthma phenotypes. The occurrence and development of asthma follow a natural process. Children often present with atopic dermatitis (eczema) as the first disease. As the condition progresses, more than 2 / 3 of patients with atopic dermatitis develop allergic rhinitis, and 1 / 3 of patients with atopic dermatitis progress to asthma. The key to asthma management lies in the "four earlys": early detection, early diagnosis, early treatment, and early management. Therefore, an early screening model for the risk of childhood asthma occurrence is constructed based on children's allergy history, early risk factors, pulmonary function tests, and allergen test results and implanted into the system. Physicians can assist in decision-making through the prediction model based on the data uploaded by the children and give corresponding guidance and suggestions to the children and their parents, with a view to achieving the "four early management" of allergic asthma.

[0109] Allergy remission monitoring refers to the home self - detection and recording during the allergy remission period. Parents can log in through the patient - side entrance of the system. In the health center interface, they can record the daily medications and acute - attack medications for the child, and also record the results of each examination / laboratory test for the child on a daily basis. They can also conduct monthly assessment records such as the asthma control assessment scale, rhinitis symptom assessment scale, and skin assessment scale. In addition, the dynamic lung function of the child can be monitored daily through a portable spirometer, and wheezing sounds can be monitored through wearable devices such as electronic stethoscopes to collect the child's health information in real - time, timely grasp the child's allergy control situation and possible risk factors. At the same time, the system will automatically give medication and monitoring reminders, and push relevant allergy health popular science knowledge to enhance the initiative of parents and children in taking medications and guide parents to conduct home self - management of allergic children.

[0110] The intelligent allergy - test reading module refers to the intelligent interpretation of allergy test reports. The realization of this function relies on the allergy intelligent interpretation report system implanted in the "Allergy - child Management" system. Parents can upload the allergy test report form in the "My Allergens" module. The system can record the test results of allergens at different times, conduct intelligent interpretation and analysis based on the allergy test report form, and present the corresponding allergen avoidance management guidance in real - time. Parents can learn the allergen interpretation results in advance after the diagnosis.

[0111] The acute - allergy - plan module refers to the treatment plan for dealing with already - occurred acute allergic events or allergic events at risk of occurrence. This function is for the allergy - risk warning system to assess the possible risk of allergic attacks in children. When it captures respiratory symptoms such as wheezing and breath - holding, skin symptoms such as rashes and urticaria, digestive symptoms such as vomiting and diarrhea, and severe allergic reactions involving multiple - system symptoms through recording or wearable devices, according to the onset and severity of different symptoms, it will push personalized emergency treatment guidance suggestions.

[0112] The external - health - management module for allergies refers to the health management related to allergies and non - allergic diseases for allergic children. Parents can search for popular science knowledge related to the daily care, diet feeding, and vaccine injection of allergic children in the built - in allergy popular science knowledge of the system, or leave a message in the background.

[0113] In a specific embodiment, the system further includes a data pre - processing module. Data pre - processing includes a data calculation module (for data calculation), a data cleaning module (for data processing), and a data management module (for structured query, addition, deletion, modification, and other applications of data). After the data acquisition module collects data, it is processed by the data pre - processing module and then input to the risk assessment module; or after data pre - processing, it is expanded by the data expansion module and then input to the risk assessment model; or it is first expanded and then pre - processed by the data pre - processing module and then input to the risk assessment model.

[0114] The system further includes a data mining module, which performs data mining through integrity rules, label features, statistical analysis, and data profiling to obtain mined data. The data from the preprocessing module is mined using the data mining module to obtain mined data, or the augmented data and original data from the data augmentation module are mined to obtain mined data.

[0115] The system further includes a data service module, and the mined data is trained through a data model module to obtain a data model (risk assessment model).

[0116] The system further includes a data monitoring module (for real-time monitoring of data changes), a scale push module (for pushing allergy-related scales), and a service API module.

[0117] In a specific embodiment, the functions of the data storage module include:

[0118] 1) Hierarchical data storage: According to the type and access frequency of data, data is divided into hot data and cold data:

[0119] 2) Hot data: Data that is accessed in real time, such as the latest monitoring indicators of patients;

[0120] 3) Cold data: Historical data that is stored for a long time, such as allergy attack records, diagnosis and treatment records, etc.

[0121] In addition, the data storage module supports multiple data types, including:

[0122] 1) Structured data: Such as patient basic information, allergen test reports;

[0123] 2) Semi-structured data: Such as symptom records, medication records;

[0124] 3) Unstructured data: Such as uploaded pictures, test reports in PDF format.

[0125] In a specific embodiment, the warning and notification module: sets warning levels according to risk scores and performs notification pushing. The pushing channels include: real-time notification of patients or their families via text messages, WeChat mini-programs, emails, etc.

[0126] In addition, for high-risk situations, the system can trigger a phone reminder or manual customer service intervention.

[0127] In a specific embodiment, the user interaction module: on the patient / family side, supports viewing real-time health status and risk assessment results, and supports manual entry of information such as symptoms and medication records to supplement data collection. In addition, personalized health management suggestions and popular science content can also be obtained.

[0128] On the doctor side, it supports browsing the risk assessment results and health trend charts of patients, and at the same time supports receiving the list of high-risk patients pushed by the system, intervening, and being able to formulate personalized health management plans for patients.

[0129] In a specific embodiment, the security and privacy protection module: encrypts the transmitted data using TLS, stores, manages, and ensures that only authorized personnel can access the relevant data.

[0130] In a specific embodiment, the system integration and interface module: accesses hospital information systems such as HIS and LIS to obtain patients' medical data in real time. In addition, it obtains environmental information such as meteorological data, air quality index, and pollen concentration in real time through the API.

[0131] The standardized interface provides RESTful API to support docking with third-party applications and achieve data sharing.

[0132] In a specific embodiment, a treatment plan for acute allergic events that have occurred or allergic events at risk of occurring. This function is for the allergy risk warning system to push personalized emergency treatment guidance suggestions according to the onset and severity of different symptoms when assessing the possible risk of allergic attacks or allergic reactions in children.

[0133] In another specific embodiment, the children's allergy management system is divided into a basic resource layer, a data resource layer, a data management layer, a system application layer, and a user layer. Communication between each layer is carried out through a standardized interface to ensure the flexibility and scalability of the system. As Figure 2 shown. The basic resource layer provides the hardware and network infrastructure required for the system to run, including network devices, computing devices, storage devices, and terminal devices. This layer supports the deployment and operation of Spring Boot microservices to ensure the high availability and elasticity of the system. In addition, the basic resource layer also integrates information security components such as Web application firewalls, web page anti-tampering, and anti-virus gateways to protect the system from external attacks. The data resource layer is mainly responsible for the generation, collection, and storage of data. This layer uses Postgres as the main database to support complex queries and transaction processing to ensure data consistency and integrity. At the same time, Redis is used as a cache database to accelerate the reading speed of frequently accessed data and improve the response efficiency of the system. The data resource layer also implements data interfaces with external systems to support real-time data synchronization and batch import.

[0134] The data management layer is built on top of the data resource layer and is responsible for the processes of data integration, cleaning, transformation, and loading (ETL), such as physical examination data, questionnaire data, risk models, and indicator data. This layer adopts the method of domain-driven design (DDD), encapsulating business logic in domain models to ensure that the business rules of the system are clear and easy to maintain. The data management layer also implements fine-grained data permission management, based on the Spring Security framework, to ensure the security and compliance of data access. Specifically, it includes a data preprocessing module, a data mining module, and a data service module. The data system application layer is the layer directly facing users and provides rich functional applications. This layer is built based on Spring Cloud, supporting functions such as service discovery, configuration management, circuit breakers, and load balancing to ensure reliable communication between microservices and the high availability of the system. The system application layer also implements business logic based on DDD, such as "Internet +", artificial intelligence, and big data analysis. Specifically targeting the specialty characteristics of childhood allergic diseases, it provides a closed-loop management solution. In addition, this layer also implements user identity authentication and access authorization to ensure the security and privacy of user data.

[0135] The user layer provides an intuitive user interface and interaction experience, supporting multi-terminal access, including PCs, smartphones, and tablets. This layer communicates with the system application layer through RESTful APIs to ensure the rapid response to user requests and the real-time update of data. The design of the user layer focuses on the user experience, providing a simple and easy-to-use interface and a smooth operation process.

[0136] In a specific embodiment, childhood allergic diseases are chronic inflammatory diseases that require a long-term, continuous, and standardized treatment management process, which includes steps such as screening, monitoring, evaluation, diagnosis and treatment, intervention, and follow-up. A closed-loop management system of interaction and linkage among hospitals, society, family doctors, the Internet, children, and parents should be established to enhance personalized treatment for the unique needs of each child and enable children to receive effective management throughout the process. By designing an effective management process throughout the whole process and applying it to daily home detection and health guidance, it promotes the transformation of childhood allergic diseases from "disease diagnosis and treatment-centered" to "health promotion-centered" and "active health", with the aim of achieving the goals of controlling diseases, improving the quality of life, and improving the prognosis. Based on this, the present invention has developed a digital and intelligent chronic disease management tool for allergic children - "Mintong Management" (risk assessment system for allergy occurrence) based on the WeChat platform. The application interfaces of the user end and the doctor end are as Figure 3 , Figure 4 shown.

[0137] Figure 5 The present invention provides an embodiment with a method for risk assessment of allergy occurrence, including:

[0138] Obtain the basic data of the person to be tested, including physiological index data, environmental data, and medical record data;

[0139] Input the basic data into a risk assessment model for evaluation to obtain a risk result, and the risk result includes one or more of the following: no allergy risk, low allergy risk, medium allergy risk, high allergy risk;

[0140] The construction process of the risk assessment model is as follows:

[0141] Obtain the basic data set of the person to be tested and the labels of risk categories;

[0142] Input the data set and the labels of risk categories into a neural network for training to obtain a risk assessment model;

[0143] Among them, the neural network is initialized with parameters based on the oscillation period, and the initial parameters are calculated through the product of the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the associated parameters of the neural network.

[0144] In one embodiment, during the construction process of the risk assessment model, the oscillation period adaptively adjusts to optimize the iterative process of the neural network. When the loss of the loss function fluctuates violently during the iteration process, the oscillation amplitude is adjusted and then the iteration is carried out.

[0145] In one embodiment, during the construction process of the risk assessment model, random perturbations are added after the iterative parameter update to optimize the parameter update range of the neural network, and the size of the random perturbation is dynamically adjusted according to the number of training iterations.

[0146] In one embodiment, the construction of the evaluation model further includes data augmentation; the data set and the labels of risk categories are augmented to obtain augmented data, and the augmented data is input into the neural network model for training to obtain a risk assessment model; the data augmentation is carried out through a generative adversarial network, and the loss function calculation of the generative adversarial network includes adversarial loss and feature contrast loss. The calculation of the adversarial loss includes random noise data, feature data of allergy medical data, and the calculation of the feature contrast loss includes allergy medical data. The feature data of the allergy medical data includes one or more of the following: statistical features, semantic features, image features, time series features, specific medical index features; the allergy medical data includes one or more of the following: heart rate, blood pressure, air quality index, pollen concentration, allergy history, type of medication used, age, gender, temperature and humidity index, family allergy history.

[0147] In one embodiment, the generative adversarial network further includes a distribution adaptation evaluation mechanism, including a diversity loss and a quality evaluation loss. The quality evaluation loss measures the quality difference by comparing the mean and variance between the generated allergy medical data and the real allergy medical data. The diversity loss is obtained by calculating the diversity loss factor and the generated samples.

[0148] The disclosed embodiments of the present invention also provide a computer program product or system, including a computer program, which when executed by a processor implements the steps of the above-mentioned risk assessment method for allergy occurrence.

[0149] Figure 6 The schematic diagram of the risk assessment device for allergy occurrence provided by the embodiments of the present invention specifically includes:

[0150] A memory and a processor; the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, any of the above-mentioned risk assessment methods for allergy occurrence.

[0151] The disclosed embodiments of the present invention also provide a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, any of the above-mentioned risk assessment methods for allergy occurrence.

[0152] The verification results of this verification embodiment show that allocating fixed weights for indications can improve the performance of this method compared to the default settings. Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here. In several embodiments provided by this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in each embodiment of the present invention, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units. Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The storage medium can include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk, or optical disc, etc.

[0153] Those of ordinary skill in the art can understand that all or part of the steps in implementing the methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned medium storage can be read-only memory, magnetic disk, or optical disc, etc.

[0154] The above has introduced in detail a computer device provided by the present invention. For those of ordinary skill in the art, according to the idea of the embodiments of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A management system for children's allergies, characterized in that: include: Danger warning Module: used for early warning of risks of acute allergic reactions; The risk warning module includes: Data collection module: used to obtain basic data of the subject, including physiological index data, environmental data, and medical record data; Risk assessment module: used to input the basic data into the risk assessment model for assessment to obtain risk results; The construction process of the risk assessment model is as follows: Obtain the basic data set of the subject and the label of the risk category; Inputting the data set and risk category labels into a neural network for training to obtain a risk assessment model; The neural network performs parameter initialization based on the oscillation period, and the initialization parameters are calculated by multiplying the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the associated parameters of the neural network.

2. The children's allergy management system according to claim 1, characterized in that: During the construction of the risk assessment model, the oscillation period is adaptively adjusted to optimize the iteration process of the neural network. When the loss of the loss function fluctuates violently during the iteration process, the iteration is performed after adjusting the oscillation amplitude.

3. The children's allergy management system according to claim 1, characterized in that: The risk assessment model construction process adds random perturbations to optimize the parameter update range of the neural network after iterative parameter updates, and the size of the random perturbations is dynamically adjusted according to the number of training iterations.

4. The management system for children's allergies according to claim 1, characterized in that: The construction of the assessment model also includes data expansion; data expansion is performed on the data set and the labels of the risk categories to obtain expanded data, and the expanded data is input into a neural network model for training to obtain a risk assessment model; the data expansion is performed through a generative adversarial network, and the loss function calculation of the generative adversarial network includes adversarial loss and feature contrast loss, the calculation of the adversarial loss includes random noise data, feature data of allergic medical data, and the feature contrast loss calculation includes allergic medical data.

5. The management system for children's allergies according to claim 4, characterized in that: The characteristic data of the allergy medical data include one or more of the following: statistical features, semantic features, image features, time series features, and specific medical indicator features; the allergy medical data include one or more of the following: heart rate, blood pressure, air quality index, pollen concentration, allergy history, type of medication used, age, gender, temperature and humidity index, and family allergy history.

6. The management system for children's allergies according to claim 1, characterized in that: The generative adversarial network also includes a distribution adaptive evaluation mechanism, including diversity loss and quality assessment loss. The quality assessment loss is measured by comparing the mean and variance between the generated allergy medical data and the real allergy medical data to measure the quality difference; the diversity loss is calculated by the diversity loss factor and the generated samples.

7. The management system for children's allergies according to claim 1, characterized in that: The system may also include one or more of the following: Allergy early identification module: used for early identification and screening of symptoms related to allergic diseases; Allergy relief monitoring module: used to record the detection data during the allergy relief period; Allergy detection intelligent reading module: used for intelligent interpretation of allergen detection reports; Allergy emergency plan module: used to push the treatment plan for acute allergic events that have occurred or the risk of allergic events; Allergy and non-allergic disease management module: used for health management of allergic and non-allergic diseases related to children with allergies.

8. A method for assessing the risk of allergy, characterized in that: include: Obtain basic data of the subject, including physiological index data, environmental data, and medical record data; The basic data is input into a risk assessment model for evaluation to obtain a risk result, wherein the risk result includes one or more of the following: no allergy risk, low allergy risk, moderate allergy risk, and high allergy risk; The construction process of the risk assessment model is as follows: Obtain the basic data set of the subject and the label of the risk category; Inputting the data set and risk category labels into a neural network for training to obtain a risk assessment model; The neural network performs parameter initialization based on the oscillation period, and the initialization parameters are calculated by multiplying the weight adjustment factor, the oscillation period, the sine term or cosine term of the oscillation period, and the associated parameters of the neural network.

9. A computer device comprising a processor, a memory and a computer program or instruction stored in the memory, characterized in that: The computer program or instructions are executed by the processor to implement the risk assessment method for allergy occurrence according to claim 8.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: The computer program or instructions are executed by a processor to implement the risk assessment method for allergy occurrence as described in claim 8.

Images