System and method for preventing respiratory tract infection of children

Abstract

The invention relates to the technical field of health risk assessment, in particular to a system and a method for preventing respiratory tract infection of children, an environmental data analysis module is used for collecting air temperature, humidity and air quality parameters focused on a pediatric activity place and carrying out data cleaning and standardization processing; according to the invention, through deep processing of environment monitoring data of a pediatric activity place, including hyperfine data cleaning and precise standardization, the reliability of analysis is comprehensively guaranteed, and a foundation is built for subsequent model operation. Innovative daily mean value calculation and optimized time sequence analysis are closely attached to the environment change rhythm of the pediatric department, and the dynamic trend of environment parameters is clearly drawn. A composite machine learning strategy of a random forest and a support vector machine is fused, pediatric exclusive features are deeply embedded, the precision and adaptability of a disease prediction model are greatly improved, and the complexity and variability of pediatric data are leisurely handled.

Description

Technical Field

[0001] The present invention relates to the technical field of health risk assessment, and in particular to a system and method for preventing respiratory tract infections in children. Background Art

[0002] Pediatric respiratory infections are common diseases among children and have a great impact on their healthy growth. The system for preventing pediatric respiratory infections aims to accurately predict the probability of respiratory disease outbreaks and potential health impacts in children through in-depth analysis of the environmental data of the pediatric population and real-time health monitoring data, providing strong support for early prevention and control.

[0003] However, the current technologies for predicting pediatric respiratory infections have many shortcomings. On the one hand, existing disease prediction models are not well adapted to the pediatric population, and it is difficult to fully consider the unique physiological characteristics and living environment of children. As a result, when faced with volatile pediatric data, the effectiveness of the model is greatly reduced, and it is impossible to accurately capture the potential patterns of disease occurrence. On the other hand, the traditional fixed warning threshold lacks flexibility in pediatric scenarios. Because children have relatively weak immunity and their condition changes more rapidly, fixed thresholds are difficult to match the rapid fluctuations in children's health status in real time, which can easily lead to untimely or inaccurate warnings and miss the best time for prevention and control. Therefore, there is an urgent need to improve existing technologies to meet the urgent needs of pediatric respiratory infection prevention. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a system and method for preventing respiratory tract infections in children.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: A system for preventing respiratory tract infections in children comprises:

[0006] The environmental data analysis module collects and focuses on collecting the temperature, humidity and air quality parameters of pediatric activity places, performs data cleaning and standardization, calculates the daily average value of each environmental parameter, and generates the daily average value data of the pediatric environment; performs time series analysis on the daily average value data of the pediatric environment, determines the parameter change trend, and generates the pediatric environment trend analysis results;

[0007] The disease prediction module, based on the pediatric environmental trend analysis results, combined with the pediatric respiratory infection case data, fully considers the personalized factors of children's age, physique, and vaccination history, and performs data training through random forests and support vector machines to obtain a preliminary disease prediction model designed specifically for pediatrics. The preliminary disease prediction model is used for simulation prediction to obtain comprehensive pediatric disease prediction results;

[0008] A threshold dynamic adjustment module performs early warning detection and generates a dynamic early warning threshold according to the comprehensive pediatric disease prediction result;

[0009] The early warning information output module obtains the pediatric infection risk level of the region according to the dynamic early warning threshold, and then integrates the regional pediatric health data and real-time monitoring data to obtain the regional health risk results, and outputs highly targeted early warning information based on the regional pediatric health risk results to obtain early warning results.

[0010] Preferably, the steps for obtaining the daily average data of the pediatric environment are:

[0011] Collect the temperature, humidity and air quality parameters of pediatric activity places, perform strict data cleaning on the original data, remove noise and outliers, and obtain cleaned pediatric environmental data;

[0012] Based on the cleaned pediatric environment data, the weighted daily average of each parameter was calculated using the following formula:

[0013]

[0014] Among them, q i represents the parameter value of a single data point, N is the total number of data points, and w i is the weight coefficient associated with the data point time, AVG q is the weighted daily average;

[0015] Based on the weighted daily average value, the daily environmental average value data for each day is summarized.

[0016] Preferably, the steps for obtaining the pediatric environment trend analysis results are:

[0017] Apply time series analysis to the daily average data of the pediatric environment, use differential operation to preprocess each parameter, and after eliminating seasonal fluctuations, obtain the processed pediatric parameter change sequence;

[0018] Based on the processed pediatric parameter change sequence, the overall trend value of each parameter is calculated using the following formula:

[0019]

[0020] Among them, T trend is the overall trend value, Δq i is the change of environmental parameters within a time period, γ i is the adjustment factor of the time period, and M is the number of time periods;

[0021] Based on the overall trend value, a pediatric environment trend analysis result is generated to obtain analyzed pediatric environment trend data.

[0022] Preferably, the steps for obtaining the preliminary pediatric disease prediction model are:

[0023] Collect the pediatric environmental trend analysis results and pediatric respiratory infection historical data, integrate the data and perform comprehensive preprocessing, including missing value processing and outlier processing rules based on the child's growth stage segmentation data, to obtain an integrated pediatric data set;

[0024] Based on the integrated pediatric data set, the random forest algorithm training data is applied, and the random forest prediction model is obtained by fine-tuning the number of trees and the maximum depth in the model parameters;

[0025] Based on the random forest prediction model, the support vector machine algorithm was applied, the kernel function type and regularization parameters were adjusted to optimize the classification boundary, and the model was trained until convergence to obtain a preliminary pediatric disease prediction model.

[0026] Preferably, the steps for obtaining the dynamic warning threshold are:

[0027] Based on the comprehensive pediatric disease prediction results, missing value processing and outlier detection are performed, and the data is normalized to obtain a standardized prediction result data set;

[0028] Based on the standardized prediction result data set, the dynamic warning threshold is calculated using the following formula:

[0029]

[0030] Among them, R i is the infection risk score of the i-th prediction result, is the weight factor of risk score, V i is the variance of each prediction result, SN is the total number of results, α and β are adjustment coefficients, and T is the dynamic warning threshold.

[0031] Preferably, the steps for obtaining the pediatric health risk results of the region are:

[0032] According to the dynamic warning threshold, the risk level of pediatric health monitoring data in each region is evaluated, and the risk is graded according to whether the probability of infection is higher or lower than the dynamic warning threshold, so as to obtain a preliminary pediatric risk level classification for each region;

[0033] Collect and integrate pediatric health monitoring data and real-time medical monitoring data from various regions, including children's hospital admission rates, pediatric outpatient visit symptoms, and children's vaccination status, and fuse the data to form a comprehensive regional pediatric health data set;

[0034] Based on the comprehensive regional pediatric health data set and preliminary pediatric risk level classification, risk comparison and trend prediction were conducted to evaluate the comprehensive pediatric health risks in each region. The regional pediatric health risk results were obtained by referring to the pediatric epidemic development trend and the regional pediatric public health resource distribution.

[0035] Preferably, the steps of obtaining the warning result are:

[0036] Based on the pediatric health risk results of the regions, high-risk regions are identified for the pediatric infection rate and severe illness rate in each region, and a regional pediatric risk level list is obtained;

[0037] Design and compile warning information content based on the pediatric risk level list of the region, including child-friendly protective action guidelines, children's hospitals, kindergartens, and primary and secondary school places with high child density, and obtain customized warning information drafts;

[0038] The customized draft warning information is broadcast through the regional education department notification, the official website of the medical institution and the news media platform, and the coverage of the information and the response of the public are monitored to obtain the warning results.

[0039] The present invention provides a method for preventing respiratory tract infection in children, comprising the following steps:

[0040] Collect temperature, humidity and air quality parameters at meteorological stations and air quality monitoring points around pediatric activity sites, record and summarize data, clean and standardize the summarized data, remove abnormal data and correct deviations, calculate the daily average values ​​of temperature, humidity and air quality index, and generate daily average records of pediatric environments;

[0041] Use the daily average value records of the pediatric environment for time series analysis, use moving average and exponential smoothing to analyze the changing trends of temperature, humidity, and air quality index, predict the pediatric environment conditions, and generate pediatric environment trend forecasts;

[0042] Combining the pediatric environmental trend forecast with the pediatric respiratory infection case data, the number of seasonal pediatric infection cases was collected, and the correlation between the two was analyzed using linear regression. A preliminary disease transmission model based on pediatric environmental factors was constructed, and simulation predictions were performed to obtain a pediatric disease transmission risk assessment.

[0043] According to the pediatric disease transmission risk assessment, the current pediatric infection risk level is assessed, and the warning threshold is dynamically adjusted by calculating the deviation between the current infection data and the threshold to generate a dynamic warning threshold;

[0044] Based on the dynamic warning threshold, the pediatric health risk level of each region is assessed, the real-time pediatric health monitoring data of the region is collected and compared with the dynamic warning threshold, high-risk areas are identified, customized warning information is generated and released to the public and health departments, and pediatric health warning reports are obtained in the monitored areas.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, through the in-depth processing of environmental monitoring data of pediatric activity places, including ultra-fine data cleaning and precise standardization, the reliability of the analysis is fully guaranteed, laying a solid foundation for subsequent model operations. The innovative daily average calculation and optimized time series analysis closely fit the rhythm of pediatric environmental changes and clearly outline the dynamic trend of environmental parameters. The composite machine learning strategy that integrates random forests and support vector machines deeply embeds pediatric-specific features, greatly improves the accuracy and adaptability of disease prediction models, and calmly copes with the complexity and variability of pediatric data. The high-frequency adaptive adjustment of dynamic warning thresholds closely tracks children's real-time health data to ensure that the early warning system always maintains a high degree of sensitivity and accurately captures every change in health risks. The early warning information customized for the specific health risk situation of pediatrics in the region perfectly meets the needs of children and parents, and is both interesting and practical. It provides all-round and strong support for the prevention and control of pediatric respiratory infections, and effectively protects the healthy growth of children. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0049] See also Figure 1 The present invention provides a technical solution: a system for preventing respiratory tract infections in children comprises:

[0050] The environmental data analysis module collects and focuses on collecting the temperature, humidity and air quality parameters of pediatric activity places, performs data cleaning and standardization, calculates the daily average value of each environmental parameter, and generates the daily average value data of the pediatric environment; performs time series analysis on the daily average value data of the pediatric environment, determines the parameter change trend, and generates the pediatric environment trend analysis results;

[0051] The disease prediction module, based on the pediatric environmental trend analysis results, combined with the pediatric respiratory infection case data, fully considers the personalized factors of children's age, physique, and vaccination history, and performs data training through random forests and support vector machines to obtain a preliminary disease prediction model designed specifically for pediatrics. The preliminary disease prediction model is used for simulation prediction to obtain comprehensive pediatric disease prediction results;

[0052] A threshold dynamic adjustment module performs early warning detection and generates a dynamic early warning threshold according to the comprehensive pediatric disease prediction result;

[0053] The early warning information output module obtains the pediatric infection risk level of the region according to the dynamic early warning threshold, and then integrates the regional pediatric health data and real-time monitoring data to obtain the regional health risk results, and outputs highly targeted early warning information based on the regional pediatric health risk results to obtain early warning results.

[0054] The steps for obtaining the daily average data of the pediatric environment are as follows:

[0055] Collect the temperature, humidity and air quality parameters of pediatric activity places, perform strict data cleaning on the original data, remove noise and outliers, and obtain cleaned pediatric environmental data;

[0056] Based on the cleaned pediatric environment data, the weighted daily average of each parameter was calculated using the following formula:

[0057]

[0058] Among them, q i represents the parameter value of a single data point, N is the total number of data points, and w i is the weight coefficient associated with the data point time, AVG q is the weighted daily average;

[0059] Based on the weighted daily average value, the daily environmental average value data for each day is summarized.

[0060] Specifically, based on the collection of temperature, humidity and air quality parameters in the pediatric activity area, the three parameters are first retrieved one by one in chronological order and the collection time is marked. Then, the duplicated parameter records are compared, and the records with overlapping timestamps and excessively large numerical differences are deleted. In combination with the range provided by the on-site sensor equipment, it is checked whether there are abnormal values ​​outside the range. Finally, according to the range limits of the temperature, humidity and air quality sensors in the instrument calibration records, the records that exceed the range are directly eliminated, and the noise fluctuations are checked according to the hourly segments. After aggregation, the detection software performs low-amplitude smoothing on the remaining data points, marks the time information corresponding to each data record and compares it with the mean value of the same type of parameters pre-counted in the previous step, and identifies and eliminates suspicious data by judging the deviation of multiple consecutive data points from the mean within the sampling period. After data cleaning, the cleaned environmental data is obtained.

[0061] The benefit of the formula is that by incorporating the time-related weight factors into the calculation, the observation frequency and data validity can be fully reflected in multiple sampling periods, making the calculation of the daily average value of the pediatric environment more accurate and avoiding the excessive impact of extreme data on the results. The steps for obtaining each parameter are as follows: air temperature data is collected in real time through an ambient temperature sensor, humidity data is obtained through a capacitive humidity sensor, and air quality parameters use optical particle sensors and electrochemical sensors to monitor inhalable particles and major gases. The weight coefficient is calculated based on the stability of the sampling period and the sampling frequency of the equipment. The size of the weight parameter fluctuates with the reliability of the data during the day or night.

[0062] Calculation process:

[0063] Let the temperature be The humidity value is The air quality value is For example, a total of N=6 groups of data were collected in 6 time periods on a certain day. The temperature values ​​monitored by the temperature sensor were 24.2℃, 25.0℃, 23.8℃, 24.5℃, 25.3℃, and 24.1℃, respectively. After weight calculation, the corresponding weights of each time period are w1=1.05, w2=1.00, w3=0.98, w4=1.03, w5=1.02, and w6=1.01. In the weighted calculation of the temperature, first find the numerator:

[0064] (1.05×24.2)+(1.00×25.0)+(0.98×23.8)+(1.03×24.5)+(1.02×25.3)+(1.01×24.1);

[0065] =25.41+25.00+23.324+25.235+25.806+24.341=149.115;

[0066] Find the denominator:

[0067] 1.05+1.00+0.98+1.03+1.02+1.01=6.09;

[0068] Divide the two:

[0069]

[0070] The results show that the weighted average daily temperature for that day was approximately 24.49°C. If this value deviates from a predetermined comfortable temperature range, it can be used as an important reference for subsequent adjustments to the air conditioning or ventilation system.

[0071] Based on the weighted daily average values ​​of temperature, humidity and air quality obtained by weighted calculation in the previous step, record the time stamps corresponding to these weighted daily average values, store the three parameters in parallel according to the weighted calculation results on the same day, retrieve the environmental data log that has been cleaned and calculated, and summarize the weighted daily average value of temperature, humidity and air quality on the same date into the same entry according to the matching date or corresponding index, and identify them with a unified date label. On this basis, perform a one-time summary of the weighted daily average values ​​of temperature, humidity and air quality on that day, and then further group them according to the daily classification principle set previously. Number the sorted data day by day and mark them in the environmental monitoring database. Finally, construct the daily environmental average data for each day based on the grouped daily data.

[0072] The steps to obtain the results of the pediatric environment trend analysis are:

[0073] Apply time series analysis to the daily average data of the pediatric environment, use differential operation to preprocess each parameter, and after eliminating seasonal fluctuations, obtain the processed pediatric parameter change sequence;

[0074] Based on the processed pediatric parameter change sequence, the overall trend value of each parameter is calculated using the following formula:

[0075]

[0076] Among them, T trend is the overall trend value, Δq i is the change of environmental parameters within a time period, γ i is the adjustment factor of the time period, and M is the number of time periods;

[0077] Based on the overall trend value, a pediatric environment trend analysis result is generated to obtain analyzed pediatric environment trend data.

[0078] Specifically, time series analysis is applied to the daily average environmental data. Combined with the daily average environmental data obtained in the previous step, the parameter values ​​of adjacent days are located through the clear date labels in the monitoring records, and the temperature, humidity, air quality and other data between adjacent days are arranged in order. An appropriate differential order is determined and a differential operation is performed in the data. The difference between adjacent elements of each parameter in each step of the differential calculation is recorded. The obtained differential sequence is then tested item by item, and the time period with abnormal jumps is reviewed. By referring to other time periods in the same day or combining stable sections of multiple days, it is determined whether the jump exceeds a limit threshold calculated based on the historical daily average. The threshold is usually obtained based on the coefficient of variation method in statistics. First, the ratio of the standard deviation to the mean of the daily average of the same parameter in the same area in history is calculated, and then multiplied by a proportional factor determined in the field experience data, the result is To the observable limiting threshold, if the jump of the differential sequence is higher than this threshold, the data segment is marked as a mutation segment and further difference smoothing is performed. The abnormal values ​​are corrected by consulting the previously stored data of the monitoring point in the same period to ensure the continuity and reliability of the data after the difference. Then, according to the seasonal cycle of the area to which the collected data belongs, the previous differenced sequence is divided according to a cycle length such as 7 days or 30 days. The mean of the sequence in each cycle unit is calculated and compared with the mean of the adjacent cycle unit to identify the components with periodic repeated rise and fall. Then, the multi-month average values ​​of temperature and humidity in the same area announced by the historical meteorological station are used to infer the possible seasonal amplitude range. If it is found in the comparison that the current differential sequence obviously repeats the cycle fluctuation, this part of the seasonal component is removed from the data, and the remaining values ​​are recombined as the final difference preprocessing result to obtain the processed parameter change sequence.

[0079] The formula is useful in that by introducing the adjustment factor γ i And combined with Δq i The numerical value can reflect both the magnitude of parameter changes and the impact of different time periods in trend analysis, so that the overall trend value can better reflect the true evolution of the parameter over time.

[0080] Δq i The steps of obtaining γ are as follows: based on the daily average data obtained above, record the numerical difference of the same environmental parameter in adjacent time periods, γ i The steps for obtaining M are to quantitatively evaluate the working status of the sampling equipment, the stability of the monitoring position and the intensity of external interference in each time period, convert these evaluation results into numerical values ​​according to the set conversion rules, and change with environmental fluctuations. The steps for obtaining M are to determine it according to the number of time periods selected for analysis.

[0081] Calculation process:

[0082] Take M = 4, record Δq1 = 1.2, Δq2 = -0.4, Δq3 = 0.9, Δq4 = 1.6, and record the adjustment factors as γ1 = 0.15, γ2 = 0.10, γ3 = 0.35, γ4 = 0.20, and first calculate the numerator:

[0083]

[0084] Specific expansion:

[0085] (1.2×0.8696) 2 +(-0.4×0.9091) 2 +(0.9×0.7407) 2 +(1.6×0.8333) 2 ;

[0086] =(1.0435) 2 +(-0.3636) 2 +(0.6666) 2 +(1.3333) 2 ;

[0087] =1.0893+0.1322+0.4444+1.7777;

[0088] =3.4436;

[0089] Taking the square root of this result, we get Divide by M=4, we get

[0090]

[0091] The results show that in the four selected time periods, the overall trend value is around 0.4635. If the value is higher than 1.0, it means that the parameter fluctuation range is larger and the adjustment factor is subject to stronger interference. A value around 0.5 usually indicates a relatively mild change.

[0092] Based on the overall trend value, first read the overall trend value calculated in the previous paragraph from the parameter change sequence after differential preprocessing and seasonal fluctuation elimination and combine it with the differential data of the current monitoring period, then select the length of an observation period such as one week or ten days, superimpose the environmental parameter sequence within this period one by one and generate the corresponding change index, and detect whether each superimposed value exceeds a benchmark threshold obtained from historical records. The benchmark threshold depends on the conventional floating range when the same parameter is measured in the same area. The benchmark threshold is calculated by recording the fluctuation of the daily average value measured on the spot for many times and using the maximum and minimum range. If the superimposed value exceeds the threshold, further retrieve the daily average analysis content summarized in the previous step, check the distribution of the interference factors that can be confirmed in the interval, and make numerical corrections to the time period with obvious interference. By querying the equipment calibration information or the parameters of the adjacent time period to adjust the abnormal sampling amount in the record of this section, recalculate the differential and superimposed results after the correction, compare the new superimposed value with the benchmark threshold again, and continue to check the processing results of all time periods. Finally, summarize the data of all time periods to form the overall result of trend analysis and obtain the analyzed environmental trend data.

[0093] The steps to obtain the preliminary pediatric disease prediction model are:

[0094] Collect the pediatric environmental trend analysis results and pediatric respiratory infection historical data, integrate the data and perform comprehensive preprocessing, including missing value processing and outlier processing rules based on the child's growth stage segmentation data, to obtain an integrated pediatric data set;

[0095] Based on the integrated pediatric data set, the random forest algorithm training data is applied, and the random forest prediction model is obtained by fine-tuning the number of trees and the maximum depth in the model parameters;

[0096] Based on the random forest prediction model, the support vector machine algorithm was applied, the kernel function type and regularization parameters were adjusted to optimize the classification boundary, and the model was trained until convergence to obtain a preliminary pediatric disease prediction model.

[0097] Specifically, we collected the results of environmental trend analysis and historical data on respiratory infections, checked the record sources and formats of the two parts of data, unified the field names and value types, confirmed the time series correspondence of each parameter, used neighboring interpolation for missing values ​​and determined the inserted value based on the difference between adjacent observations and the time span, set upper and lower limits of the constraint range for outliers according to historical fluctuations, and eliminated or corrected them if they exceeded the upper and lower limits. After all data were initially processed, they were aligned to the same time series index and processed centrally, and all processed data information was sorted out to obtain an integrated data set.

[0098] Based on the previously integrated data set, several environmental feature fields and respiratory infection records were selected as analysis targets. The time order was maintained when dividing the training set and the test set to prevent information leakage. Relevant parameters were extracted and quantified for each training sample record. When setting the number of trees in the random forest algorithm, the statistical distribution of the reference data in different feature dimensions was considered, and a better total number of trees was selected in combination with the changes in training error and the node division level. The maximum depth was used to control the number of branch layers of each tree and reduce the problem of overfitting. After the training converged, the parameters with the best performance were selected to obtain the random forest prediction model.

[0099] Based on the result output of the random forest prediction model, its probability distribution of sample classification is input into the support vector machine algorithm. First, different kernel function types are tried according to the linear separability of sample features. After comparing the classification accuracy and convergence speed of each kernel function after training, the appropriate kernel function is selected and different regularization parameters are tested. Each adjustment of the regularization intensity will cause changes in the classification boundary, so it is necessary to repeat the training and observe the accuracy and convergence iteration. When the training process tends to be stable, the optimal regularization parameter is determined and the overall training of the support vector machine is completed to obtain a preliminary disease prediction model.

[0100] The steps for obtaining the dynamic warning threshold are:

[0101] Based on the comprehensive pediatric disease prediction results, missing value processing and outlier detection are performed, and the data is normalized to obtain a standardized prediction result data set;

[0102] Based on the standardized prediction result data set, the dynamic warning threshold is calculated using the following formula:

[0103]

[0104] Among them, R i is the infection risk score of the i-th prediction result, is the weight factor of risk score, V i is the variance of each prediction result, SN is the total number of results, α and β are adjustment coefficients, and T is the dynamic warning threshold.

[0105] Specifically, based on the comprehensive disease prediction results, first confirm the basic structure of each field in the data set and match the identification information such as date and sample number one by one. After viewing the original records, retrieve those fields that are displayed as empty values, and determine the filling method by comparing the numerical differences between adjacent observation records and the time span of the detection interval. For example, use the average value of the two time periods to replace the missing value. If the numerical difference between adjacent records is too large, use the insertion ratio principle and insert the value in segments. Next, perform outlier detection on all samples, determine the regular floating range with reference to historical statistical data, and define the limit interval in combination with the standard deviation multiple or quantile difference method. When performing the comparison, traverse all fields and use the recorded upper and lower limits to judge each sample value. If an obvious Abnormal points that obviously exceed the limit range are reviewed in combination with the surrounding time periods and sample information of the same batch. Those that are determined to be occasional or caused by unstable measuring instruments are eliminated or reassigned. After the abnormal values ​​are processed, the numerical distribution is considered as a whole. The maximum and minimum normalization or mean-variance normalization scheme is applied to parameter fields with different dimensions. The values ​​of each field are mapped to the same comparable interval through traversal. After the numerical range tends to be consistent, all normalized results are integrated into a list, in which each sample has an aligned and standardized value under the same attribute column. Finally, they are sorted and numbered in combination with the previous identification information, and each sample record is matched with its corresponding comprehensive disease prediction result to form the final standardized prediction result data set.

[0106] The benefit of the formula is that by multiplying and superimposing the infection risk score of each predicted result with the corresponding weight factor, and combining the variance term and adjustment coefficient of each predicted result to appropriately amplify or reduce them, the dynamic warning threshold can reflect the diversified risk scores while taking into account the fluctuations caused by the variance of the results.

[0107] R i The steps of obtaining are as follows: combining the infection risk score calculated for each sample record in the comprehensive disease prediction results obtained above, and recording the numerical value range generally between 0 and 1 through a unified scale. The acquisition steps are to quantify and assign values ​​based on multiple factors such as the stability of on-site testing equipment, data credibility, and the susceptibility of the target population, and there will be certain fluctuations in different periods of time. i The steps for obtaining are to select the scores of the same target group in the same period from the output results of multiple prediction models, and calculate the variance value of the group of scores. The steps for obtaining SN are to count the number of prediction results that have been collected. The steps for obtaining α and β are to refer to historical research data and expert evaluations, and to set them in combination with the volatility requirements of risk scores and the denominator smoothing effect.

[0108] Calculation process:

[0109] Select SN = 5 prediction results, set the infection risk scores to R1 = 0.30, R2 = 0.45, R3 = 0.55, R4 = 0.70, R5 = 0.25, and the corresponding weight factors W R1 =1.05,W R2 =0.90,W R3 =1.10,W R4 =1.00,W R5 =0.95, the variance values ​​are recorded as V1=0.03, V2=0.02, V3=0.05, V4=0.07, V5=0.02, the adjustment coefficients α=0.60, β=0.80 (these values ​​are obtained according to the previous monitoring and calculation process), first calculate the weighted sum of the risk scores in the numerator:

[0110] (0.30×1.05)+(0.45×0.90)+(0.55×1.10)+(0.70×1.00)+(0.25×0.95);

[0111] =0.315+0.405+0.605+0.70+0.2375=2.2625;

[0112] Then calculate the sum of the variance parts and multiply by α:

[0113]

[0114] α×0.0954=0.60×0.0954=0.05724;

[0115] Take the sum of the two parts as the numerator and divide it by the denominator SN + β = 5 + 0.80 = 5.80:

[0116]

[0117] The results show that under the combined influence of the weighted risk scores and variances of the five samples, the dynamic warning threshold is approximately 0.4. When the threshold is greater than 0.6, it often indicates that the overall risk score is concentrated to a certain extent and has a high variance, while a threshold less than 0.3 indicates that the overall risk score is low or the fluctuation is relatively stable.

[0118] The steps to obtain regional pediatric health risk results are:

[0119] According to the dynamic warning threshold, the risk level of pediatric health monitoring data in each region is evaluated, and the risk is graded according to whether the probability of infection is higher or lower than the dynamic warning threshold, so as to obtain a preliminary pediatric risk level classification for each region;

[0120] Collect and integrate pediatric health monitoring data and real-time medical monitoring data from various regions, including children's hospital admission rates, pediatric outpatient visit symptoms, and children's vaccination status, and fuse the data to form a comprehensive regional pediatric health data set;

[0121] Based on the comprehensive regional pediatric health data set and preliminary pediatric risk level classification, risk comparison and trend prediction were conducted to evaluate the comprehensive pediatric health risks in each region. The regional pediatric health risk results were obtained by referring to the pediatric epidemic development trend and the regional pediatric public health resource distribution.

[0122] Specifically, according to the dynamic warning threshold T obtained earlier, we first perform field analysis on the health monitoring data of each region, read the infection probability value reported by each monitoring point and compare it with T one by one. If the infection probability exceeds T, the record is marked as potentially high risk; if it is lower than T, it is marked as relatively low risk. Then, we create a list for each region, in which the risk marking status of all records in the region is summarized. The overall risk level of the region is measured according to the proportion of high-risk marks in the records. After traversing all regions, a risk ratio statistic is generated. When the high-risk ratio of a region exceeds an empirical interval, it is classified as a higher risk level. The empirical interval is usually determined based on historical observation data in the past monitoring cycle. For example, the average ratio of similar periods in the previous year is selected and compared with the upper and lower limits of fluctuations. Finally, the risk levels of each region are marked in a comparison table and sorted. In this way, a preliminary risk level classification of each region is obtained.

[0123] For the health monitoring data and real-time medical monitoring data of different regions, the hospital admission rate field is first read and arranged in chronological order. The infection symptom results reported by each institution are divided into mild, moderate or severe, and are also arranged in chronological order. The vaccination date and vaccine type of the collected vaccination information are confirmed for each person, and this information is matched with the monitoring data of the corresponding area. If it is found that some records have date misalignment or missing vaccination type, they are supplemented and compared with the previously retained time series association or personal identification information. When the vaccination record number provided by the on-site management unit matches the infection symptom result, the coverage status of the current population can be judged. All matched and confirmed data are put into the same data table for unified management. After summarizing the fields, a comprehensive regional health data set containing admission rate, symptom results and vaccination information is formed.

[0124] Based on the integrated regional health data set and the preliminary risk level classification obtained earlier, the target time window is first determined. Within this time range, the historical monitoring records of each region are extracted and the risk change trajectory is statistically analyzed. The admission rate, symptom data and vaccine coverage are compared with the trajectory. If the historical risk level in a certain region is higher than the median risk value previously recorded in the region for several consecutive days, it is marked as a continuous upward trend. If the admission rate increases over time and the symptom distribution moves toward a higher severity, it is further confirmed that there is a larger health threat. After obtaining the comparison results of all regions, this trend information is associated with the number of medical beds or human resources allocation in the public health resource statistics table. By examining the relationship between resource utilization and risk level, it is determined which regions have the possibility of further escalation of risk levels. Finally, all comparison and prediction information are integrated to obtain regional health risk results.

[0125] The steps to obtain the warning results are as follows:

[0126] Based on the pediatric health risk results of the regions, high-risk regions are identified for the pediatric infection rate and severe illness rate in each region, and a regional pediatric risk level list is obtained;

[0127] Design and compile warning information content based on the pediatric risk level list of the region, including child-friendly protective action guidelines, children's hospitals, kindergartens, and primary and secondary school places with high child density, and obtain customized warning information drafts;

[0128] The customized draft warning information is broadcast through the regional education department notification, the official website of the medical institution and the news media platform, and the coverage of the information and the response of the public are monitored to obtain the warning results.

[0129] Specifically, based on the regional health risk results obtained previously, the infection rate and severity rate values ​​of each region are retrieved one by one, and the average infection rate and severity rate of the region in the previous statistical period are used as a reference benchmark. A step-by-step comparison is used to determine whether the value increment in the current monitoring period is higher than the previous fluctuation range of the region. If it is higher than the upper limit of the experience range, it is recorded as a sudden increase state, and the region is marked as a candidate high-risk region during risk verification. All regions that meet the sudden increase state are collected and their infection rate and severity rate values ​​are summarized for secondary screening to detect whether the conditions for a significant increase in the severity rate are met at the same time. A threshold range can be set for this condition, which is often derived from the load bearing capacity of medical institutions of the same level and historical observation data. When a region is found to reach or exceed this severity rate threshold, it is directly included in the list of high-risk areas. Finally, the risk level records of each region that have been finally confirmed are integrated to obtain a regional risk level list.

[0130] According to the regional risk level list obtained earlier, the population characteristics corresponding to each risk level are analyzed, and the information of susceptible populations is retrieved from the historical monitoring records of each district to confirm the type of intervention measures required. When compiling the early warning information content, first select the appropriate key points of the resident action guide according to the local traffic environment and the guidance of the traffic management department, and then list the specific addresses and contact information of pediatric activity venues based on the distribution of dangerous areas. Next, select appropriate key items from the common epidemic prevention requirements obtained from local health management units, and list these key points together with the action guide. When arranging the manuscript, follow the principle of conciseness and conciseness, write text descriptions paragraph by paragraph according to the pre-agreed layout, and attach a list of key points. After review and proofreading, finalize the draft and generate a customized early warning information draft.

[0131] Through notifications from local education departments, official websites of medical institutions and news media platforms, we first review the previously compiled list of target audience distribution, schedule the broadcast of the draft warning information in accordance with the external release rules of each region, confirm the push time and release scope one by one and match them with the communication channels used by the main local communities, then upload the draft warning information to the management end in text and image form. The system will send the warning information to the target population in the wide area network in sequence according to the established schedule, and record the push completion rate and feedback interaction volume of each platform in the background. If the coverage rate of certain areas is lower than a certain benchmark, the sender settings will be reviewed separately and the push process will be restarted. Finally, a data set including the push coverage and public response is formed, and the warning results are compiled.

[0132] The present invention provides a method for preventing respiratory tract infection in children, comprising the following steps:

[0133] Collect temperature, humidity and air quality parameters at meteorological stations and air quality monitoring points around pediatric activity sites, record and summarize data, clean and standardize the summarized data, remove abnormal data and correct deviations, calculate the daily average values ​​of temperature, humidity and air quality index, and generate daily average records of pediatric environments;

[0134] Use the daily average value records of the pediatric environment for time series analysis, use moving average and exponential smoothing to analyze the changing trends of temperature, humidity, and air quality index, predict the pediatric environment conditions, and generate pediatric environment trend forecasts;

[0135] Combining the pediatric environmental trend forecast with the pediatric respiratory infection case data, the number of seasonal pediatric infection cases was collected, and the correlation between the two was analyzed using linear regression. A preliminary disease transmission model based on pediatric environmental factors was constructed, and simulation predictions were performed to obtain a pediatric disease transmission risk assessment.

[0136] According to the pediatric disease transmission risk assessment, the current pediatric infection risk level is assessed, and the warning threshold is dynamically adjusted by calculating the deviation between the current infection data and the threshold to generate a dynamic warning threshold;

[0137] Based on the dynamic warning threshold, the pediatric health risk level of each region is assessed, the real-time pediatric health monitoring data of the region is collected and compared with the dynamic warning threshold, high-risk areas are identified, customized warning information is generated and released to the public and health departments, and regional pediatric health warning reports are obtained.

Claims

1. A system for preventing respiratory tract infections in children, characterized in that: The system comprises: The environmental data analysis module collects and focuses on collecting the temperature, humidity and air quality parameters of pediatric activity places, performs data cleaning and standardization, calculates the daily average value of each environmental parameter, and generates the daily average value data of the pediatric environment; performs time series analysis on the daily average value data of the pediatric environment, determines the parameter change trend, and generates the pediatric environment trend analysis results; The disease prediction module, based on the pediatric environmental trend analysis results, combined with the pediatric respiratory infection case data, fully considers the personalized factors of children's age, physique, and vaccination history, and performs data training through random forests and support vector machines to obtain a preliminary disease prediction model designed specifically for pediatrics. The preliminary disease prediction model is used for simulation prediction to obtain comprehensive pediatric disease prediction results; A threshold dynamic adjustment module performs early warning detection and generates a dynamic early warning threshold according to the comprehensive pediatric disease prediction result; The early warning information output module obtains the pediatric infection risk level of the region according to the dynamic early warning threshold, and then integrates the regional pediatric health data and real-time monitoring data to obtain the regional health risk results, and outputs highly targeted early warning information based on the regional pediatric health risk results to obtain early warning results.

2. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps for obtaining the daily average data of the pediatric environment are as follows: Collect the temperature, humidity and air quality parameters of pediatric activity places, perform strict data cleaning on the original data, remove noise and outliers, and obtain cleaned pediatric environmental data; Based on the cleaned pediatric environment data, the weighted daily average of each parameter was calculated using the following formula: Among them, q i represents the parameter value of a single data point, N is the total number of data points, and w i is the weight coefficient associated with the data point time, AVG q is the weighted daily average; Based on the weighted daily average value, the daily environmental average value data for each day is summarized.

3. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps for obtaining the pediatric environment trend analysis results are as follows: Apply time series analysis to the daily average data of the pediatric environment, use differential operation to preprocess each parameter, and after eliminating seasonal fluctuations, obtain the processed pediatric parameter change sequence; Based on the processed pediatric parameter change sequence, the overall trend value of each parameter is calculated using the following formula: Among them, T trend is the overall trend value, Δq i is the change of environmental parameters within a time period, γ i is the adjustment factor of the time period, and M is the number of time periods; Based on the overall trend value, a pediatric environment trend analysis result is generated to obtain analyzed pediatric environment trend data.

4. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps for obtaining the preliminary pediatric disease prediction model are: Collect the pediatric environmental trend analysis results and pediatric respiratory infection historical data, integrate the data and perform comprehensive preprocessing, including missing value processing and outlier processing rules based on the child's growth stage segmentation data, to obtain an integrated pediatric data set; Based on the integrated pediatric data set, the random forest algorithm training data is applied, and the random forest prediction model is obtained by fine-tuning the number of trees and the maximum depth in the model parameters; Based on the random forest prediction model, the support vector machine algorithm was applied, the kernel function type and regularization parameters were adjusted to optimize the classification boundary, and the model was trained until convergence to obtain a preliminary pediatric disease prediction model.

5. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps for obtaining the dynamic warning threshold are: Based on the comprehensive pediatric disease prediction results, missing value processing and outlier detection are performed, and the data is normalized to obtain a standardized prediction result data set; Based on the standardized prediction result data set, the dynamic warning threshold is calculated using the following formula: Among them, R i is the infection risk score of the i-th prediction result, is the weight factor of risk score, V i is the variance of each prediction result, SN is the total number of results, α and β are adjustment coefficients, and T is the dynamic warning threshold.

6. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps to obtain the pediatric health risk results in the region are: According to the dynamic warning threshold, the risk level of pediatric health monitoring data in each region is evaluated, and the risk is graded according to whether the probability of infection is higher or lower than the dynamic warning threshold, so as to obtain a preliminary pediatric risk level classification for each region; Collect and integrate pediatric health monitoring data and real-time medical monitoring data from various regions, including children's hospital admission rates, pediatric outpatient visit symptoms, and children's vaccination status, and fuse the data to form a comprehensive regional pediatric health data set; Based on the comprehensive regional pediatric health data set and preliminary pediatric risk level classification, risk comparison and trend prediction were conducted to evaluate the comprehensive pediatric health risks in each region. The regional pediatric health risk results were obtained by referring to the pediatric epidemic development trend and the regional pediatric public health resource distribution.

7. The system for preventing respiratory tract infections in children according to claim 1, characterized in that: The steps for obtaining the warning result are: Based on the pediatric health risk results of the regions, high-risk regions are identified for the pediatric infection rate and severe illness rate in each region, and a regional pediatric risk level list is obtained; Design and compile warning information content based on the pediatric risk level list of the region, including child-friendly protective action guidelines, children's hospitals, kindergartens, and primary and secondary school places with high child density, and obtain customized warning information drafts; The customized draft warning information is broadcast through the regional education department notification, the official website of the medical institution and the news media platform, and the coverage of the information and the response of the public are monitored to obtain the warning results.

8. A method for preventing respiratory tract infections in children, characterized in that: The system for preventing respiratory tract infections in children according to any one of claims 1 to 7 comprises the following steps: Collect temperature, humidity and air quality parameters at meteorological stations and air quality monitoring points around pediatric activity sites, record and summarize data, clean and standardize the summarized data, remove abnormal data and correct deviations, calculate the daily average values ​​of temperature, humidity and air quality index, and generate daily average records of pediatric environments; Use the daily average value records of the pediatric environment for time series analysis, use moving average and exponential smoothing to analyze the changing trends of temperature, humidity, and air quality index, predict the pediatric environment conditions, and generate pediatric environment trend forecasts; Combining the pediatric environmental trend forecast with the pediatric respiratory infection case data, the number of seasonal pediatric infection cases was collected, and the correlation between the two was analyzed using linear regression. A preliminary disease transmission model based on pediatric environmental factors was constructed, and simulation predictions were performed to obtain a pediatric disease transmission risk assessment. According to the pediatric disease transmission risk assessment, the current pediatric infection risk level is assessed, and the warning threshold is dynamically adjusted by calculating the deviation between the current infection data and the threshold to generate a dynamic warning threshold; Based on the dynamic warning threshold, the pediatric health risk level of each region is assessed, the real-time pediatric health monitoring data of the region is collected and compared with the dynamic warning threshold, high-risk areas are identified, customized warning information is generated and released to the public and health departments, and regional pediatric health warning reports are obtained.

Images