Multi-factor collaborative influenza prevalence trend prediction method and system

Abstract

The invention belongs to the technical field of digital medical treatment, and discloses a multi-factor collaborative influenza prevalence trend prediction method and system, and the method comprises the steps: building an extreme value distribution, gamma distribution and propagation dynamics model, comprehensively predicting the influenza prevalence trend from multiple angles, carrying out the weight distribution and normalization through a root-mean-square error, and carrying out the prediction of the influenza prevalence trend. A more reasonable preliminary prediction result is obtained; combining a DLNM model and an artificial intelligence model, and comprehensively considering the influence of meteorological conditions on influenza propagation; the influence of the activeness of different types of individuals is calculated, and linear addition is performed to obtain the influence of the total population activeness on the influenza prevalence trend, so that the prediction result is closer to reality; according to the public health decision-making method, the accuracy and the practicability of prediction are improved, and powerful support is provided for public health decision-making.

Description

Technical Field

[0001] The present invention belongs to the field of digital medical technology and relates to a method and system for predicting influenza epidemic trends with multi-factor collaboration. Background Art

[0002] Currently, the prediction of influenza epidemic trends mainly relies on transmission dynamics models, such as the classic SEIR model. However, such models have shown significant limitations in practical applications, and their prediction accuracy is often questioned. The main reasons include: First, the model is highly dependent on parameter settings, and it is often difficult to accurately capture the uncertainty of parameters or the complexity of non-linear relationships between parameters; Second, the model is difficult to comprehensively consider all influencing factors, especially the existence of latent variables, such as meteorological factors and population traffic and social behaviors, which have important impacts on influenza epidemics but are difficult to quantify in the model, thus affecting the prediction accuracy.

[0003] To address the above problems, there have been some attempts to improve prediction accuracy by integrating multiple influenza prediction models or introducing new prediction methods. For example, Patent CN118866395A proposes a method and device for predicting influenza trends based on multi-modal data, which uses deep learning technology to incorporate meteorological factors as independent variables into the prediction model. However, there is often a contradiction between the generalization and accuracy of deep learning methods, and a large amount of historical data is required for training to obtain better prediction results. Moreover, it has high requirements for computing power, and the interpretability of the model is relatively insufficient. On the other hand, Patent CN202410434526.4 attempts to use a complex social contact network to quantitatively study the impact of population traffic and social behaviors on the spread of infectious diseases. Although this method is innovative in theory, the parameter settings are complex in actual operation, and for an infectious disease like influenza with non-discrete case distribution and insignificant age distribution characteristics, its prediction significance is relatively limited.

[0004] Therefore, how to quantitatively calculate the meteorological influencing factors in influenza prediction, simplify and quantify the impact of population traffic and social behaviors, and at the same time maintain the interpretability of the model has become a major challenge in current influenza prediction research. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem in the prior art that it is difficult to accurately quantify the impacts of meteorological factors and population traffic and social behaviors in influenza prediction, and at the same time simplify the model parameter settings, and provide a method and system for predicting influenza epidemic trends with multi-factor collaboration.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A method for predicting influenza epidemic trends with multi-factor collaboration includes:

[0008] Obtain the time series data of influenza transmission cases for the influenza subtype to be studied;

[0009] Establish an extreme value distribution model, a gamma distribution model, and a transmission dynamics model, and input the time series data of influenza transmission cases into the established models to obtain the influenza epidemic trend prediction results of the three models respectively;

[0010] Based on the prediction results of the extreme value distribution model, the gamma distribution model, and the transmission dynamics model and their respective corresponding true values, obtain the root mean square error of each model, assign weights to the root mean square error, and obtain the preliminary prediction result after normalization;

[0011] Establish a DLNM model and an artificial intelligence model to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend;

[0012] Based on the meteorological forecast time series data and the preliminary epidemic trend prediction data, calculate the impact of the two on the activity levels of different types of individuals, multiply the individual impact by the total number of people in that type of population to obtain the activity levels of different types of populations and linearly sum them to obtain the impact of the total population activity level on the influenza epidemic trend;

[0013] Quantify the impact of meteorological conditions on the influenza epidemic trend and the impact of the total population activity level on the influenza epidemic trend, and correct the obtained preliminary prediction result to obtain the final prediction result.

[0014] A further improvement of the present invention lies in:

[0015] Further, the extreme value distribution model is

[0016]

[0017] The gamma distribution model is

[0018]

[0019] where y0 is the baseline, A is the multiple relationship, and x c is the x value when y is the maximum value, and α, β, and w are shape parameters.

[0020] Further, the transmission dynamics model divides the population into susceptibles S, latent individuals E, infectives I, and recovered individuals R, where

[0021] The expression for susceptibles S is:

[0022]

[0023] The expression for latent individuals E is:

[0024]

[0025] The expression of the diseased I is:

[0026]

[0027] The expression of the recovered R is:

[0028]

[0029] The boundary condition is

[0030] S (0，a) +E (0，a) +I (0，a) +R (0,a) = 1

[0031] where t is time and α is age; γ (t,a) is the contact infection rate, k (t,a) is the vaccine effectiveness, u (t,a) is the mortality rate, b (t) is the birth rate, σ (t,a) is the proportion of recovered individuals turning into susceptible individuals, a (t,a) is the probability of latent individuals turning into diseased individuals, λ (t,a) is the probability of diseased individuals turning into recovered individuals, d (t,a) is the case fatality rate.

[0032] Furthermore, the weight assignment for the root mean square error is specifically as follows:

[0033]

[0034] where w i is the weight of a certain algorithm, and MSE i is the root mean square error of a certain algorithm; i represents the model index, 0 represents the extreme value distribution model, 1 represents the gamma distribution model, and 2 represents the propagation dynamics model;

[0035] The preliminary prediction result obtained after normalization is specifically: The preliminary predicted value is the weighted average of the predicted values of each model,

[0036]

[0037] where is the predicted value of the i-th model.

[0038] Furthermore, establish a DLNM model and an artificial intelligence model to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend, specifically:

[0039] Based on the historical meteorological conditions and the contemporaneous influenza surveillance data input into the DLNM model, the relative risk value of meteorological conditions on influenza prevalence is calculated, and the obtained relative risk value of meteorological conditions on influenza prevalence is substituted into the meteorological forecast data to obtain the relative risk prediction value RR of meteorological factors on influenza;

[0040] The artificial intelligence model is an ensemble learning model, and the ensemble learning module includes the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm; the collected average temperature, relative humidity, first-order temperature gradient, temperature interval identifier, and the reported number of cases of another influenza subtype are used as data sets and input into the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm respectively, and are integrated through the BPNN algorithm to obtain the infection rate λ of the influenza subtype to be studied t Predicted value;

[0041] Among them, the infection rate λ of the influenza subtype to be studied at time t t is:

[0042]

[0043] Q t =N t -I” t

[0044] Among them, I t is the number of infected people of the influenza subtype to be studied at time t, N t is the total number of patients at time t, I” t is the number of infected people of another influenza subtype at time t, Q t is the difference between the total number of patients at time t and the number of infected people of another influenza subtype, k is the recovery rate, and m is the immune inactivation rate; I t+1 is the uncorrected number of infected people of the influenza subtype to be studied at the next moment; the I t+1 is the preliminary prediction result.

[0045] The RR value coefficient w of the DLNM model and the contact infection rate λ 2-1 coefficient w t are set respectively to obtain the influence of meteorological conditions, which is expressed by the following formula: 2-2 Among them,

[0046]

[0047] Among them, is the influence of meteorological factors on influenza prevalence, and coef is a fixed coefficient.

[0048] Further, based on the meteorological forecast time series data and the preliminary prediction data of the epidemic trend, calculate the influence of both on the activity levels of different types of individuals, multiply the individual influence by the total number of people in that type to obtain the activity levels of different types of people, and linearly sum them up to obtain the influence of the total population activity level on the influenza epidemic trend. Specifically:

[0049] The individual activity level is:

[0050]

[0051] Wherein, is the activity level of the i-th type of individual, is the function of the influence of meteorological factors on the activity level of the i-th type of individual at time t, is the function of the number of reported cases at time t on the activity level of the i-th type of individual, γ i is the i-th type of individual affected by and The coefficient of the combined influence;

[0052] The population activity level is:

[0053]

[0054] Wherein, is the population activity level, k is the total number of individual types, n i is the total number of the i-th type of individual.

[0055] Further, quantify the influence of meteorological conditions on the influenza epidemic trend and the influence of the total population activity level on the influenza epidemic trend, and correct the obtained preliminary prediction results to obtain the final prediction results. Specifically:

[0056]

[0057] Wherein, I' t is the predicted number of reported cases at time t after correction, α is the meteorological factor correction factor, β is the correction factor after taking the logarithm of the population activity factor, and coef2 is the baseline, which is related to the number of people in the area to be predicted and the population classification characteristics.

[0058] The present invention discloses a multi-factor collaborative influenza epidemic trend prediction system, including:

[0059] An acquisition module, which acquires the time series data of influenza transmission cases of the influenza subtype to be studied;

[0060] A construction module, which establishes an extreme value distribution model, a gamma distribution model and a transmission dynamics model, and inputs the time series data of influenza transmission cases into the established models to respectively obtain the influenza epidemic trend prediction results of the three models;

[0061] A normalization module, which obtains the root mean square error of each model based on the prediction results of the extreme value distribution model, gamma distribution model, and propagation dynamics model and their respective corresponding true values, assigns weights to the root mean square errors, and obtains a preliminary prediction result after normalization;

[0062] A prediction module, which establishes a DLNM model and an artificial intelligence model to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend;

[0063] A calculation module, which calculates the impact of meteorological forecast time series data and preliminary epidemic trend prediction data on the activity levels of different types of individuals, multiplies the individual impact by the total number of people of that type to obtain the activity levels of different types of people, and linearly sums them to obtain the impact of the total population activity level on the influenza epidemic trend;

[0064] A correction module, which quantifies the impact of meteorological conditions on the influenza epidemic trend and the impact of the total population activity level on the influenza epidemic trend, and corrects the obtained preliminary prediction result to obtain a final prediction result.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] The present invention comprehensively predicts the influenza epidemic trend from multiple angles by establishing an extreme value distribution, gamma distribution, and propagation dynamics model, improving the accuracy and reliability of the prediction; uses the root mean square error for weight assignment and normalization to obtain a more reasonable preliminary prediction result; combines the DLNM model and the artificial intelligence model to comprehensively consider the impact of meteorological conditions on influenza transmission, enhancing the scientificity and pertinence of the prediction; calculates the impact of the activity levels of different types of individuals and linearly sums them to obtain the impact of the total population activity level on the influenza epidemic trend, making the prediction result closer to the actual situation; finally, corrects the preliminary prediction result to obtain a final prediction result, further improving the accuracy and practicality of the prediction, and providing strong support for public health decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0068] Figure 1 It is a schematic flowchart of a method for predicting the influenza epidemic trend with multi-factor collaboration of the present invention;

[0069] Figure 2Schematic structural diagram of a multi-factor collaborative influenza epidemic trend prediction system according to the present invention;

[0070] Figure 3 Another process schematic diagram of a multi-factor collaborative influenza epidemic trend prediction method according to the present invention;

[0071] Figure 4 Schematic diagram of the prediction and fitting results of the extreme value distribution model, gamma distribution model and propagation dynamics model;

[0072] Figure 5 Schematic diagram of the input and output of the integrated learning model;

[0073] Figure 6 Schematic diagram of the correction results of meteorological factors and population activity;

[0074] Figure 7 Schematic diagram of the true value, preliminary prediction value and final prediction value. Detailed implementation manners

[0075] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, 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. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0076] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0077] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0078] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the inventive product is usually placed during use, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, terms such as "first", "second", etc. are only used for descriptive distinction and cannot be construed as indicating or implying relative importance.

[0079] In addition, when the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and it does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0080] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, when the terms "arranged", "installed", "connected", and "coupled" appear, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0081] The present invention will be further described in detail below with reference to the accompanying drawings:

[0082] See Figure 1 , the present invention discloses a multi-factor collaborative influenza epidemic trend prediction method, including:

[0083] S101, obtaining the time series data of influenza transmission cases of the influenza subtype to be studied;

[0084] S102, establishing an extreme value distribution model, a gamma distribution model, and a transmission dynamics model, and inputting the time series data of influenza transmission cases into the established models to respectively obtain the influenza epidemic trend prediction results of the three models;

[0085] The extreme value distribution model is

[0086]

[0087]

[0088] The gamma distribution model is

[0089]

[0090] where y0 is the baseline, A is the multiple relationship, and x c is the x value when y is the maximum value, and α, β, and w are shape parameters.

[0091] The transmission dynamics model divides the population into susceptibles S, latent individuals E, infected individuals I, and recovered individuals R, where

[0092] The expression of susceptibles S is:

[0093]

[0094] The expression of the latent individuals R is:

[0095]

[0096] The expression of the infected individuals I is:

[0097]

[0098] The expression of the recovered individuals R is:

[0099]

[0100] The boundary conditions are

[0101] S (0，a) +E (0，a) +I (0，a) +R (0，a) = 1

[0102] where t is time and α is age; γ (t,a) is the contact infection rate, k (t,a) is the vaccine effectiveness, u (t,a) is the mortality rate, b (t) is the birth rate, σ (t,a) is the proportion of recovered individuals turning into susceptible individuals, a (t,a) is the probability of latent individuals turning into infected individuals, λ (t,a) is the probability of infected individuals turning into recovered individuals, d (t,a) is the case fatality rate.

[0103] S103. Based on the prediction results of the extreme value distribution model, the gamma distribution model, and the propagation dynamics model and their respective corresponding true values, obtain the root mean square error of each model, assign weights to the root mean square error, and after normalization, obtain the preliminary prediction result;

[0104] The assignment of weights to the root mean square error is specifically:

[0105]

[0106] where w i is the weight of a certain algorithm, and MSE i is the root mean square error of a certain algorithm; i represents the model index, 0 represents the extreme value distribution model, 1 represents the gamma distribution model, and 2 represents the propagation dynamics model;

[0107] The result obtained after normalization to get the preliminary prediction result is specifically: The preliminary predicted value is the weighted average of the predicted values of each model,

[0108]

[0109] where, is the predicted value of the i-th model.

[0110] S104. Establish a comprehensive prediction of the impact of meteorological conditions on the influenza epidemic trend by using a DLNM model and an artificial intelligence model;

[0111] Input historical meteorological conditions and contemporaneous influenza surveillance data into the DLNM model to calculate the relative risk value of meteorological conditions on influenza epidemic, and substitute the obtained relative risk value of meteorological conditions on influenza epidemic into meteorological forecast data to obtain the relative risk prediction value RR of meteorological factors on influenza;

[0112] The artificial intelligence model is an ensemble learning model, and the ensemble learning module includes the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm; input the collected average temperature, relative humidity, first-order temperature gradient, temperature interval identifier, and the reported number of cases of another influenza subtype as a data set into the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm respectively, and perform integration through the BPNN algorithm to obtain the infection rate λ of the influenza subtype to be studied t Predicted value;

[0113] Among them, the infection rate λ of the influenza subtype to be studied at time t t is:

[0114]

[0115] Q t = N t - I” t ;

[0116] Among them, I t is the number of infected people of the influenza subtype to be studied at time t, N t is the total number of patients at time t, I” t is the number of infected people of another influenza subtype at time t, Q t is the difference between the total number of patients at time t and the number of infected people of another influenza subtype, k is the recovery rate, and m is the immune inactivation rate; I t+1 is the uncorrected number of infected people of the influenza subtype to be studied at the next moment; the I t+1 is the preliminary prediction result.

[0117] Set the RR value coefficient w 2-1 of the DLNM model and the contact infection rate λ t coefficient w 2-2 , and obtain the impact of meteorological conditions, which is expressed by the following formula:

[0118]

[0119] Among them, For the influence of meteorological factors on the influenza epidemic, coef is a fixed coefficient.

[0120] S105. Based on the meteorological forecast time series data and the preliminary epidemic trend prediction data, calculate the influence of both on the activity levels of different types of individuals. Multiply the individual influence by the total number of people in that type to obtain the activity levels of different types of people and linearly sum them to obtain the influence of the total population activity level on the influenza epidemic trend.

[0121] The individual activity level is:

[0122]

[0123] Where is the activity level of the i-th type of individual, is the function of the influence of meteorological factors on the activity level of the i-th type of individual at time t, is the function of the influence of the reported number of cases at time t on the activity level of the i-th type of individual, γ i is the i-th type of individual affected by and The coefficient of the combined influence;

[0124] The population activity level is:

[0125]

[0126] Where is the population activity level, k is the total number of individual types, n i is the total number of the i-th type of individual.

[0127] S106. Quantify the influence of meteorological conditions on the influenza epidemic trend and the influence of the total population activity level on the influenza epidemic trend, and correct the obtained preliminary prediction results to obtain the final prediction results.

[0128]

[0129] Where I' t is the predicted reported number of cases at time t after correction, α is the meteorological factor correction factor, β is the correction factor after taking the logarithm of the population activity level factor, and coef2 is the baseline, which is related to the number of people in the area to be predicted and the population classification characteristics.

[0130] See Figure 2 , The present invention discloses a multi-factor collaborative influenza epidemic trend prediction system, including:

[0131] An acquisition module, which acquires the time series data of influenza transmission cases of the influenza subtype to be studied.

[0132] A construction module that establishes an extreme value distribution model, a gamma distribution model, and a propagation dynamics model, and inputs the influenza transmission case time series data into the established models to obtain the influenza epidemic trend prediction results of the three models respectively;

[0133] A normalization module that, based on the prediction results of the extreme value distribution model, the gamma distribution model, and the propagation dynamics model and their respective corresponding true values, obtains the root mean square error of each model, assigns weights to the root mean square error, and after normalization, obtains a preliminary prediction result;

[0134] A prediction module that establishes a DLNM model and an artificial intelligence model to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend;

[0135] A calculation module that, based on the meteorological forecast time series data and the preliminary epidemic trend prediction data, calculates their impacts on the activity levels of different types of individuals, multiplies the individual impact by the total number of people of that type to obtain the activity levels of different types of people, and linearly sums them to obtain the impact of the total population activity level on the influenza epidemic trend;

[0136] A correction module that quantifies the impact of meteorological conditions on the influenza epidemic trend and the impact of the total population activity level on the influenza epidemic trend, and corrects the obtained preliminary prediction result to obtain the final prediction result.

[0137] Example:

[0138] See Figure 3 , the present invention discloses a method for predicting the influenza epidemic trend with multi-factor collaboration. According to the existing influenza surveillance data in a certain province from the 44th to the 51st week of 2024, the meteorological forecast data in the 52nd week of 2024 and the 1st week of 2025, and the relevant data of the population in the province, the influenza epidemic trend in the 52nd week of 2024 and the 1st week of 2025 in the province is predicted.

[0139] Step 1: Obtain the time series data of influenza reported cases from the 44th to the 51st week of 2024 at weekly intervals;

[0140] The acquisition of the number of influenza cases is not limited to the influenza surveillance system, and can also be obtained from relevant records such as hospitals or disease research institutes.

[0141] Step 2, establish an extreme value distribution model, a gamma distribution model, and a propagation dynamics (SEIR) model, input the infectious disease time series data in Step 1 into the above three models, and respectively obtain the influenza epidemic trend prediction results.

[0142] The extreme value distribution model is

[0143]

[0144] The gamma distribution model is

[0145]

[0146] where y0 is the baseline, A is the multiple relationship, and x c is the x value when y is the maximum, and α, β, and w are shape parameters.

[0147] The propagation dynamics algorithm divides the population into four categories: susceptibles S, latents E, infecteds I, and recovered R. Among them,

[0148] The expression for susceptibles S is:

[0149]

[0150] The expression for latents E is:

[0151]

[0152] The expression for infecteds I is:

[0153]

[0154] The expression for recovered R is:

[0155]

[0156] The boundary condition is

[0157] S (0，a) +E (0,a) +I (0，a) +R (0，a) = 1

[0158] The proportion of the number of each category in the total population is affected by two states: time and age. The boundary condition is that S + E + I + R = 1 when t = 0. In the mathematical structure, t is time, α is age, γ (t,a) is the contact infection rate, k (t,a) is the vaccine effectiveness, u (t,a) is the mortality rate, b (t) is the birth rate, σ (t,a) is the proportion of recovered individuals turning into susceptibles, a (t,a) is the probability of latents turning into infecteds, λ (t,a) is the probability of infecteds turning into recovered individuals, d (t,a) is the case fatality rate. In single epidemic prediction, the influencing factors of population birth rate and mortality rate can be ignored. Therefore, in this embodiment, b (t) = 0, u (t,a) = 0.

[0159] Calculate the root mean square error (MSE) between the prediction results of the above three models and the existing true values, automatically assign weights, and obtain the preliminary prediction results of the integration after normalization. The weight calculation method is as follows:

[0160]

[0161] where w i is the weight of a certain algorithm, and MSE i is the root mean square error of a certain algorithm. It can be seen from the principle of this algorithm that the preliminary prediction results obtained by integrating the three algorithms and automatically assigning weights according to MSE have a smaller error than the currently commonly used traditional single propagation dynamics algorithm. The results are shown in Figure 4 . Figure 4 Among them, comparing the MSE of the SEIR prediction result = 38.66 and the MSE of the integrated algorithm prediction result = 36.19 proves that the integrated algorithm has a lower MSE and higher accuracy.

[0162] Step 3: Obtain the meteorological forecast data for the 52nd week of 2024 and the 1st week of 2025 on a weekly basis.

[0163] In this embodiment, based on the historical influenza surveillance and meteorological data from 2012 to 2022, the DLNM model is used to obtain the relative risk values RR of average temperature for influenza A and influenza B respectively. The temperature points with an RR value of 1 are 15°C for influenza A, and -1.9°C and 25°C for influenza B. Substitute the meteorological forecast data for the 52nd week of 2024 and the 1st week of 2025 into the DLNM model to obtain the predicted relative risk values.

[0164] See Figure 5 , based on the historical influenza surveillance and meteorological data from 2012 to 2022, an artificial intelligence model is established to predict the risk of contact infection. The created model is an integrated learning model, where the input independent variables are average temperature, relative humidity, first-order temperature gradient, temperature interval identifier, and the reported number of cases of another influenza subtype. They are respectively input into the BPNN algorithm, RFR algorithm, XGBoostR algorithm, and kNNR algorithm, and integrated through the BPNN algorithm to obtain the infection rate λ t predicted value.

[0165] In Figure 5Among them, for the temperature range identification, for influenza A, when the average temperature T < 8 degrees Celsius, Flag = -1; when T ∈ [8, 15], Flag = 0; when T > 15 degrees Celsius, Flag = 1. For influenza B, when the average temperature T < 10 degrees Celsius, Flag = 1; when T >= 10 degrees Celsius, Flag = 0. The BPNN algorithm is the backpropagation neural network algorithm, RFR is the random forest regression algorithm, XGBoostR is the extreme gradient ascent regression algorithm, and kNNR is the k-nearest neighbor regression algorithm. After each of the above algorithms calculates the result, the BPNN algorithm is used for integration.

[0166] Among them, the contact infection rate λ at time t t is defined by the following formula:

[0167]

[0168] Q t = N t - I” t

[0169] Among them, I t is the number of infected people with the influenza subtype to be studied at time t, N t is the total number of medical consultations at time t, and I” t is the number of infected people with another influenza subtype at time t, all of which can be obtained from the influenza surveillance system. Q t is the difference between the total number of medical consultations at time t and the number of infected people with another influenza subtype. k is the recovery rate set to 0.2, and m is the immune inactivation rate set to 0.00274. I t+1 is the uncorrected number of infected people with the influenza subtype to be studied at the next moment, which is predicted by step 2. After the model is trained, the meteorological forecast data for the 52nd week of 2024 and the 1st week of 2025 are substituted to obtain the t predicted value of λ.

[0170] The influenza subtypes are divided into influenza A and influenza B, and there is a correlation between influenza A and influenza B. When performing ensemble learning, when predicting the λ of a certain subtype of influenza t the number of cases I” of another subtype is input t which can improve the prediction performance. For example, if predicting influenza A, the result will be better when the number of cases of influenza B is used as the independent variable. Therefore, there is such a logic: if predicting the λ of future influenza A t+1 it is necessary to know the future I” of influenza B in advance t+1 . In practical operations, the I” of influenza B t+1 does not need to be accurate. Using the time series of the previous influenza B subtype cases as the original data, the I” of influenza B is obtained through the preliminary prediction method (step 2) t+1 、I” t+2 、I” t+3……The preliminary prediction results are used as one of the independent variables and input into this ensemble learning model to predict the λ of future influenza A t+1 , λ t+2 , λ t+3 …….

[0171] The RR value coefficient w 2-1 and the contact infection rate λ t coefficient w 2-2 are set respectively. The influence of meteorological factors is as follows:

[0172]

[0173] In the formula is the influence of meteorological factors on the influenza epidemic, coef is a fixed coefficient. In this embodiment, preferably, it is 0.220 for influenza A and 0.215 for influenza B, w 2-1 = 1, w 2-2 = 48.3. It should be noted that the above parameters are related to the intrinsic characteristics of the area to be predicted and do not need to be changed once set. The influence results of meteorological factors are shown in Figure 6 .

[0174] Step 4, the population in population activity consists of multiple types of people, and the distribution characteristics of the population types are obtained through publicly released statistical bulletins. In this embodiment, the total population is 85.05 million people, and according to the active characteristics related to the influenza epidemic, it is divided into four major types, among which there are 16 million students (type 1), 38.8 million workers (type 2), 21.28 million farmers (type 3), and 8.97 million others (type 4).

[0175] To reduce the calculation amount, the present invention calculates the mean value of the activity of all individuals in each type of population. The individual activity is defined by the following formula:

[0176]

[0177] Among them, is the activity of the i-th type of individual, is the function of the influence of meteorological factors on the activity of the i-th type of individual at time t, is the function of the number of reported cases at time t on the activity of the i-th type of individual, γ i is the coefficient of the i-th type of individual affected by and synergistic influence.

[0178] In this embodiment, preferably:

[0179] For type 1 individuals, is a piecewise function of temperature T:

[0180] When T < 10;

[0181] When 10 ≤ T < 15;

[0182] When 15 ≤ T < 23;

[0183] When 23 ≤ T < 29;

[0184] When T > 29.

[0185] It is the reported number of cases I t Function:

[0186]

[0187] Cross - term γ1 = 0.15

[0188] For type 2 individuals, It is a piece - wise function of temperature T:

[0189] When T < 10;

[0190] When 10 ≤ T < 15;

[0191] When 15 ≤ T < 23;

[0192] When 23 ≤ T < 29;

[0193] When T > 29.

[0194] It is the reported number of cases I t Function:

[0195]

[0196] Cross - term γ2 = 0.2

[0197] For type 3 individuals, It is a piece - wise function of temperature T:

[0198] When T < 5;

[0199] When 5 ≤ T < 15;

[0200] When 15 ≤ T < 25;

[0201] When 25 ≤ T < 30;

[0202] When T > 30.

[0203] is the reported number of cases I t Function:

[0204]

[0205] The cross - term γ3 = 0.1

[0206] For type 4 individuals, is a piece - wise function of temperature T:

[0207] When T < 10;

[0208] When 10 ≤ T < 15;

[0209] When 15 ≤ T < 23;

[0210] When 23 ≤ T < 29;

[0211] When T > 29.

[0212] is the reported number of cases I t Function:

[0213]

[0214] The cross - term γ4 = 0.175

[0215] Substitute the above equations and parameters into the following formula:

[0216]

[0217] where, is the population activity, k is the total number of population types, in this embodiment k = 4, n i is the total number of the i - th type of individuals, substitute the meteorological and population - related parameters to obtain the influence of population activity.

[0218] Substitute the temperature parameter and the population activity factor for correction, the equation is:

[0219]

[0220] where, I' tLet \(I\) be the predicted number of patients in the prediction report at time \(t\) after correction, \(\alpha\) be the meteorological factor correction factor, \(\beta\) be the correction factor after taking the logarithm of population activity; \(coef2\) is the baseline value, which is related to the total number of people in the area to be predicted and the population classification characteristics. In this embodiment, \(\alpha = 0.0119\), \(\beta = 0.25\), and \(coef2 = 18.56\).

[0221] All equations and their parameters in step 4 have been built into the system.

[0222] Step 5, the finally corrected result \(I'\) t is plotted as a time series curve, and the result is as Figure 7 shown. Figure 7 It can be seen that the deviation between the initial prediction result and the true value in the 52nd week of 2024 is -15.1%, and the deviation from the 1st week of 2025 is -19.7%; the deviation between the finally corrected prediction result and the true value in the 52nd week of 2024 is -7.5%, and the deviation from the 1st week of 2025 is -13.0%. The results show that compared with the initial prediction value, after quantitatively correcting the meteorological influence factor and the population activity influence factor, the finally predicted result can be greatly approximated to the true value, and the influenza epidemic trend can be predicted more accurately, which has a significant advantage compared with the existing technology.

[0223] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-factor collaborative influenza epidemic trend prediction method, characterized in that include: Obtain time series data on influenza transmission cases of the influenza subtype to be studied; Establish extreme value distribution model, gamma distribution model and transmission dynamics model, input influenza transmission case time series data into the established models, and obtain influenza epidemic trend prediction results of the three models respectively; Based on the prediction results of the extreme value distribution model, gamma distribution model and propagation dynamics model and their corresponding true values, the root mean square error of each model is obtained, the root mean square error is weighted and normalized to obtain the preliminary prediction results; Establish DLNM model and artificial intelligence model to comprehensively predict the impact of meteorological conditions on influenza epidemic trends; Based on the meteorological forecast time series data and preliminary epidemic trend forecast data, the impact of the two on the activity of different types of individuals was calculated. The individual impact was multiplied by the total number of people of that type to obtain the activity of different types of people and then linearly added to obtain the impact of the total population activity on the influenza epidemic trend; The impact of meteorological conditions and the impact of total population activity on influenza epidemic trends are quantified, and the preliminary forecast results are revised to obtain the final forecast results.

2. A multi-factor collaborative influenza epidemic trend prediction method according to claim 1, characterized in that, The extreme value distribution model is The gamma distribution model is Among them, y0 is the baseline, A is the multiple relationship, and x c is the x value when y is at its maximum, and α, β, and w are shape parameters.

3. A multi-factor collaborative influenza epidemic trend prediction method according to claim 1, characterized in that The transmission dynamics model divides the population into susceptible people S, latent people E, sick people I and recovered people R, where: The expression of susceptible person S is: The expression of the lurker E is: The expression of patient I is: The expression of the recovered person R is: The boundary condition is S (0，a) +E (0，a) +I (0，a) +R (0，a) = 1 Among them, t is time, α is age; γ (t,a) is the contact infection rate, k (t,) is the vaccine effectiveness, u (t,a) is the mortality rate, b (t) is the birth rate, σ (t,a) is the proportion of recovered individuals turning into susceptible individuals, a (t,a) is the probability of latent individuals turning into patients, λ (t,a) is the probability of patients turning into recovered individuals, d (t,a) is the case fatality rate.

4. A multi-factor collaborative influenza epidemic trend prediction method according to claim 3, characterized in that, The weight distribution of the root mean square error is specifically as follows: where w i is the weight of a certain algorithm, and MSE i is the root mean square error of a certain algorithm; i represents the model index, 0 represents the extreme value distribution model, 1 represents the gamma distribution model, and 2 represents the propagation dynamics model; After the normalization, the initial prediction result is obtained, specifically: the initial prediction value is the weighted average of the prediction values ​​of each model, Among them, is the predicted value of the i-th model.

5. A multi-factor collaborative influenza epidemic trend prediction method according to claim 4, characterized in that, The DLNM model and the artificial intelligence model are established to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend, specifically: Based on the historical meteorological conditions and influenza surveillance data of the same period, the relative risk value of meteorological conditions for influenza epidemics was calculated and substituted into the meteorological forecast data to obtain the relative risk prediction value RR of meteorological factors for influenza. The artificial intelligence model is an ensemble learning model, and the ensemble learning module includes the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm; the collected average temperature, relative humidity, first-order temperature gradient, temperature range identifier, and the reported number of cases of another influenza subtype are used as data sets and input into the BPNN algorithm, the RFR algorithm, the XGBoostR algorithm, and the kNNR algorithm respectively, and are integrated through the BPNN algorithm to obtain the infection rate λ of the influenza subtype to be studied t Predicted value; Among them, the infection rate λ of the influenza subtype to be studied at time t t is as follows: Q t = N t - I″ t Among them, I t is the number of infected people with the influenza subtype to be studied at time t, N t is the total number of patients visiting the doctor at time t, I″ t is the number of infected people with another influenza subtype at time t, Q t is the difference between the total number of patients visiting the doctor at time t and the number of infected people with another influenza subtype, k is the recovery rate, and m is the immune inactivation rate; I t+1 is the uncorrected number of infected people with the influenza subtype to be studied at the next moment; the I t+1 is the preliminary prediction result. Set the RR value coefficient w of the DLNM model separately 2-1 and the contact infection rate λ t Coefficient w 2-2 , to obtain the influence of meteorological conditions, which is expressed by the following formula: Among them, is the impact of meteorological factors on the influenza epidemic, and coef is a fixed coefficient.

6. A multi-factor collaborative influenza epidemic trend prediction method according to claim 5, characterized in that Based on the weather forecast time series data and the preliminary epidemic trend forecast data, the impact of the two on the activity of different types of individuals is calculated, and the individual impact is multiplied by the total number of people of this type to obtain the activity of different types of people and linearly add them to obtain the impact of the total population activity on the influenza epidemic trend, specifically: The individual activity is: Among them, is the activity level of the i-th type of individual, is a function of the influence of meteorological factors on the activity level of the i-th type of individual at time t, is a function of the influence of the reported number of patients on the activity level of the i-th type of individual at time t, γ i is the i-th type of individual affected by and is the coefficient of the combined influence; The activity of the population is: Among them, is the population activity, k is the total number of individual type categories, and n i is the total number of the i-th type of individuals.

7. A multi-factor collaborative influenza epidemic trend prediction method according to claim 6, characterized in that The influence of meteorological conditions on the influenza epidemic trend and the influence of the total population activity on the influenza epidemic trend are quantified, and the obtained preliminary prediction results are corrected to obtain the final prediction results, which are specifically: where I' t is the predicted number of patients in the corrected prediction report at time t, α is the meteorological factor correction factor, β is the correction factor after taking the logarithm of the population activity factor, and coef2 is the baseline, which is related to the number of people in the area to be predicted and the population classification characteristics.

8. A multi-factor collaborative influenza epidemic trend prediction system, characterized in that, include: An acquisition module, wherein the acquisition module acquires time series data of influenza transmission cases of the influenza subtype to be studied; A construction module, wherein the construction module establishes an extreme value distribution model, a gamma distribution model, and a transmission dynamics model, and inputs the influenza transmission case time series data into the established models to obtain influenza epidemic trend prediction results of the three models respectively; A normalization module, which obtains the root mean square error of each model based on the prediction results and their corresponding true values of the extreme value distribution model, gamma distribution model, and propagation dynamics model, assigns weights to the root mean square error, and obtains a preliminary prediction result after normalization; A prediction module, which establishes a DLNM model and an artificial intelligence model to comprehensively predict the impact of meteorological conditions on the influenza epidemic trend; A calculation module, which calculates the impacts of both on the activity levels of different types of individuals based on the meteorological forecast time series data and the preliminary epidemic trend prediction data, multiplies the individual impact by the total number of people in that type of population, linearly sums up the activity levels of different types of populations, and obtains the impact of the total population activity level on the influenza epidemic trend; A correction module, which quantifies the impact of meteorological conditions on the influenza epidemic trend and the impact of the total population activity level on the influenza epidemic trend, and corrects the obtained preliminary prediction result to obtain a final prediction result.

Images