Multi-element coupling epidemic propagation prediction method and system

Abstract

The invention discloses a multivariate coupling epidemic propagation prediction method and system, and relates to the technical field of epidemic prediction, and the method comprises the steps: obtaining a current optimal state estimation value and policy intervention intensity based on an extended Kalman filtering algorithm in combination with an optimal state estimation value at a previous moment and a real observation error; aggregating the social public opinion text risk scores to obtain a social information pressure index, generating group alertness in combination with a three-state model, and determining climate regulation factors according to meteorological data; constructing a dynamic regulation factor and a dynamic propagation rate based on the group alertness, the climate regulation factor, the policy intervention intensity and the external input risk, and substituting the dynamic regulation factor and the dynamic propagation rate into the differential equation set to calculate a state evolution sequence; and through game theory multi-objective optimization, iteratively minimizing a comprehensive decision loss function containing economic, healthy and fatigue weights, and outputting an epidemic situation state prediction curve and optimal policy intervention intensity. Unified coupling of multiple subsystems is achieved, prediction precision is improved, quantitative decision support is provided for scientific prevention and control, and the method is suitable for complex epidemic propagation scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of epidemic prediction, in particular to a multi-element coupled epidemic transmission prediction method and system. BACKGROUND

[0002] The existing epidemic transmission prediction method is based on the classic SEIR (Susceptible-Exposed-Infected-Recovered) model, and attempts to integrate multiple factors by improving the model (such as multi-region cell SEIR, etc.). For example, multi-region cell SEIR model, information-disease coupled model (UA-WA-FA), climate coupled beta (T, RH) regression model.

[0003] The multi-region cell model regards the city as a node, represents the flow of the population through a fixed migration matrix, and describes the cross-regional transmission by dividing the geographical unit and introducing a fixed migration matrix. However, the model is connected by a fixed M in each cell, and the transmission mechanism does not consider dynamic adjustment or external disturbance; the states in each cell are independent of each other, lack a unified modeling framework, and are prone to state dimension fragmentation; the information-disease coupled model (UA-WA-FA) describes the influence of information diffusion on protective behavior and simulates the dynamic adjustment of social media information on individual protective behavior, but the model introduces a behavior layer, which is separated from the SEIR transmission structure and cannot be directly coupled with spatial transmission, policy optimization, etc. The behavior mechanism lacks a mechanism for dynamic interaction with the transmission chain. The climate coupled model uses a linear or exponential function to fit the influence of temperature and humidity on the transmission rate, and associates temperature and humidity with the transmission rate through an empirical regression function. These methods have been locally optimized in a single dimension (such as spatial transmission, behavior feedback, or climate modulation), but have not achieved system integration of multiple modules. It is unable to handle multiple viral strains competition and immune escape at the same time.

[0004] The above defects collectively result in the poor performance of existing models in complex epidemic scenarios: static parameters and fragmented module design make it difficult to capture the real-time dynamics of multi-viral strain competition, cross-border transmission, and the linkage effects of behavior-policy-climate; the lack of a unified framework and dynamic optimization capability limits the prediction accuracy and decision support value. These problems highlight the need to build a multi-factor coupled, real-time updated, and scalable prediction system. SUMMARY

[0005] In view of the problems in the prior art, the present application provides a multi-element coupled epidemic transmission prediction method, which couples disease dynamics, information diffusion, policy game and climate through a common state space, thereby significantly improving the prediction accuracy and providing quantitative intervention suggestions.

[0006] The present application is realized by the following technical solutions: A multi-element coupled epidemic transmission prediction method, comprising the following steps: Step 1, based on the extended Kalman filter algorithm according to the optimal state estimation value of the last time predict the state prior estimate value at the current time, correct the state prior estimate value at the current time according to the error of the state estimate value at the current time and the real observation, obtain the optimal state estimate value at the current time and the policy intervention intensity; Step 2, aggregate the risk score of each text of the social public opinion data to obtain the social information pressure index, and use the UA-WA-FA three-state model combined with the total infection data at the last time to obtain the group vigilance degree; determine the climate adjustment factor according to the current meteorological data; Step 3, construct a dynamic adjustment factor according to the current group vigilance degree, climate adjustment factor, policy intervention intensity and external input risk, determine the dynamic transmission rate according to the dynamic adjustment factor and the current optimal estimate value, and calculate the state evolution sequence at the next time by substituting the dynamic transmission rate into the SEIR differential equation set; Step 4, construct a comprehensive decision loss function according to the decision weight, repeat steps 1-3, and iteratively solve the state evolution sequence based on the multi-objective strategy optimization of game theory until the decision loss function is minimized, to obtain the epidemic state prediction curve at the future time and the current optimal policy intervention intensity.

[0007] Preferably, the method for obtaining the optimal state estimate value at the current time and the policy intervention intensity in step 1 is as follows: The extended Kalman filter algorithm according to the state optimal estimate value at t 1 time , and the state prior estimate value at t time is predicted combined with the dynamic equation; The latest daily new cases are used as the real observation data at t time, the error between the state prior estimate value at t time and the real observation data is calculated, the state prior estimate value at t time is corrected according to the current Kalman gain K(t) and the error, and the state optimal estimate value at t time and the policy intervention intensity are obtained.

[0008] Preferably, the expression of the Kalman gain is as follows:

[0009] Wherein, K(t) is the Kalman gain, is the covariance matrix of the prior estimate, H is the observation matrix, and R is the covariance matrix of the observation noise.

[0010] Preferably, the risk score of each text of the social public opinion data is aggregated to obtain the social information pressure index, and the UA-WA-FA three-state model is used combined with the total infection data at the last time to obtain the group vigilance degree, which includes: According to the negative emotion intensity of each text of social public opinion data and combined with the risk theme weight, the risk score of each text is obtained The risk scores of all texts are aggregated to obtain a single-day total stress value, and the single-day total stress value is normalized to obtain a social information stress index According to the social information stress index and combined with the total number of infected people at the previous time, a UA-WA-FA three-state model is used to simulate the evolution process of public awareness state under the stimulation of epidemic information and severity, and a group alertness is obtained.

[0011] Preferably, according to the negative emotion intensity of each text of social public opinion data and combined with the risk theme weight, the risk score of each text is obtained, the risk scores of all texts are aggregated to obtain a single-day total stress value, and the single-day total stress value is normalized to obtain a social information stress index, comprising: Using a mature sentiment analysis model to score each text to obtain its negative emotion intensity; According to the set risk theme dictionary, different weights are given to different types of keywords, and the risk score of each text is calculated according to the keyword weight; The negative emotion intensity and risk score of all texts are weighted and aggregated to obtain social information stress of the day, and the social information stress of the day is normalized to obtain a social information stress index.

[0012] Preferably, the climate adjustment factor is determined according to the current meteorological data, comprising: Obtain the temperature and relative humidity of the current environment and combine with the regression function to calculate the suitability of the current meteorological condition for virus transmission to obtain the climate adjustment factor.

[0013] Preferably, the dynamic transmission rate is determined according to the dynamic adjustment factor and the current optimal estimate value, comprising: According to the condition function, the dynamic adjustment factor and the current policy intervention intensity, the dynamic transmission rate at the current time is determined As follows:

[0014] Wherein, The condition function is The policy intervention intensity is The climate adjustment factor is.

[0015] Preferably, the dynamic transmission rate and the SEIR differential equation set are substituted to calculate the state evolution sequence at the next time, comprising: The dynamic transmission rate is substituted into the SEIR differential equation set ​And substitute into the SEIR differential equations, the state evolution sequence of next time is calculated through the Runge-Kutta method.

[0016] A multi-element coupled epidemic transmission prediction system comprises: The state prediction module is configured to determine a state prior estimate value of a current time based on an extended Kalman filtering algorithm according to an optimal state estimation value of a previous time The state prior estimate value of the current time is corrected according to an error between the state estimation value of the current time and a real observation to obtain an optimal state estimation value of the current time and a policy intervention intensity; The dynamic factor module is configured to aggregate a risk score of each text of social public opinion data to obtain a social information pressure index, and obtain a group alertness degree by using a UA-WA-FA three-state model in combination with total infection data of a previous time; A climate adjustment factor is determined according to current meteorological data; The state evolution module is configured to construct a dynamic adjustment factor according to the current group alertness degree, the climate adjustment factor, the policy intervention intensity and external input risk, determine a dynamic transmission rate according to the dynamic adjustment factor and the current optimal estimation value, and calculate the state evolution sequence of next time by substituting the dynamic transmission rate into the SEIR differential equations; The intervention module is configured to construct a comprehensive decision loss function according to a decision weight, repeat steps 1-3, and iteratively solve the state evolution sequence based on a multi-objective strategy optimization of game theory until the decision loss function is minimized to obtain an epidemic state prediction curve of a future time and a current optimal policy intervention intensity.

[0017] An electronic device comprises: A memory is configured to store a computer program; A processor is configured to implement the steps of the multi-element coupled epidemic transmission prediction method when the computer program is executed.

[0018] Compared with the prior art, the present application has the following beneficial technical effects: The multi-element coupled epidemic transmission prediction method provided by the application has obvious technical scheme advantages. The core is to construct a multi-dimensional dynamic coupling framework to realize the dual improvement of prediction accuracy and decision support value. First, by extending the Kalman filter algorithm, the optimal state estimation value at the last moment is used to predict the current prior estimation value, and the real observation error is corrected, which can calibrate the state parameters and policy intervention intensity in real time, solve the problem that the static parameters of the traditional model cannot adapt to the dynamic changes of the epidemic, and ensure the accuracy of the basic data and intervention intensity. Secondly, the social public opinion and meteorological data are innovatively integrated. The social information pressure index is generated by aggregating the text risk score, and the group alertness is output by combining the UA-WA-FA three-state model, and the climate adjustment factor is determined according to the meteorological data, which breaks through the limitation of single factor analysis and fully captures the influence of human behavior and natural environment on epidemic transmission. Finally, based on the multi-dimensional factors, the dynamic adjustment factor and the transmission rate are constructed, the state evolution sequence is calculated by integrating the SEIR differential equation set, and the decision loss function is minimized by game theory multi-objective optimization iteration, which realizes accurate prediction of the epidemic state and outputs the optimal policy intervention intensity, and provides practical and forward-looking technical support for scientific prevention and control in complex epidemic scenarios.

[0019] The application also provides a multi-element coupled epidemic transmission prediction system, an electronic device and a computer storage medium, which have all the advantages of the multi-element coupled epidemic transmission prediction method. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0021] Figure 1 The flowchart of the multi-element coupled epidemic transmission prediction method of the application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the embodiments of the application more clear, the technical scheme in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, not all the embodiments. The components of the embodiments of the application described and shown in the drawings here can be arranged and designed in various different configurations.

[0023] Therefore, the following detailed description of embodiments of the application provided in the accompanying drawings is not intended to limit the scope of the application claimed, but merely represents selected embodiments of the application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0024] Referring to Figure 1 A multi-source heterogeneous data coupling epidemic transmission prediction method, comprising the following steps: Step 1, obtaining multi-source heterogeneous data and preprocessing.

[0025] The multi-source heterogeneous data includes public health data, meteorological environmental data, social public opinion data, population statistics data and economic data.

[0026] The multi-source heterogeneous data includes structured data and unstructured data. The structured data is public health data, traffic flow data, meteorological environmental data and population statistics data, and the unstructured data includes social public opinion data.

[0027] The public health data refers to the daily new / cumulative cases in each region .

[0028] The meteorological environmental data refers to the daily temperature T and relative humidity RH in each region.

[0029] The social public opinion data refers to the related text M(t) on social media, news and other platforms.

[0030] The population statistics data refers to the total population N in each region, and the initial behavior state distribution of the total population N is: (proportion of unaware, alert, and tired population).

[0031] Then, the ETL method is used to preprocess the multi-source heterogeneous data, which includes cleaning, parsing, missing value filling and alignment, so that all the multi-source heterogeneous data are uniformly mapped into the four-tuple standard format of {timestamp, location code, variable name, numerical value}, forming a standardized time series data set.

[0032] Step 2, the extended Kalman filter algorithm is used to predict the state prior estimate value at the current time according to the optimal state estimate value at the last time , correct the state prior estimate value at the current time according to the error of the state estimate value at the current time and the real observation, and obtain the optimal state estimate value at the current time and the policy intervention intensity, as follows: S2.1, the optimal state estimate value at time t 1 output by the extended Kalman filter algorithm ; The optimal state estimation value includes susceptible S, exposed E, infected I and recovered R.

[0033] S2.2, the extended Kalman filter algorithm is used to obtain the optimal state estimation value at time t 1 according to the state estimation value at time t 0 and the state prior estimation value at time t predicted by combining the dynamic equation.

[0034] wherein, is the optimal state estimation value at time t 1, f(...) is the state transition function, that is, the dynamic equation described by the differential equation group, is the state prior estimation value at time t.

[0035] S2.2, the latest daily new cases obtained in step 1 are taken as the real observation data at time t , the error between the state prior estimation value at time t and the real observation data is calculated, the state prior estimation value at time t is corrected according to the current Kalman gain K(t) and the error, and the optimal state estimation value at time t and the updated model parameters are obtained.

[0036] The expression of the Kalman gain is as follows:

[0037] wherein, K(t) is the Kalman gain, the value of the Kalman gain determines the degree to which the “new observation data” should be believed, the greater the gain, the more the true observation value is believed; the smaller the gain, the more the model prediction value is believed; is the covariance matrix of the prior estimation, which represents the uncertainty degree of the model to its own prediction; H is the observation matrix, which maps the high-dimensional state vector X to the directly observable quantity (the number of infected persons I in this case).

[0038] R is the covariance matrix of the observation noise, which represents the uncertainty or error of the observation data itself.

[0039] The updated model parameters include the basic transmission rate , the latent-to-infected rate , and the recovery rate . Policy intervention strength: such as policy sensitivity , behavior response strength .

[0040] According to the extended Kalman filter algorithm, the optimal estimation value at time t is predicted, and the core state variables (S, E, I, R) of the model are updated, and a series of key parameters of the model are also updated and optimized. The updated state estimation value : contains more accurate estimates of the current S, E, I, R, etc. cabin population.

[0041] Step 3, aggregate the risk score of each text of social public opinion data to get social information pressure index, and use UA-WA-FA three-state model combined with the total infection data of the last moment to get the group alertness.

[0042] According to the current meteorological data, determine the climate adjustment factor.

[0043] S3.1, according to the negative emotion intensity of each text of social public opinion data and combined with the risk theme weight, get the risk score of each text , aggregate the risk score of all texts to get the single-day total pressure value, and normalize the single-day total pressure value to get the social information pressure index ; According to the social information pressure index and combined with the total number of infections of the last moment, use the UA-WA-FA three-state model to simulate the evolution process of the public's awareness state under the stimulation of epidemic information and severity, and get the group alertness; the specific method is as follows: 1) In the social public opinion data, select each text highly related to the epidemic, and preprocess each text, including removing HTML tags, emoticons, URL links, etc. noise.

[0044] 2) Use a pre-defined keyword library to filter out posts or news containing these words, and perform Chinese word segmentation on the filtered texts to get each text highly related to the epidemic.

[0045] 3) Use a mature sentiment analysis model (such as BERT-based model) to score each text to get its negative emotion intensity (taking value 0-1, the larger the value represents the more negative).

[0046] 4) Define a risk theme dictionary, assign different weights to different types of keywords. For example, the weight of high-risk words will be higher than that of regular words. Calculate the risk score of each text .

[0047] 5) Weighted sum of the negative emotion intensity and risk score of all related texts on the same day to get the information pressure of the whole society on the same day, the aggregation formula is as follows:

[0048] Where n is the total number of relevant texts for that day. , It is the weighting coefficient of emotion and risk.

[0049] Normalization: To ensure stable input into the model, the original stress values ​​need to be normalized (e.g., using Min-Max or Z-score methods) to obtain the final social information stress index. This ensures that its value falls within [0,1] or a stable interval.

[0050] 6) Based on the social information stress index and the total number of infections at the previous moment, the UA-WA-FA three-state model (Unaware-Warned-Fatigued) is used to simulate the evolution of the public's awareness of protection under the stimulation of epidemic information and severity, and to obtain the group alertness.

[0051] The group alertness level A(t) represents the proportion of the population in a society that is in an alert state (WA), as shown below:

[0052] Among them, U, W, and F represent the number of people in an unaware, alert, and fatigued state, respectively.

[0053] The rate of change in the number of unaware individuals (U) was determined based on group alertness. ;

[0054] Where λ is the "infection rate" of information dissemination, which refers to the rate at which unsuspecting individuals become aware of the epidemic after being exposed to epidemic information (default value 0.8). ρ is the "forgetting rate", which refers to the rate at which fatigued individuals become unsuspecting again after relaxing their vigilance (default value 0.05).

[0055] Determine the rate of change in the number of alert people (W) based on the level of group alertness. ;

[0056] Where μ is the "fatigue rate", which refers to the rate at which an alert person becomes fatigued due to prolonged vigilance (default value 0.03).

[0057] The rate of change in the number of fatigued individuals (F) was determined based on group alertness. ;

[0058] S3.2, Obtain the temperature and relative humidity of the current environment and combine the regression function to calculate the suitability of the current weather conditions for virus transmission, and obtain the climate adjustment factor, which is a value between 0 and 1.

[0059] wherein, is the temperature and humidity most suitable for virus transmission ( default 8, default 50). is the sensitivity coefficient of temperature and humidity on the transmission rate ( default 0.01, default 0.02).

[0060] Step 4, According to the current group alertness, climate adjustment factor and combined external input risk to build dynamic adjustment factor, according to the dynamic adjustment factor and the current policy intervention intensity to determine the dynamic transmission rate.

[0061] S4.1, Policy intervention intensity is the current policy intervention intensity P(t), policy intervention intensity in the first round of the default value is set to 0.1, continuously updated in the iteration process.

[0062] External input risk is the rate of i-type virus infection imported from other areas by means of transportation, the calculation formula is .

[0063] S4.2, Set the adjustment function , the adjustment function it combines the inhibitory effect of public behavior and policy intervention intensity:

[0064] S4.3, According to the dynamic adjustment factor and the current optimal estimate value to determine the dynamic transmission rate at the current time.

[0065]

[0066] S4.4, The dynamic transmission rate and into the SEIR differential equation set, through numerical integration (such as Runge-Kutta method) to calculate the next time state evolution sequence, which represents the complete sequence of the number of people in each compartment over time in the future (such as the next 30 days).

[0067] The expression of the SEIR differential equation set is as follows:

[0068] wherein, is the change rate of susceptible (Susceptible) number, the rate of susceptible individuals being infected by the exposed individuals of type i virus, the rate of susceptible individuals being immunized, default value is 0.001.

[0069]

[0070] where, is the rate of change of the number of exposed individuals of type i virus, is the rate of exposed individuals developing into infected individuals.

[0071]

[0072] where, is the rate of change of the number of infected individuals of type i virus, is the rate of infected individuals recovering or dying.

[0073] is the rate of infected individuals of type i virus being imported from other areas by transportation, calculated by ; is the policy sensitivity, is the policy intervention intensity is the number of observed cases divided by the normalized intensity.

[0074]

[0075] is the rate of change of the number of individuals recovering from being infected with a specific virus strain i.

[0076] Substitute the dynamic transmission rate into the SEIR differential equation set, and calculate the state evolution sequence at the next time by methods such as the Runge-Kutta method, which represents the complete sequence of the number of individuals in each compartment changing over time in a future period of time (such as the next 30 days), expressed as follows:

[0077] Step 5, construct a comprehensive decision-making loss function according to the decision weight, repeat steps 2-4, and iteratively solve the state evolution sequence based on the multi-objective strategy optimization of game theory. The Adam optimizer calculates the gradient and updates the candidate policy intervention intensity P(t) according to the loss value L(t), so that it moves in the direction that makes the loss value L(t) smaller, until the decision-making loss function is minimized, obtaining the epidemic state prediction curve at the future time and the current optimal policy intervention intensity.

[0078] The decision weight includes the economic weight preset by the decision maker , health weight and fatigue weight .

[0079] The comprehensive decision loss function is constructed according to the decision weight as follows:

[0080] where, =GDP_loss_rate P(t) is the economic cost brought by the policy intervention intensity P(t). = is the health cost caused by the epidemic (measured by the proportion of infection). =1 A(t) is the public's fatigue cost of epidemic prevention (the lower the alertness, the higher the fatigue).

[0081] Adam optimizer update rule:

[0082] where, is the policy intervention intensity of the current iteration. , : are the first-order (momentum) and second-order (velocity) gradient matrix estimates, respectively, which are the core of the Adam algorithm. is the learning rate, which controls the step size of each update, and Clip[]: limits the updated value to the interval $[0, 1]$ to ensure that the policy intervention intensity is meaningful.

[0083] Step 6, according to the epidemic state prediction curve and the optimal policy intervention intensity to intervene in the epidemic.

[0084] The embodiment of the application is a multi-element coupled epidemic transmission prediction method. The multi-element coupled epidemic transmission prediction method takes a unified coupling framework as a core architecture, realizes the standardized integration of disease, information, policy, and climate data through multi-source heterogeneous data preprocessing, and then relies on modules such as an extended Kalman filter, a UA-WA-FA model, and SEIR differential equations to complete the linkage calculation of multiple subsystems in a single state space, thereby completely solving the problem of the fragmentation of traditional model modules and the inability to work together. At the same time, the design of the dynamic transmission rate in the embodiment provides an adaptive basis for the multi-strain competition mechanism, and subsequent immune escape and cross-protection parameters can be directly integrated without significantly adjusting the core framework, which has excellent scalability. In terms of UA-WA-FA behavior closed-loop modulation β, the group alertness generated by social public opinion and the intensity of policy intervention are jointly included in the behavior adjustment function, which directly acts on the dynamic transmission rate calculation to form a real-time feedback closed loop of information transmission influencing epidemic transmission, thereby avoiding the disconnection between behavior analysis and transmission model. In addition, based on the iterative solving process of the Adam optimizer, the “three-target policy game optimization” is accurately landed, the economic, health, and fatigue costs are balanced through the comprehensive decision loss function, and the optimal policy intervention intensity is output online, which not only ensures the accuracy of the epidemic prediction, but also provides a scientific basis for the prevention and control decision.

[0085] Correspondingly, the application also provides a multi-element coupled epidemic transmission prediction system, comprising: a state prediction module configured to determine, based on an extended Kalman filter algorithm, an optimal state estimation value at a previous moment predict a state prior estimation value at a current moment, correct the state prior estimation value at the current moment according to an error between a state estimation value at the current moment and a real observation, and obtain a current optimal state estimation value and a policy intervention intensity; a dynamic factor module configured to aggregate a risk score of each text of social public opinion data to obtain a social information pressure index, and obtain a group alertness by using a UA-WA-FA three-state model in combination with total infection data at a previous moment; determine a climate adjustment factor according to current meteorological data; a state evolution module configured to construct a dynamic adjustment factor according to the current group alertness, the climate adjustment factor, the policy intervention intensity, and external input risks, determine a dynamic transmission rate according to the dynamic adjustment factor and the current optimal estimation value, and calculate a state evolution sequence at a next moment by substituting the dynamic transmission rate into SEIR differential equations; an intervention module configured to construct a comprehensive decision loss function according to a decision weight, repeat steps 1-3, and iteratively solve the state evolution sequence based on a multi-objective strategy optimization of game theory until the decision loss function is minimized, so as to obtain an epidemic state prediction curve at a future moment and a current optimal policy intervention intensity.

[0086] It should be noted that in several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. For example, the apparatus embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules can be combined or integrated into another apparatus, or some features can be ignored or not executed. The modules described as separate components can be or can not be physically separated, and the components displayed as modules can be one physical unit or multiple physical units, that is, can be located in one place or can be distributed to multiple different places. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0087] In addition, each module in each embodiment of the present application can be integrated in one processing unit, or each module can be physically present separately, or two or more modules can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0088] The electronic device provided in the embodiment of the present application includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the multi-coupling epidemic propagation prediction method described in any of the above embodiments when executing the computer program.

[0089] The electronic device provided in another embodiment of the present application can further include: an input port connected with the processor, used for transmitting the multi-modal data collected by the external collection device to the processor; and a display unit connected with the processor, used for displaying the processing result of the processor to the outside world; and a communication module connected with the processor, used for realizing the communication between the electronic device and the outside world. The display unit can be a display panel, a laser scanning display, etc.; the communication mode adopted by the communication module includes but is not limited to mobile high-definition link technology (HML), universal serial bus (USB), high-definition multimedia interface (HDMI), wireless connection (including wireless fidelity technology (WiFi), Bluetooth communication technology, low-power Bluetooth communication technology, IEEE 802.11s-based communication technology).

[0090] The computer readable storage medium provided in the embodiment of the present application stores a computer program, and the computer program is executed by the processor to realize the steps of the multi-coupling epidemic propagation prediction method described in any of the above embodiments.

[0091] The related parts in the multi-element coupling epidemic spread prediction system, the electronic device and the computer readable storage medium provided by the embodiments of the present application are described in detail in the corresponding part of the multi-element coupling epidemic spread prediction method provided by the embodiments of the present application, and will not be described here. In addition, the part of the above technical solution provided by the embodiments of the present application which is consistent with the implementation principle of the corresponding technical solution in the prior art is not described in detail, so as not to be too redundant.

[0092] The above is only an illustration of the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made on the basis of the technical solution according to the technical idea of the present application falls within the protection scope of the claims of the present application.

Claims

1. A multi-coupling method for predicting the spread of an epidemic, characterized in that, Includes the following steps: Step 1: Based on the extended Kalman filter algorithm, estimate the optimal state value from the previous time step. Predict the prior state estimate at the current moment, and correct the prior state estimate at the current moment based on the error between the current state estimate and the actual observation, to obtain the optimal state estimate at the current moment and the intensity of policy intervention; Step 2: Aggregate the risk scores of each text in the social sentiment data to obtain the social information pressure index, and combine it with the total infection data of the previous moment to obtain the group vigilance level using the UA-WA-FA three-state model; Determine climate regulation factors based on current meteorological data; Step 3: Construct a dynamic adjustment factor based on the current group alertness, climate adjustment factor, policy intervention intensity, and external input risk. Determine the dynamic propagation rate based on the dynamic adjustment factor and the current optimal estimate. Substitute the dynamic propagation rate into the SEIR differential equation system to calculate the state evolution sequence at the next moment. Step 4: Construct a comprehensive decision loss function based on the decision weights. Repeat steps 1-3 and iteratively solve the state evolution sequence based on game theory multi-objective strategy optimization until the decision loss function is minimized. This will yield the predicted curve of the epidemic status at future moments and the current optimal policy intervention intensity.

2. The multi-coupling method for predicting the spread of an epidemic according to claim 1, characterized in that, The method for obtaining the optimal state estimate and the intensity of policy intervention at the current moment, as described in step 1, is as follows: The extended Kalman filter algorithm is based on the optimal state estimate at time t-1. And combine the dynamic equations to predict the prior state estimate at time t; The latest daily new cases are used as the actual observation data at time t. The error between the state prior estimate at time t and the actual observation data is calculated. The state prior estimate at time t is corrected according to the current Kalman gain K(t) and the error to obtain the optimal state estimate at time t and the intensity of policy intervention.

3. The multi-coupling method for predicting the spread of an epidemic according to claim 2, characterized in that, The expression for the Kalman gain is as follows: Where K(t) is the Kalman gain, Let H be the covariance matrix of the test estimate, H be the observation matrix, and R be the covariance matrix of the observation noise.

4. The multi-coupling method for predicting the spread of an epidemic according to claim 1, characterized in that, The social information pressure index is obtained by aggregating the risk scores of each text in the social sentiment data, and then using the UA-WA-FA three-state model combined with the total infection data from the previous time point to obtain the group alertness, including: Based on the intensity of negative sentiment in each text from social media data and combined with the weight of risk topics, a risk score is obtained for each text. The risk scores of all texts are aggregated to obtain the total daily stress value. The social information stress index is then obtained by normalizing the total daily stress value. ; Based on the social information stress index and the total number of infections at the previous moment, the UA-WA-FA three-state model is used to simulate the evolution of public awareness of protection under the stimulation of epidemic information and severity, and to obtain the group vigilance level.

5. The multi-coupling method for predicting the spread of an epidemic according to claim 4, characterized in that, The risk score for each text is obtained by combining the negative sentiment intensity of each text in the social sentiment data with the risk topic weight. The risk scores of all texts are aggregated to obtain the total daily stress value. The social information stress index is obtained by normalizing the total daily stress value, including: Each text is scored using a mature sentiment analysis model to obtain its negative sentiment intensity; Based on the established risk topic dictionary, different weights are assigned to different types of keywords, and the risk score of each text is calculated based on the keyword weights. We weight and aggregate the negative emotion intensity and risk score of all texts to obtain the social information pressure of the day. We then normalize the social information pressure of the day to obtain the social information pressure index.

6. The multi-coupling method for predicting the spread of an epidemic according to claim 1, characterized in that, The determination of climate adjustment factors based on current meteorological data includes: By obtaining the current ambient temperature and relative humidity and combining them with a regression function, the suitability of the current meteorological conditions for virus transmission is calculated, and the climate regulation factor is obtained.

7. The multi-coupling method for predicting the spread of an epidemic according to claim 1, characterized in that, The determination of the dynamic propagation rate based on the dynamic adjustment factor and the current optimal estimate includes: The dynamic propagation rate at the current moment is determined based on the set condition function, dynamic adjustment factor, and current policy intervention intensity. ,as follows: in, For conditional functions, As for the intensity of policy intervention, It is a climate regulating factor.

8. The multi-coupling method for predicting the spread of an epidemic according to claim 1, characterized in that, The step of calculating the state evolution sequence at the next time step by substituting the dynamic propagation rate into the SEIR differential equation system includes: Dynamic propagation rate Substituting these equations into the SEIR differential equations, the state evolution sequence at the next time step is calculated using the Runge-Kutta method.

9. A multi-coupling epidemic transmission prediction system, characterized in that, include: The state prediction module is used to predict the state based on the optimal state estimate from the previous time step using the extended Kalman filter algorithm. Predict the prior state estimate at the current moment, and correct the prior state estimate at the current moment based on the error between the current state estimate and the actual observation, to obtain the optimal state estimate at the current moment and the intensity of policy intervention; The dynamic factor module is used to aggregate the risk score of each text in the social sentiment data to obtain the social information pressure index, and combine it with the total infection data of the previous moment to obtain the group vigilance level using the UA-WA-FA three-state model; Determine climate regulation factors based on current meteorological data; The state evolution module is used to construct dynamic adjustment factors based on the current group alertness, climate adjustment factors, policy intervention intensity, and external input risks. It determines the dynamic propagation rate based on the dynamic adjustment factors and the current optimal estimate, and calculates the state evolution sequence for the next moment by substituting the dynamic propagation rate into the SEIR differential equation system. The intervention module is used to construct a comprehensive decision loss function based on decision weights. Steps 1-3 are repeated to iteratively solve the state evolution sequence based on game theory multi-objective strategy optimization until the decision loss function is minimized, thereby obtaining the predicted curve of the epidemic status at future moments and the current optimal policy intervention intensity.

10. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the multi-coupling epidemic transmission prediction method as described in any one of claims 1-8 when executing the computer program.

Images