Infectious disease prediction mixed data assimilation method based on time-varying set Kalman filtering and KNN classifier

Abstract

The invention relates to an infectious disease prediction mixed data assimilation method based on time-varying set Kalman filtering and a KNN classifier, and belongs to the field of data processing, and the method comprises the steps: constructing a time-varying parameter model, and obtaining a weight function carrying specific observation information through Newton interpolation; a KNN classifier is introduced to classify the set parameters of the random updating step of the set Kalman filtering based on the weighting function; carrying out resampling processing on the classified new set to obtain an updated set; data assimilation is carried out on data by using the constructed mixed data assimilation model, model parameters are adjusted in real time in the operation process to adapt to new observation data, and it is ensured that the model can reflect the actual propagation situation; a simulation result is analyzed and compared with independent observation data for verification, the accuracy and reliability of the mixed data assimilation model are evaluated, and more accurate prediction data are obtained.

Description

Technical Field

[0003] The present invention relates to the field of data processing, and in particular to an infectious disease prediction method based on data assimilation technology. Background Art

[0005] In the current field of infectious disease prediction, traditional prediction models usually rely on limited observational data and simplified mathematical frameworks, which makes it difficult for these methods to capture complex transmission pathways. In addition, due to the high dynamics and complexity of transmission pathways, traditional models have the problem of ignoring the specificity of virus transmission and amplifying its integrity when processing data, making it difficult to accurately predict the spread of infectious diseases. Therefore, effectively integrating infectious disease observation data and assimilating it into complex prediction models to improve the prediction accuracy and adaptability of the models has become a major challenge in research. Traditional prediction models are not flexible enough, and there is a significant deviation between the observed values ​​and the model predicted values. Therefore, there is an urgent need for a new data assimilation method that can integrate multi-source observational data and correct the prediction accuracy of infectious disease models in real time, thereby improving the accuracy and flexibility of prediction models in infectious disease prediction. Summary of the invention

[0007] Based on the above objectives, the present invention provides a hybrid data assimilation method for infectious disease prediction based on time-varying ensemble Kalman filtering and KNN classifier.

[0008] The present invention proposes an infectious disease prediction method based on hybrid data assimilation, comprising:

[0009] S1: Based on the existing infectious disease model, initialize the infectious disease prediction model to simulate the infectious disease transmission process;

[0010] S2: Collect multi-source infectious disease observation data;

[0011] S3: Construct a time-varying parameter model and use Newton interpolation to obtain a weight function that carries specific observation information;

[0012] S4: Introduce a KNN classifier to classify the ensemble parameters of the random update step of the ensemble Kalman filter based on the above weight function;

[0013] S5: resampling the new set after the above classification to obtain an updated set;

[0014] S6: Use the constructed hybrid data assimilation model to assimilate the data, and adjust the model parameters in real time during operation to adapt to the new observation data to ensure that the model can reflect the actual propagation situation;

[0015] S7: Analysis and verification of simulation results, analyze the simulation results, compare and verify them with independent observational data, and evaluate the accuracy and reliability of the data assimilation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of an overall process based on hybrid data assimilation provided by the present invention.

[0018] Figure 2 A schematic diagram of the classification process provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the time-varying ensemble Kalman filter provided by the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work should fall within the scope of protection of the present invention.

[0022] Furthermore, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts disclosed in the present invention.

[0023] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. The terms "installed", "connected", and "connected" 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 a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. 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.

[0024] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of methods and systems consistent with some aspects of the present invention as detailed in the appended claims.

[0025] The present invention proposes a hybrid data assimilation method for infectious disease prediction based on time-varying ensemble Kalman filtering and KNN classifier, in order to solve the problems that traditional prediction models usually rely on limited observation data and simplified mathematical frameworks in the prediction process, which makes it difficult for prediction models to capture the complex transmission pathways of viruses, and due to the high dynamics and complexity of the transmission pathways, traditional models ignore the specificity of virus transmission when processing data and amplify its integrity, making it difficult to accurately predict the transmission of infectious diseases. Specifically, the method includes steps S1 to S7:

[0026] S1, based on the existing infectious disease model, initialize the infectious disease prediction model to simulate the infectious disease transmission process.

[0027] Optionally, the infectious disease model selected by the present invention is specifically the SEAIQR model. The SEAIQR model is an extended model for infectious disease transmission modeling, which adds the exposure period (E), asymptomatic infected persons (A), isolation (I), and hospitalized patients (Q) to the classic SIR (susceptible-infected-recovered) model to describe and simulate the transmission process of infectious diseases in more detail.

[0028] S2: Collect multi-source infectious disease observation data.

[0029] Optionally, the infectious disease observation data specifically include the initial number of susceptible persons at the municipal level, the initial number of isolated susceptible persons, the initial number of exposed persons, the initial number of isolated exposed persons, the initial number of asymptomatic infected persons, the initial number of infected persons, the initial number of hospitalized persons, the initial number of immune persons, the infection rate, the exponential decay rate, the transmission coefficient of exposed persons, the transmission coefficient of asymptomatic infected persons, the isolation rate, the conversion rate from latent period to infected persons, the ratio of asymptomatic infected persons to infected persons, the virus mortality rate, the isolation release rate, the isolation rate of infected persons, the conversion rate of isolated latent persons to hospitalized patients, the recovery rate of infected persons, the recovery rate of asymptomatic infected persons, the recovery rate of hospitalized persons, the immunity threshold, and the decrease in antibody levels.

[0030] S3: Construct a time-varying parameter model and use Newton interpolation to obtain a weight function that carries specific observation information.

[0031] Among them, the time-varying parameter model can be constructed by considering the main influencing parameters, which can effectively respond to the changing trend of the prediction vector.

[0032] Specifically, a suitable interpolation method may be selected to construct a weight function carrying specific observation information. In a possible implementation, S3 specifically includes sub-steps S301, S302, and S303:

[0033] S301: construct a time-varying parameter model;

[0034] S302: Determine a weight function that carries specific observation information.

[0035] S303: Construct a weight model of different parameters in the random update step of the ensemble Kalman filter.

[0036] According to the potential relationship between different features in the parameter set, a time-varying parameter model is constructed:

[0037]

[0038] in, Indicates the observation period The time-varying parameters of the day, Indicates the observation period The number of infections per day, Indicates the observation period The number of potentially infected people per day, Indicates Number of people in quarantine per day, Indicates the number of days in the observation period.

[0039] By fitting the time-varying parameters Get the weight function that carries specific observation information:

[0040]

[0041] in, Indicates the number of days in the observation period.

[0042] In one possible implementation, a weight model for randomly updating parameters is constructed. :

[0043]

[0044] in Indicates updating the Tiandi The predicted value of the number of infected people in the vector, Indicates updating the Tiandi The predicted value of the number of potential infected persons in the vector, Indicates updating the Tiandi The predicted value of the number of people in quarantine is a vector, Indicates updating the Tiandi The predicted isolation rate of the vector, In the present invention, the specific observation information of different virus transmissions may be different. By comparing the transmission rates of different viruses, the importance of each parameter to the specific observation information can be more reasonably considered to calibrate the weight model. .

[0045] S4: A KNN classifier is introduced to classify the ensemble parameters of the random update step of the ensemble Kalman filter based on the above weight function.

[0046] Among them, K-Nearest Neighbors (KNN) is a commonly used supervised learning algorithm, mainly used for classification and regression tasks. Its core idea is to predict the category or value of new data points by comparing the distance between the new data points and the data points in the training data set.

[0047] S4 specifically includes sub-steps S401 and S402:

[0048] S401: Construct classification indicators ;

[0049] S402: Determine the classification scope.

[0050] Among them, the weight function that carries specific observation information Get classification index :

[0051]

[0052] in is the initial time-varying parameter of the observation period.

[0053] Among them, through classification indicators Get the classification range:

[0054]

[0055] in is a classification indicator, represents the number of days of the observation period, represents the variance of classification interval 1, represents the variance of classification interval 2, and represents the classification range of classification interval 1, and Represents the classification range of classification interval 2, and The present invention uses the above classification index to classify the target set, where , , and is an adjustable classification index. In the present invention, the classification index , , and It can be adjusted according to the different sources of infection to obtain a new set with higher fit .

[0056] S5: Resample the new set after the above classification to obtain an updated set.

[0057] S6: Use the constructed hybrid data assimilation model to assimilate the data, and adjust the model parameters in real time during operation to adapt to the new observation data to ensure that the model can reflect the actual propagation situation.

[0058] Among them, the time-varying ensemble Kalman filter prediction error covariance matrix and analysis value are obtained by hybrid data assimilation method:

[0059]

[0060] in, For the The prediction error covariance matrix of the time-varying ensemble Kalman filter obtained by the update is: is the number of set vectors of the target set obtained after classification, for The model prediction value obtained by the update is For the The average value of the updated analysis value of the time-varying ensemble Kalman filter obtained by the update.

[0061] S7: Analysis and verification of simulation results, analyze the simulation results, compare and verify them with independent observational data, and evaluate the accuracy and reliability of the data assimilation model.

[0062] The beneficial effects of the present invention are as follows:

[0063] The KNN classifier is used to classify the data, and the time-varying parameters and resampling methods are used to integrate the data, which improves the prediction accuracy of the model. The hybrid data assimilation method is used to adjust and modify the infectious disease prediction model, which improves the universality and prediction stability of the traditional infectious disease model and ensures better consistency between the assimilation process and the infectious disease prediction model.

Claims

1. An infectious disease prediction method based on hybrid data assimilation is characterized by: include: S1: Based on the existing infectious disease model, initialize the infectious disease prediction model to simulate the infectious disease transmission process; S2: Collect multi-source infectious disease observation data; S3: Construct a time-varying parameter model and use Newton interpolation to obtain a weight function that carries specific observation information; S4: Introduce a KNN classifier to classify the ensemble parameters of the random update step of the ensemble Kalman filter based on the above weight function; S5: resampling the new set after the above classification to obtain an updated set; S6: Use the constructed hybrid data assimilation model to assimilate the data, and adjust the model parameters in real time during operation to adapt to the new observation data to ensure that the model can reflect the actual propagation situation; S7: Analysis and verification of simulation results, analyze the simulation results, compare and verify them with independent observational data, and evaluate the accuracy and reliability of the data assimilation model.

2. The infectious disease prediction method based on hybrid data assimilation according to claim 1, characterized in that: The S3 specifically includes: S301: construct a time-varying parameter model; S302: Determine a weight function that carries specific observation information. S303: Construct a weight model of different parameters in the random update step of the ensemble Kalman filter.

3. The infectious disease prediction method based on hybrid data assimilation according to claim 2, characterized in that: The S301 specifically includes: According to the potential relationship between different features in the parameter set, a time-varying parameter model is constructed: in, Indicates the observation period The time-varying parameters of the day, Indicates the observation period The number of infections per day, Indicates the observation period The number of potentially infected people per day, Indicates Number of people in quarantine per day, Indicates the number of days in the observation period.

4. The infectious disease prediction method based on hybrid data assimilation according to claim 2, characterized in that: The S302 specifically includes: By fitting the time-varying parameters Get the weight function that carries specific observation information: in, Indicates the number of days in the observation period.

5. The infectious disease prediction method based on hybrid data assimilation according to claim 2, characterized in that: The S303 specifically includes: Construct a weight model with random update parameters : in Indicates updating the Tiandi The predicted value of the number of infected people in the vector, Indicates updating the Tiandi The predicted value of the number of potential infected persons in the vector, Indicates updating the Tiandi The predicted value of the number of people in quarantine is a vector, Indicates updating the Tiandi The predicted isolation rate of the vector, Indicates the number of days in the observation period.

6. The infectious disease prediction method based on hybrid data assimilation according to claim 1, characterized in that: The S4 specifically includes: S401: Construct classification indicators ; S402: Determine the classification scope.

7. The infectious disease prediction method based on hybrid data assimilation according to claim 6, characterized in that: The S401 specifically includes: Through the weight function that carries specific observation information Get classification index : in is the initial time-varying parameter of the observation period.

8. The infectious disease prediction method based on hybrid data assimilation according to claim 6, characterized in that: The S402 specifically includes: By classification index Get the classification range: in is a classification indicator, represents the number of days of the observation period, represents the variance of classification interval 1, represents the variance of classification interval 2, and represents the classification range of classification interval 1, and Represents the classification range of classification interval 2, and Represents a classified collection.

9. The infectious disease prediction method based on hybrid data assimilation according to claim 1, characterized in that: The S6 specifically includes: The time-varying ensemble Kalman filter prediction error covariance matrix and analysis value are obtained by hybrid data assimilation method: in For the The prediction error covariance matrix of the time-varying ensemble Kalman filter obtained by the update is: is the number of set vectors of the target set obtained after classification, for The model prediction value obtained by the update is For the The average value of the updated analysis value of the time-varying ensemble Kalman filter obtained by the update.

Images