CN-122025201-A - Epidemic situation emergency response decision method, system, computer equipment and storage medium

CN122025201ACN 122025201 ACN122025201 ACN 122025201ACN-122025201-A

Abstract

The application discloses an epidemic emergency response decision method, a system, computer equipment and a storage medium. The epidemic emergency response decision method comprises the steps of fitting parameters of an infectious disease dynamics model according to an L-BFGS-B optimization algorithm to generate a Hessian matrix, detecting and eliminating abnormal data of the Hessian matrix according to the pathological data, sampling the Hessian matrix after eliminating the pathological data according to a self-adaptive MCMC algorithm, and inputting the sampled parameters in the Hessian matrix into the infectious disease dynamics model to generate a prediction curve beam. The method is used for realizing rapid output of epidemic emergency response decisions.

Inventors

YANG FENGLEI
GAO WEI
FENG HANG
JIAO CHENGYANG
KUANG RUIJIE

Assignees

联通数智医疗科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260311

Claims (10)

1. The epidemic emergency response decision method is characterized by comprising the following steps: fitting parameters of an infectious disease dynamics model according to an L-BFGS-B optimization algorithm to generate a Hessian matrix; Detecting the pathological data of the Hessian matrix and eliminating abnormal data; Sampling the Hessian matrix with the pathological data removed according to the self-adaptive MCMC algorithm, and inputting the sampled parameters in the Hessian matrix into the infectious disease dynamics model to generate a prediction curve beam.
2. The epidemic emergency response decision method of claim 1, further comprising the steps of: and carrying out risk assessment on the prediction curve bundles to obtain decision suggestions.
3. The epidemic emergency response decision method of claim 2, further comprising the steps of: and optimizing the decision proposal according to the control target under the constraint condition to obtain a final decision scheme.
4. The epidemic emergency response decision method according to claim 1, wherein one of an infectious disease dynamics model SI model, an SIR model, an SIRs model, an SEIR model is used.
5. The epidemic emergency response decision method according to claim 1, wherein the step of performing the detection of the pathological data to reject abnormal data comprises: detecting a condition number of the Hessian matrix to identify a disease state matrix; Singular value decomposition identification is carried out on the disease state matrix, and a colinear parameter pair is removed; constructing a Hessian matrix according to an adaptive regularization coefficient algorithm; calculating an inverse matrix of the Hessian matrix to obtain a parameter covariance matrix; extracting parameter standard errors from diagonal elements of the parameter covariance matrix, calculating a parameter confidence interval of the Hessian matrix according to the parameter standard errors, and deleting parameters which correspondingly fall outside the parameter confidence interval in the Hessian matrix.
6. The epidemic emergency response decision method of claim 5, wherein sampling according to an adaptive MCMC algorithm comprises: dynamically determining the sampling quantity of each type of parameter in the Hessian matrix after removing the pathological data based on the size of the parameter standard error; Based on the determined sampling quantity and a normal approximation principle, adopting a multivariate normal distribution algorithm to randomly sample the Hessian matrix after removing the pathological data.
7. The epidemic emergency response decision method according to claim 1, wherein when generating the prediction curve bundles, calculating the fractional number of each prediction curve outputted by the infectious disease dynamics model, and synthesizing the prediction curves with the fractional number within the 95% confidence interval into the prediction curve bundles.
8. The epidemic emergency response decision system is characterized by comprising an infectious disease dynamics module, a Hessian matrix generation module, an anomaly processing module and a self-adaptive MCMC algorithm sampling module; The Hessian matrix generation module is used for generating a Hessian matrix by fitting parameters of the infectious disease dynamics model according to an L-BFGS-B optimization algorithm; the anomaly processing module is used for detecting the pathological data of the Hessian matrix and eliminating the anomaly data; the self-adaptive MCMC algorithm sampling module is used for sampling the Hessian matrix after the pathological data is removed according to the self-adaptive MCMC algorithm; The infectious disease dynamics module is used for inputting the sampled parameters in the Hessian matrix into the infectious disease dynamics model to generate a predictive curve bundle.
9. A computer device comprising a memory, a processor, the memory provided with a computer program which, when executed by the processor, is adapted to carry out the epidemic emergency response decision method according to any one of claims 1 to 7.
10. A storage medium having stored thereon computer instructions for execution by a processor to implement the epidemic emergency response decision method of any one of claims 1-7.

Description

Epidemic situation emergency response decision method, system, computer equipment and storage medium Technical Field The application relates to the technical field of large model data analysis, in particular to an epidemic emergency response decision method, a system, computer equipment and a storage medium. Background Infectious disease dynamics models (such as SIR, SIRS, SEIR model) are widely used for epidemic prediction. In the infectious disease dynamics model, the uncertainty quantization of parameters is needed, and the uncertainty quantization of parameters in the prior art comprises the following steps: (1) The complete Bayesian MCMC method is based on the Fischer-Tropsch information matrix theory, and parameter posterior distribution inference is carried out through a Markov chain Monte Carlo method; (2) Approximate bayesian computational methods, using simulation experiments and cross-validation; (3) Neural network acceleration methods, using deep learning acceleration parameter inference. There are a number of problems with the prior art. The first problem is that the calculation time is too long, the emergency response aging requirement cannot be met, the single analysis time consumption in the complete Bayesian MCMC method is usually 3-10 minutes, the comparison time consumption is 30-100 minutes if 10 alternatives are adopted, the decision scheme evaluation cannot be completed within 2 hours due to the time consumption of data processing, the approximate Bayesian calculation single analysis is 5-15 minutes, cross verification and multiple iterations are needed, the time limit is exceeded, the neural network method is fast in reasoning, the new epidemic data are retrained for a plurality of hours, and the neural network method lacks generality and is not suitable for emergency response of sudden epidemic situations. The second problem is that professional statistics is required, the base use threshold is high, and staff in the base health sector usually have a corresponding professional background. The third problem is that the requirement on computing resources is high, the basic hardware condition is too common to realize corresponding computing resources, and the cloud computing service involves potential data safety hazards. The fourth problem is that the lack of an engineering automatic processing mechanism is very easy to report errors and exit when encountering singular or excessively large condition number data, so that non-professional users cannot independently use the data, and the flow from the data to decision is not consistent. The fifth problem is the lack of a complete decision-oriented system, independent parameter estimation, uncertainty quantification and intervention assessment, requiring researchers to manually combine the steps, and difficult to automatically generate risk assessment and decision advice. Disclosure of Invention In order to overcome one or more technical problems in the prior art, the application provides an epidemic situation emergency response decision method, a system, computer equipment and a storage medium, which can realize rapid output of an epidemic situation emergency response decision. The technical scheme of the application is as follows. In a first aspect, the present application provides a method for determining an emergency response of an epidemic situation, comprising the steps of: fitting parameters of an infectious disease dynamics model according to an L-BFGS-B optimization algorithm to generate a Hessian matrix; Detecting the pathological data of the Hessian matrix and eliminating abnormal data; Sampling the Hessian matrix with the pathological data removed according to the self-adaptive MCMC algorithm, and inputting the sampled parameters in the Hessian matrix into the infectious disease dynamics model to generate a prediction curve beam. A first preferred embodiment of the first aspect further comprises the steps of: and carrying out risk assessment on the prediction curve bundles to obtain decision suggestions. In a first preferred embodiment of the first aspect, more preferably, the method further comprises the step of: and optimizing the decision proposal according to the control target under the constraint condition to obtain a final decision scheme. A second preferred embodiment of the first aspect is one of the infectious disease dynamics model SI model, SIR model, SIRs model, SEIR model used. In a third preferred embodiment of the first aspect, the step of performing the detection of the pathological data and the elimination of the abnormal data includes: detecting a condition number of the Hessian matrix to identify a disease state matrix; Singular value decomposition identification is carried out on the disease state matrix, and a colinear parameter pair is removed; constructing a Hessian matrix according to an adaptive regularization coefficient algorithm; calculating an inverse matrix of the Hessian matrix to obtain a parame