Method and device for dynamic tuning and business verification of infectious disease early warning model in frequent cross-regional population flow area

Abstract

The application discloses a kind of infectious disease early warning model dynamic tuning and business verification method and device of frequent area of cross-regional population flow, belong to infectious disease prevention and control technical field.Method obtains epidemic monitoring, population flow, regional heterogeneity, meteorological environment, intervention measure and business feedback multi-class data;Based on target infectious disease dynamics model, combined with multi-source data, construct cross-regional coupling propagation model, output the predicted case number and comprehensive propagation risk index of each region in predetermined time window;Combined with prediction result and business feedback data, dynamically tune model parameter, obtain optimized early warning model;Based on tuning model, predict epidemic trend, output early warning result and generate business application guide.The present application can realize model parameter dynamic tuning, improve cross-regional prediction accuracy, form complete business closed loop.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, and in particular to a method and apparatus for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement. Background Technology

[0002] The main problem with current technologies is not the lack of infectious disease prediction models, but rather that most models remain at the stage of "offline modeling, historical backtesting, and one-time deployment," lacking the ability to continuously validate and optimize them for real-world disease control scenarios. Especially in regions with frequent population movement, such as the Beijing-Tianjin-Hebei region, the Yangtze River Delta, and the Guangdong-Hong Kong-Macau Greater Bay Area, epidemic transmission is influenced by multiple factors, including cross-city commuting, imported risks, regional heterogeneity, pathogen mutations, population immunity levels, and changes in prevention and control policies. Transmission patterns continuously change, while existing model parameters are often fixed after deployment, making it difficult to identify and correct conceptual drift in a timely manner, leading to a decline in prediction accuracy over time.

[0003] In related technologies, the predicted epidemic trends of infectious diseases typically do not consider factors such as regional population movement or the natural mortality rate of patients. This leads to the inability to determine the epidemic trend of infectious diseases in a specific region. Furthermore, the factors considered in determining the epidemic trend are not comprehensive enough, resulting in inaccurate predictions. Meanwhile, existing systems generally emphasize case data and the model algorithm itself, neglecting the feedback value from operational personnel in real-world early warning and response. They fail to translate operational feedback such as early warning confirmation, false alarms, underreporting, intervention measures, and response effects into a basis for model optimization. Therefore, there is a tendency for model output to become disconnected from actual disease control assessments, leading to problems such as delayed early warnings, persistent false positives, persistent false negatives, or risk levels that do not match operational understanding. In addition, existing model evaluations largely rely on statistical indicators such as mean squared error and accuracy, lacking a comprehensive assessment of early warning lead time, peak bias, persistence of false positives / false negatives, and operational usability, making it difficult to truly reflect the model's application value in public health practice.

[0004] More importantly, many early warning models lack the ability to deeply integrate with national or provincial infectious disease monitoring and early warning platforms, making it impossible to establish a standardized process from data access, model operation, early warning output, operational feedback, parameter optimization to report generation. As a result, even if the models perform well in experimental data, they are difficult to operate stably in real-world cross-regional scenarios in the long term, and it is even more difficult to form a replicable, scalable, and auditable operational application model. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement, which can achieve dynamic optimization of model parameters, improve cross-regional prediction accuracy, and form an operational closed loop.

[0006] Firstly, this application proposes a method for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement. The method includes: Acquire epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data; Based on the target infectious disease dynamics model, a cross-regional coupled transmission model is constructed using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window. The target infectious disease dynamics model is used to indicate the state transition relationships between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions. These multiple regions have different disease transmission risks, and a coupling relationship is established between them through a population flow matrix. Based on the predicted number of cases and the comprehensive transmission risk index, and combined with the business feedback data, the parameters of the target infectious disease dynamics model are dynamically optimized to obtain the optimized infectious disease early warning model. Using the optimized infectious disease early warning model, the epidemic trend of infectious diseases in the multiple regions is predicted, early warning results are output, and business application guidelines are generated.

[0007] Furthermore, the cross-regional coupled transmission model based on the target infectious disease dynamics model, utilizing the epidemic monitoring data, the population flow data, the regional heterogeneity data, the meteorological environment data, and the intervention measure data, includes: Based on the target infectious disease dynamics model, the epidemic status variables for each of the multiple regions are determined, including the size of susceptible individuals, the size of latent individuals, the size of infected individuals, and the size of removed individuals. Based on the population flow data, an inter-regional population flow matrix is ​​constructed, where the elements of the population flow matrix represent the population size or normalized flow intensity flowing from the first region to the second region. Based on the inter-regional population flow matrix, standardized flow weights are calculated, and external input infection pressure is calculated based on the standardized flow weights. The overall infection pressure is calculated based on the external input infection pressure, local transmission risk, and regional heterogeneity risk factors. Based on the overall infection pressure, a set of regional heterogeneous transmission differential equations with population mobility coupling terms is established. The set of transmission differential equations includes equations for the rate of change of susceptible individuals, the rate of change of latent individuals, the rate of change of infected individuals, and the rate of change of removed individuals.

[0008] Furthermore, the formula for calculating the standardized flow weight is as follows: wij ( t )=∑ k =1 Nmik ( t )+ ϵmij ( t ); in, wij ( t ) represents the standardized flow weights from region i to region j at time t. mij ( t () represents the population size that flows from region i to region j at time t. N Indicates the number of regions included in the model. ϵ To avoid extremely small positive numbers with a denominator of zero; The formula for calculating the externally input infection pressure is: Λ iout ( t )=∑ j = iwji ( t ) Pj ( t ) Ij ( t ); Among them, Λ iout ( t This represents the external infection pressure in region i. wji ( t ) represents the standardized flow weights from region j to region i. Ij ( t () represents the number of infected individuals in region j at time t. Pj ( t ) represents the resident or real-time population size of region j.

[0009] Furthermore, the formula for calculating the overall infection pressure is as follows: λi ( t )= βi ( t ) Pi ( t ) Ii ( t )+ ηi ( t )∑ j = iwji ( t ) Pj ( t ) Ij ( t ); in, λi( t ) represents the overall infection pressure in region i. βi ( t ) represents the local effective contact rate of region i. ηi ( t () represents the external input risk amplification factor. Ii ( t () represents the number of infected individuals in region i at time t. Pi ( t ) represents the resident or real-time population size of region i; The set of regional heterogeneous propagation differential equations with population flow coupling terms is as follows: dtdSi ( t )=- λi ( t ) Si ( t ); dtdEi ( t )= λi ( t ) Si ( t )- σiEi ( t ); dtdIi ( t )= σiEi ( t )- γiIi ( t ); dtdRi ( t )= γiIi ( t ); in, Si ( t ), Ei ( t ), Ii ( t ), Ri ( t ) represent the susceptible population, latent population, infected population, and removed population of region i at time t, respectively. σi This indicates the rate at which latent individuals become infected. γi This indicates the rate at which infected individuals recover or are removed.

[0010] Furthermore, the dynamic optimization of the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, includes: Based on the predicted number of cases and the actual number of monitored cases, the point prediction error index, peak prediction deviation index, early warning advance index, false positive rate index, and false negative rate index are calculated. Based on the business feedback data, extract the business confirmation results, intervention measures information, false positive labels, false negative labels, and expert consensus scores; Based on the point prediction error index, the peak prediction deviation index, the early warning advance index, the false positive rate index, the false negative rate index, and the business feedback data, determine whether the parameter tuning triggering conditions are met. Under the condition that the parameter tuning triggering condition is met, the population flow weight, local effective contact rate, external input risk amplification coefficient, and risk warning threshold of the target infectious disease dynamics model are dynamically corrected based on the verification dataset and business feedback dataset within the rolling time window, so as to obtain the tuned model parameter set.

[0011] Furthermore, the point prediction error indicators include mean absolute error, root mean square error, and mean absolute percentage error. The formula for calculating the mean absolute error is: MAEi = h 1∑ τ =1 hYi ( t + τ | t )- Yi ( t + τ ); The formula for calculating the root mean square error is: RMSEi = h 1∑ τ =1 h ( Yi ( t + τ | t )- Yi ( t + τ ))2; The formula for calculating the mean absolute percentage error is: MAPEi = h 1∑ τ =1 hYi ( t + τ )+ ϵYi ( t + τ | t )- Yi ( t +τ ); in, Yi ( t + τ | t ) represents the prediction of region i at prediction starting point t for the future. t + τ Predicted number of cases at any given time Yi ( t + τ ) indicates that region i is in t + τ Real-time monitoring of case numbers h Indicates the prediction window length. ϵ To avoid extremely small positive numbers with a denominator of zero.

[0012] Furthermore, the peak prediction deviation index includes peak time deviation and peak intensity deviation; The formula for calculating the peak time deviation is: Ditime = tipeak - tipeak; The formula for calculating the peak intensity deviation is: Dipeak = Yi ( tipeak )+ ϵYi ( tipeak )- Yi ( tipeak ); in, tipeak This represents the actual peak time of the epidemic in region i. tipeak This indicates the predicted peak time of the epidemic in region i. Yi ( tipeak ) represents the actual number of cases in region i at the true peak time. Yi ( tipeak ) represents the predicted number of cases in region i at the predicted peak time.

[0013] Furthermore, the formula for calculating the advance warning amount is as follows: Li = tievent - tiwarn; in, tievent This indicates the time when the actual epidemic situation in region i reaches the warning threshold. tiwarn This indicates the time when the model first issues a warning for region i; The parameter tuning triggering conditions include at least one of the following: the overall prediction error exceeds a preset error tolerance threshold, the peak time deviation exceeds a preset peak deviation threshold, the number of false negatives exceeds a preset false negative tolerance threshold, the number of false positives exceeds a preset false positive tolerance threshold, and the number of continuous abnormal cycles reaches a preset cycle threshold.

[0014] Furthermore, the dynamic correction of the population movement weights in the target infectious disease dynamics model includes: The population mobility weight adjustment coefficient is calculated based on the ratio of the predicted residual to the actual number of cases. The original crowd flow weights are updated based on the crowd flow weight adjustment coefficients to obtain the updated crowd flow weights. The updated crowd flow weights are normalized to obtain normalized crowd flow weights. The update formula for the crowd flow weight is: wjinew ( t )= wjiold ( t )[1+ κwYi ( t )+ ϵei ( t )]; in, wjinew ( t () represents the updated population flow weight from region j to region i. wjiold ( t () represents the population flow weight from region j to region i before the update. ei ( t ) represents the predicted residual for region i. Yi ( t ) represents the actual number of monitored cases in region i. κw This represents the learning rate for population mobility weights.

[0015] Furthermore, the dynamic correction of the local effective contact rate of the target infectious disease dynamics model includes: The contact rate adjustment index is calculated based on the ratio of the predicted residuals to the actual number of cases. Based on the contact rate adjustment index, the original local effective contact rate is updated to obtain the updated local effective contact rate. The update formula for the local effective contact rate is: βinew ( t )= βiold ( t ) exp ( κβYi ( t )+ϵYi ( t )- Yi ( t )); in, βinew ( t () represents the updated local effective contact rate for region i. βiold ( t This represents the local effective contact rate of region i before the update. Yi ( t ) represents the actual number of monitored cases in region i. Yi ( t ) represents the predicted number of cases in region i. κβ This indicates the contact rate update step size; The updated local effective contact rate meets the preset parameter boundary constraints: βi min≤ βinew ( t )≤ β i max.

[0016] Furthermore, the dynamic correction of the risk warning threshold of the target infectious disease dynamics model includes: Statistically analyze the number of false positives and false negatives in each region within the most recent validation window; Based on the number of false positives, the number of false negatives, and the business loss weights for false positives and false negatives, the threshold adjustment amount is calculated. The original risk warning threshold is updated based on the threshold adjustment amount to obtain the updated risk warning threshold. The update formula for the risk warning threshold is: ρinew = ρiold + κρ ( cFPFPi - cFNFNi ); in, ρinew This represents the updated risk warning threshold for region i. ρiold This represents the risk warning threshold for region i before the update. FPi This represents the number of false positives in region i within the most recent validation window. FNi This represents the number of false negatives in region i within the most recent validation window. cFP Indicates the business loss weighting for false positives. cFN The weight of business losses indicated by false negatives. κρ This indicates the threshold adjustment step size.

[0017] Secondly, this application also proposes a device for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement, the device comprising: The acquisition module is configured to acquire epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data; The construction module is configured to build a cross-regional coupled transmission model based on the target infectious disease dynamics model, using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data, to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window; wherein, the target infectious disease dynamics model is used to indicate the state transition relationship between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions, the multiple regions have different disease transmission risks, and the multiple regions are coupled through a population flow matrix; The optimization module is configured to dynamically optimize the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, to obtain the optimized infectious disease early warning model. The prediction module is configured to use the optimized infectious disease early warning model to predict the epidemic trend of infectious diseases in the multiple regions, output the early warning results, and generate business application guidelines.

[0018] Thirdly, this application also proposes an electronic device, comprising: processor; Memory for storing the executable instructions of the processor; The processor is configured to execute the aforementioned method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement by executing the executable instructions.

[0019] Fourthly, this application also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for dynamic optimization and operational verification of an infectious disease early warning model for areas with frequent cross-regional population movement.

[0020] Compared with existing technologies, this application provides a method and apparatus for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement. By fusing multi-source data to construct a cross-regional coupled propagation model and combining it with operational feedback to achieve dynamic parameter optimization, it effectively solves the problems of fixed model parameters, lack of coupling relationships between regions, and insufficient integration of operational feedback. This can improve prediction accuracy, enhance the timeliness of early warning, and form an operational closed loop. Attached Figure Description

[0021] Figure 1This is a schematic diagram of a method for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement, provided in an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of the architecture of a device for dynamic optimization and operational verification of an infectious disease early warning model in a region with frequent cross-regional population movement, provided in an embodiment of this application. Detailed Implementation

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] like Figure 1 As shown in the embodiments of this application, a method for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement is proposed. The method includes: S100: Obtain epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data; Epidemic monitoring data can be obtained through manual entry, export from hospital or CDC databases, etc. Population mobility data can be estimated based on historical statistics or obtained through simple questionnaires. Regional heterogeneity data can be found in publicly available statistical yearbooks. Meteorological and environmental data can be obtained from historical records by meteorological departments. Intervention data can be manually entered by policymakers. Operational feedback data can be provided by disease control personnel through oral reports or unstructured text records. This data is collected and processed for use in subsequent models.

[0026] S200. Based on the target infectious disease dynamics model, using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data, a cross-regional coupled transmission model is constructed to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window; wherein, the target infectious disease dynamics model is used to indicate the state transition relationship between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions, the multiple regions have different disease transmission risks, and the multiple regions are coupled through a population flow matrix; This model can be configured to operate independently within each region and simulate population flow between regions using a preset fixed ratio, thus achieving inter-regional coupling. For example, it can be assumed that a fixed percentage of the population flows from one region to another daily, and the infection status of these mobile populations can be simply superimposed onto the target region. Therefore, within a predetermined time window, the predicted number of cases in each region can be calculated, and a rough comprehensive transmission risk index can be obtained based on the ratio of these cases to the regional population. This targeted infectious disease dynamics model is used to indicate the state transition relationships between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions with different disease transmission risks, and these regions are coupled through a population flow matrix.

[0027] S300. Based on the predicted number of cases and the comprehensive transmission risk index, and combined with the business feedback data, the parameters of the target infectious disease dynamics model are dynamically optimized to obtain the optimized infectious disease early warning model. When a significant discrepancy is observed between the predicted number of cases and the actual number of cases monitored, or when operational feedback data indicates that the model's early warning effect is poor, model experts can manually adjust some key parameters in the model based on experience, such as the transmission rate or recovery rate. This adjustment process can be iterative until the model's output of predicted case numbers and comprehensive transmission risk index matches the actual situation to a certain extent. In this way, an optimized infectious disease early warning model can be obtained.

[0028] S400. Using the optimized infectious disease early warning model, predict the epidemic trend of infectious diseases in the multiple regions, output the early warning results, and generate a business application guide.

[0029] Based on the simulation results, reports on the development trend of the epidemic in various regions can be generated, such as predicting the range of changes in the number of cases in the coming week. Based on these predictions, it can be manually determined whether an early warning needs to be issued, and according to the warning level, an expert team can manually write operational guidelines, such as advising the public to pay attention to personal protection or advising local governments to strengthen community control.

[0030] In some possible embodiments, the construction of a cross-regional coupled transmission model based on the target infectious disease dynamics model, utilizing the epidemic monitoring data, the population flow data, the regional heterogeneity data, the meteorological environment data, and the intervention measure data, includes: Based on the target infectious disease dynamics model, the epidemic status variables for each of the multiple regions are determined, including the size of susceptible individuals, the size of latent individuals, the size of infected individuals, and the size of removed individuals. The susceptible population refers to the number of people who are not infected but could potentially be infected; the latent population refers to the number of people who are infected but have not yet shown symptoms and are not yet infectious; the infected population refers to the number of people who are infected and infectious; and the removed population refers to the number of people who have recovered or died and are no longer infectious or susceptible. Determining these variables is fundamental to constructing any infectious disease dynamics model. They are interconnected through differential equations and collectively characterize the disease transmission process. In practice, these variables can be estimated through initial data settings, fitting historical epidemic data, or based on census data.

[0031] Based on the population flow data, an inter-regional population flow matrix is ​​constructed, where the elements of the population flow matrix represent the population size or normalized flow intensity flowing from the first region to the second region. Its elements represent the population size moving from one region to another at a specific time, or the normalized intensity of the movement. This matrix reflects the potential impact of population movement on disease transmission, as population movement is a primary vehicle for cross-regional viral transmission. Population movement data can come from various sources, such as mobile phone signaling data, transportation data (flights, high-speed rail, road passenger volume), social media check-in data, or census data. Normalization eliminates the impact of differences in population size across regions, making movement intensity more comparable.

[0032] Based on the inter-regional population flow matrix, standardized flow weights are calculated, and external input infection pressure is calculated based on the standardized flow weights. Standardized mobility weights are an indicator used to measure the impact of population flows from one region to another on the epidemic situation in the target region. Their calculation aims to eliminate the impact of differences in population size between regions, making mobility intensity more comparable. External infection pressure quantifies the additional transmission risk posed by infected individuals flowing into the target region from other regions to the local epidemic situation. Calculating standardized mobility weights lays the foundation for accurately assessing external infection pressure and reflects the contribution of external infection sources to the local epidemic situation.

[0033] The overall infection pressure is calculated based on the external input infection pressure, local transmission risk, and regional heterogeneity risk factors. Overall infection pressure is a comprehensive measure of the factors driving susceptible individuals within a region to become latent carriers, taking into account all sources of disease transmission. This pressure includes not only the risk of local transmission from contact between infected and susceptible individuals within the region, but also the risk of infection from population movement from external areas, as well as heterogeneous risk factors within the region itself, such as population density, healthcare resources, and hygiene habits. By integrating these factors, a more accurate characterization of the infection threats faced by susceptible individuals within the region can be achieved.

[0034] Based on the overall infection pressure, a set of regional heterogeneous transmission differential equations with population mobility coupling terms is established. The set of transmission differential equations includes equations for the rate of change of susceptible individuals, the rate of change of latent individuals, the rate of change of infected individuals, and the rate of change of removed individuals.

[0035] This system of equations dynamically describes the population changes of four groups—susceptible individuals, latent individuals, infected individuals, and removers—in each region through a series of interrelated differential equations. By introducing a population mobility coupling term, the system of equations can dynamically reflect the impact of population mobility on the epidemic state variables in each region, enabling the model to simulate the interaction and spread of the disease between different regions, thus more realistically reflecting the dynamic evolution of the epidemic.

[0036] In some possible embodiments, the formula for calculating the standardized flow weight is: wij ( t )=∑ k =1 Nmik ( t )+ ϵmij ( t ); in, wij ( t ) represents the standardized flow weight from region i to region j at time t, and its function is to quantify the relative proportion of the population flow intensity from region i to region j in the total outflow population of region i. mij ( t The value represents the population size that flows from region i to region j at time t. This data can be obtained from population flow data, such as through mobile phone signaling data, transportation data, or census data. N This indicates the number of regions included in the model, i.e., the total number of regions involved in the calculation of the coupling propagation model; ϵ To avoid extremely small positive numbers with a denominator of zero.

[0037] The formula for calculating the externally input infection pressure is: Λ iout ( t )=∑ j = iwji (t ) Pj ( t ) Ij ( t ); Among them, Λ iout ( t () represents the external infection pressure of region i, and its function is to quantify the infection risk brought about by population flow from other regions j into region i; wji ( t ) represents the standardized flow weights from region j to region i. Ij ( t The number of infected individuals in region j at time t represents the size of the infected population. This data is one of the epidemic state variables of region j in the target infectious disease dynamics model. Pj ( t The denoted represents the resident or real-time population size of region j, which can be obtained from population statistics departments or real-time population monitoring systems. By multiplying the number of infected individuals in each source region j by the population size of that region, then multiplying by the standardized flow weight from j to i, and summing this up for all source regions j, the overall infection risk faced by region i due to external population inflows can be accurately assessed.

[0038] In some possible embodiments, the formula for calculating the overall infection pressure is: λi ( t )= βi ( t ) Pi ( t ) Ii ( t )+ ηi ( t )∑ j = iwji ( t ) Pj ( t ) Ij ( t ); in, λi ( t ) represents the overall infection pressure in region i. βi ( t ) represents the local effective contact rate of region i. ηi ( t () represents the external input risk amplification factor. Ii ( t () represents the number of infected individuals in region i at time t. Pi ( t ) represents the resident or real-time population size of region i; Overall infection pressure λi( t The effective local contact rate (FDR) is a key parameter driving the conversion of susceptible individuals into latent individuals in region i, directly determining the speed and scale of disease transmission within region i. Specifically, this formula calculates the effective local contact rate in region i. βi ( t ), the resident or real-time population size of region i Pi ( t ) and the scale of infections in region i at time t Ii ( t The contribution of local transmission is represented by multiplication; simultaneously, the risk amplification factor for external input is used. Ii ( t Multiplied by the standardized flow weights from other regions wji ( t ), the resident or real-time population size of region j Pi ( t ) and the scale of infected persons in region j at time t Ii ( t The sum of the products of these factors represents the contribution of imported infections. Among them, the effective local contact rate... βi ( t The risk amplification factor for external input can be dynamically estimated and adjusted based on historical epidemic data, meteorological environmental factors, and intervention measures to reflect changes in local transmission risk; ηi ( t This is used to quantify the amplification effect of infection risk brought about by external population movement on the local spread of the epidemic. Its value can be calibrated according to disease characteristics, regional geographical location and population movement patterns.

[0039] The set of regional heterogeneous propagation differential equations with population flow coupling terms is as follows: dtdSi ( t )=- λi ( t ) Si ( t ); dtdEi ( t )= λi ( t ) Si ( t )- σiEi ( t ); dtdIi ( t )= σiEi ( t )- γiIi ( t ); dtdRi ( t )=γiIi ( t ); in, Si ( t ), Ei ( t ), Ii ( t ), Ri ( t ) represent the susceptible population, latent population, infected population, and removed population of region i at time t, respectively. σi This indicates the rate at which latent individuals become infected. γi Indicates the rate at which infected individuals recover or are removed; This set of differential equations is an extension of the SEIR (Susceptible-Latent-Infected-Removed) model to consider regional heterogeneity and the coupling effects of population mobility. It describes the situation under a given overall infection pressure. λi ( t Under the condition that the number of susceptible individuals, latent individuals, infected individuals, and removers in region i changes over time.

[0040] Specifically, the variability equation for susceptible individuals dtdSi ( t )=- λi ( t ) Si ( t This indicates that the rate of reduction in the number of susceptible individuals is related to the overall infection pressure. λi ( t ) and the current size of susceptible individuals Si ( t The rate of change in latent individuals is directly proportional to the infection pressure; that is, the greater the infection pressure, the faster the rate at which susceptible individuals become latent individuals. (Equation for the rate of change in latent individuals) dtdEi ( t )= λi ( t ) Si ( t )- σiEi ( t This indicates that the increase in the number of latent individuals stems from the conversion of susceptible individuals, while the decrease stems from latent individuals becoming infected. σi This is the rate at which the incubation period ends and individuals become infected. Equation for the rate of change in infected individuals. dtdIi ( t )= σiEi ( t )- γiIi ( t This indicates that the increase in the number of infected individuals stems from the conversion of latent individuals, while the decrease arises from the recovery or removal of infected individuals. γi This represents the rate of recovery or removal of infected individuals. Equation for the rate of change in the number of individuals removed. dtdRi ( t )=γiIi ( t This indicates that the increase in the number of removers stems from the recovery or removal of infected individuals. (Parameter) σ i and γi These are biological parameters of the disease itself, which can be estimated based on the average incubation period and average incubation period. This set of equations utilizes overall infection pressure. λi ( t This implicitly includes the external infection pressure brought about by population movement, thereby achieving inter-regional coupling.

[0041] In some possible embodiments, the dynamic optimization of the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the operational feedback data, includes: Based on the predicted number of cases and the actual number of monitored cases, the point prediction error index, peak prediction deviation index, early warning advance index, false positive rate index, and false negative rate index are calculated. These metrics are used to quantify the predictive performance of infectious disease early warning models. Point prediction error metrics, such as mean absolute error, root mean square error, or mean absolute percentage error, measure the accuracy of the model's predictions of case numbers at a specific point in time, reflecting the average deviation between predicted and actual values. Peak prediction bias metrics, such as peak time bias and peak intensity bias, focus on assessing the model's ability to predict the timing and intensity of peak outbreaks, which is crucial for the early deployment of public health resources. Early warning lead time metrics measure the lead time between the model issuing an early warning and the actual occurrence of an outbreak; a higher value indicates better timeliness. False positive rate and false negative rate metrics assess the proportion of false early warnings (warnings issued despite no outbreak) and failures to issue warnings (outbreaks occurring but no warning), respectively, directly impacting the reliability of the early warning system and operational losses. These metrics are typically calculated by comparing the predicted number of cases output by the model within a predetermined time window with the actual number of cases monitored during the same period.

[0042] Based on the business feedback data, extract the business confirmation results, intervention measures information, false positive labels, false negative labels, and expert consensus scores; Business feedback data is an indispensable component of model tuning, providing qualitative and semi-quantitative information from real-world business scenarios. Business confirmation results refer to the actual effects or decision feedback from public health departments or related institutions after taking action based on model warnings, such as whether an emergency response was initiated or whether specific intervention measures were implemented. Intervention information records the type, intensity, and duration of non-pharmaceutical or other public health interventions implemented in a specific period and region; this information is crucial for understanding the dynamics of epidemic transmission and evaluating the effectiveness of model parameters. False positive and false negative labels are manual annotations made by business personnel or experts based on actual situations; they directly indicate specific cases where model predictions do not match reality. Expert consensus scores reflect the degree of consistency among multiple experts in their assessment of model warning results or model performance, providing authoritative professional opinions for model tuning. These data are typically extracted through structured questionnaires, business system logs, expert interviews, or manual review.

[0043] Based on the point prediction error index, the peak prediction deviation index, the early warning advance index, the false positive rate index, the false negative rate index, and the business feedback data, determine whether the parameter tuning triggering conditions are met. To ensure the timeliness and effectiveness of model parameter tuning, a clear triggering mechanism needs to be established. This mechanism comprehensively considers the model's quantitative performance indicators and business feedback information. For example, when the overall prediction error consistently exceeds a preset error tolerance threshold, it indicates a significant decrease in the model's prediction accuracy; when the peak time deviation exceeds a preset peak deviation threshold, it suggests that the model's ability to predict key stages of the epidemic is insufficient; when the number of false negatives or false positives exceeds a preset tolerance threshold within a certain period, and the number of consecutive abnormal false positive periods reaches a preset period threshold, it indicates that the model has problems with the reliability of its early warning. These conditions can be used individually or in combination, and once any one of them is met, the parameter tuning process is triggered.

[0044] Under the condition that the parameter tuning triggering condition is met, the population flow weight, local effective contact rate, external input risk amplification coefficient, and risk warning threshold of the target infectious disease dynamics model are dynamically corrected based on the verification dataset and business feedback dataset within the rolling time window, so as to obtain the tuned model parameter set.

[0045] When the parameter tuning trigger conditions are met, the model enters the dynamic correction phase. This process is based on the validation dataset and business feedback dataset within a rolling time window to ensure that the corrected parameters adapt to the latest epidemic dynamics and business needs. The population mobility weight reflects the impact of inter-regional population movement on disease transmission; its correction aims to more accurately characterize the coupling strength between different regions. The local effective contact rate is a key parameter for measuring the speed of disease transmission within a region; its correction helps the model better capture the characteristics of local transmission. The external input risk amplification coefficient is used to adjust the degree of impact of external infection pressure on the local epidemic; its correction enables the model to more accurately assess external risks. The risk warning threshold is a key parameter for determining when to issue a warning; its correction aims to balance the business losses between false positives and false negatives, making the warning results more in line with actual business needs. The dynamic correction of these parameters typically employs iterative optimization algorithms, combining historical data and real-time feedback to gradually adjust parameter values ​​to minimize prediction errors and optimize warning performance.

[0046] In some possible embodiments, the point prediction error metrics include mean absolute error, root mean square error, and mean absolute percentage error. The formula for calculating the mean absolute error is: MAEi = h 1∑ τ =1 hYi ( t + τ | t )- Yi ( t + τ ); The formula for calculating the root mean square error is: RMSEi = h 1∑ τ =1 h ( Yi ( t + τ | t )- Yi ( t + τ ))2; The formula for calculating the mean absolute percentage error is: MAPEi = h 1∑ τ =1 hYi ( t + τ )+ ϵYi ( t + τ | t )- Yi (t + τ ); in, Yi ( t + τ | t ) represents the prediction of region i at prediction starting point t for the future. t + τ Predicted number of cases at any given time Yi ( t + τ ) indicates that region i is in t + τ Real-time monitoring of case numbers h Indicates the prediction window length. ϵ To avoid extremely small positive numbers with a denominator of zero.

[0047] As an implementation method, the point prediction error index is a statistic used to quantify the difference between model predictions and actual observations. Its role is to evaluate the accuracy of the model's predictions at a specific point in time or within a specific time period, and it serves as an important basis for dynamic tuning of model parameters. The mean absolute error (MAE) used in this embodiment directly reflects the average magnitude of the prediction error by calculating the average of the absolute differences between the predicted number of cases and the actual number of monitored cases. It has good robustness to outliers. The root mean square error (RMSE), on the other hand, calculates the square root of the mean of the squares of the differences between the predicted number of cases and the actual number of monitored cases. It is more sensitive to the magnitude of the error, especially imposing a heavier penalty on larger prediction deviations, thus reflecting the dispersion of the predicted values. The mean absolute percentage error (MAPE) provides a measure of relative error. By calculating the average of the absolute percentage errors between the predicted number of cases and the actual number of monitored cases, it makes it possible to compare the prediction performance of datasets with different dimensions or magnitudes. However, it is important to note that a very small positive number ϵ should be introduced when the actual number of cases is close to zero to avoid calculation errors.

[0048] In the above calculation formula, Yi ( t + τ | t ) represents the model's prediction starting point. t For the future t + τ The predicted number of cases at any given time is calculated by the cross-regional coupled propagation model with the given input data and current model parameters. Yi ( t + τ This indicates that region i is in t + τ The real-time number of monitored cases, typically obtained through actual epidemic monitoring systems, serves as a benchmark for evaluating the accuracy of model predictions. Prediction window length. hThe time range for calculating these error metrics is defined, and its specific settings can be adjusted according to actual business needs and data update frequency. Minimal positive numbers. ϵ The introduction of this method is to account for the actual number of monitored cases when calculating the mean absolute percentage error. Yi ( t + τ When the denominator is zero or very close to zero, mathematical errors such as zero denominator are avoided, thus ensuring the stability of the calculation process and the validity of the results.

[0049] In some possible embodiments, the peak prediction deviation index includes peak time deviation and peak intensity deviation; As an implementation method, the peak prediction bias index is a type of metric specifically used to evaluate the accuracy of infectious disease early warning models in predicting the peak of an epidemic. This index system focuses on the most critical turning point in the development of an epidemic—the peak. By quantifying the difference between the peak time predicted by the model and the actual peak time, as well as the difference between the predicted peak intensity and the actual peak intensity, it provides a more targeted evaluation of model performance.

[0050] The formula for calculating the peak time deviation is: Ditime = tipeak - tipeak; The formula for calculating the peak intensity deviation is: Dipeak = Yi ( tipeak )+ ϵYi ( tipeak )- Yi ( tipeak ); in, tipeak This represents the true peak time of the epidemic in region i, that is, the time when the actual number of cases reached its highest point; tipeak This represents the predicted peak time of the epidemic in region i, i.e., the time when the number of cases predicted by the model reaches its highest point. By calculating the difference between these two times, we can intuitively reflect the accuracy of the model's prediction of the epidemic peak in the time dimension. Accurately predicting the peak time is of decisive significance for public health departments to promptly activate emergency responses, allocate medical resources, and implement intervention measures. Yi ( tipeak ) represents the actual number of cases in region i at the true peak time, that is, the total number of cases during the actual peak of the epidemic; Yi ( tipeak ) represents the predicted number of cases in region i at the predicted peak time, that is, the total number of cases predicted by the model during the peak of the epidemic; ϵThis is a very small positive number, used to avoid the denominator being zero in specific computational scenarios. By calculating the deviation between the predicted peak number of cases and the actual peak number of cases, the model's ability to predict the severity of the peak of the epidemic in terms of quantity can be evaluated. Accurately predicting the peak intensity helps assess the potential impact of the epidemic, thereby enabling the development of appropriate prevention and control strategies.

[0051] In some possible embodiments, the formula for calculating the advance warning amount is: Li = tievent - tiwarn; in, tievent This indicates the time when the actual epidemic situation in region i reaches the warning threshold. tiwarn This indicates the time when the model first issues a warning for region i; the lead time for this warning. Li The indicator is used to quantify the timeliness of the model's early warning, that is, how long before an actual epidemic event occurs can the model issue an effective early warning. A large positive early warning lead time is considered a significant indicator. Li This demonstrates that the model can identify potential risks and issue warnings earlier, which is crucial for public health departments to gain valuable intervention time.

[0052] Parameter tuning trigger conditions refer to a set of preset criteria used to determine whether model parameters need dynamic adjustment. These conditions aim to ensure that when the model's performance significantly degrades or fails to meet actual business needs, the parameter tuning process can be initiated promptly and effectively, thereby avoiding unnecessary frequent adjustments or delayed adjustments.

[0053] The parameter tuning triggering conditions include at least one of the following: the overall prediction error exceeds a preset error tolerance threshold, the peak time deviation exceeds a preset peak deviation threshold, the number of false negatives exceeds a preset false negative tolerance threshold, the number of false positives exceeds a preset false positive tolerance threshold, and the number of continuous abnormal cycles reaches a preset cycle threshold.

[0054] Specifically, exceeding the preset error tolerance threshold for overall prediction error means that when the model's overall prediction accuracy (measured by metrics such as mean absolute error, root mean square error, or mean absolute percentage error) within a rolling time window consistently or significantly exceeds the acceptable error range, it indicates that the model parameters may no longer be suitable for the current transmission dynamics and require adjustment. Exceeding the preset peak time deviation threshold means that the prediction of the peak of an infectious disease epidemic curve is crucial for resource allocation and intervention strategy development. When there is a significant difference between the model's predicted peak time and the actual observed peak time, and this deviation exceeds the preset tolerance threshold, the model parameters need to be corrected to improve the accuracy of peak prediction. Exceeding the preset false negative tolerance threshold means that a false negative warning indicates that the model has failed to identify the actual epidemic risk, which may lead to serious public health consequences. When the number of false negative warnings issued by the model within a certain period exceeds the preset tolerance threshold, it indicates that the model may be too conservative or insufficiently sensitive, requiring parameter adjustment to improve its risk identification capability. A false positive count exceeding the preset false positive tolerance threshold and a sustained abnormal period reaching the preset period threshold indicates that while false positive warnings may not lead to an uncontrolled epidemic, they could cause unnecessary waste of resources and public panic. When the number of false positive warnings issued by the model within a certain period exceeds the preset tolerance threshold, and this abnormal state persists for a certain number of periods, it indicates that the model may be too aggressive or lack specificity, requiring parameter adjustments to optimize its warning accuracy. The introduction of the "sustainable abnormal period number" is to avoid frequent optimization triggers due to short-term fluctuations and to enhance the robustness of optimization decisions.

[0055] In some possible embodiments, the dynamic correction of the population movement weights in the target infectious disease dynamics model includes: The population mobility weighting adjustment coefficient is calculated based on the ratio of the predicted residuals to the actual number of cases; the predicted residuals... ei ( t The value () represents the difference between the model's predicted number of cases and the actual number of monitored cases in region i. Its sign and magnitude directly reflect the model's prediction bias. Actual number of cases. Yi ( t This provides information on the actual scale of the epidemic. The prediction residuals are calculated. ei ( t ) and the actual number of cases Yi ( tThe ratio of the predicted value to the actual value can quantify the degree of prediction bias of the model in a specific region, and generate an adjustment factor based on this. For example, when the model's predicted value is lower than the actual value, the residual is positive, indicating that the model may have underestimated the infection risk in the region, and the weight of population flow into the region needs to be increased; conversely, when the model's predicted value is higher than the actual value, the residual is negative, indicating that the model may have overestimated the infection risk, and the weight of inflow needs to be decreased. This ratio, as input to the adjustment coefficient, can guide the direction and magnitude of the correction of the population flow weight, ensuring the rationality and effectiveness of the adjustment.

[0056] The original crowd flow weights are updated based on the crowd flow weight adjustment coefficients to obtain the updated crowd flow weights; this step is the core of achieving dynamic weight correction. The update mechanism typically uses a multiplicative form, as shown in the formula... wjinew ( t )= wjiold ( t )[1+ κwYi ( t )+ ϵei ( t As shown in the formula [1+] κwYi ( t )+ ϵei ( t The term ] is the adjustment factor based on the ratio of predicted residuals to the actual number of cases. When ei ( t When ) is positive, the adjustment factor is greater than 1, making wjinew ( t (greater than) wjiold ( t That is, increase the weight of population flow from region j to region i; when ei ( t When ) is negative, the adjustment factor is less than 1, making wjinew ( t Less than wjiold ( t This means reducing the weight of population flow from region j to region i. κw It is the learning rate of the crowd flow weights, used to control the step size of each adjustment in order to balance the sensitivity and stability of the adjustment. ϵ It is a very small positive number used to avoid the case where the denominator is zero, thus ensuring the stability of the calculation.

[0057] The updated population flow weights are then normalized to obtain normalized population flow weights. The purpose of normalization is to ensure that the population flow weights are within a reasonable numerical range and maintain their physical meaning. For example, it can be ensured that the sum of the population flow weights from one region to all other regions is 1, or the weight values ​​can be limited to between 0 and 1. This helps maintain the stability and interpretability of the model, avoids abnormal model behavior caused by excessively large or small weight values, and ensures the rationality of the population flow matrix, i.e., that the total proportion of the population flowing out of a region is consistent with reality.

[0058] The update formula for the crowd flow weight is: wjinew ( t )= wjiold ( t )[1+ κwYi ( t )+ ϵei ( t )]; in, wjinew ( t () represents the updated population flow weight from region j to region i. wjiold ( t () represents the population flow weight from region j to region i before the update. ei ( t ) represents the prediction residual for region i, which is the difference between the number of cases predicted by the model for region i and the actual number of cases monitored. Yi ( t ) represents the actual number of monitored cases in region i. κw This represents the learning rate of the crowd flow weights; it is a hyperparameter used to control the magnitude of weight adjustments, and its value is usually determined based on experience or through methods such as cross-validation. ϵ It is a very small positive number, used to prevent the actual number of cases from being inflated. Yi ( t The formula improves robustness by preventing division by zero when the weight is zero. This formula allows the model to adaptively adjust the weights of population movement between regions based on actual prediction performance, making it more accurately reflect the true dynamics of population transmission.

[0059] In some possible embodiments, the dynamic correction of the local effective contact rate in the target infectious disease dynamics model aims to ensure that the model can reflect the true intensity of disease transmission within the region in real time. The local effective contact rate βi(t) is one of the core parameters in the infectious disease dynamics model; it quantifies the frequency and probability of effective transmission between susceptible individuals and infected individuals within region i. Due to the dynamic changes in various factors such as population behavior patterns, environmental factors, and public health interventions, the local effective contact rate is not constant; therefore, dynamic correction is crucial to maintaining the accuracy of model predictions. This includes: The contact rate adjustment index is calculated based on the ratio of the predicted residuals to the actual number of cases; the predicted residuals refer to the ratio of the number of cases Yi(t) predicted by the model to the actual number of cases monitored. Yi ( t The difference between ) is Yi ( t )- Yi ( t This residual is compared with the actual number of cases. Yi ( t Using the ratio as the basis for adjustment, a relative error metric can be obtained, reflecting the magnitude of the prediction bias relative to the current scale of the epidemic. For example, when the model significantly underestimates the epidemic (predicting far fewer cases than the actual number), the ratio will be positive and large, indicating a need to significantly increase the contact rate; conversely, when the model overestimates the epidemic, the ratio will be negative, indicating a need to reduce the contact rate. To avoid a zero denominator, a very small positive number is introduced when calculating the ratio. ϵ This ensures the stability of the calculation.

[0060] Based on the contact rate adjustment index, the original local effective contact rate is updated to obtain the updated local effective contact rate. The update formula for the local effective contact rate is: βinew ( t )= βiold ( t ) exp ( κβYi ( t )+ ϵYi ( t )- Yi ( t )); in, βinew ( t () represents the updated local effective contact rate for region i. βiold ( t This represents the local effective contact rate of region i before the update. Yi ( t) represents the actual number of monitored cases in region i. Yi ( t ) represents the predicted number of cases in region i, which allows the correction magnitude of the contact rate to change exponentially with the increase of the prediction error, thus enabling faster convergence when the error is large and fine-tuning when the error is small; κβ This represents the contact rate update step size; it is an adjustable hyperparameter used to control the sensitivity and speed of each update. A larger value indicates a larger step size. κβ A higher value means the model responds more quickly to prediction errors, while a lower value means the model responds more quickly to prediction errors. κβ The value makes the model update more stable.

[0061] The updated local effective contact rate meets the preset parameter boundary constraints: βi min≤ βinew ( t )≤ β i These boundary constraints define a reasonable range of values ​​for the local effective contact rate. For example, βimin can be set to a positive value close to zero to avoid negative or zero contact rates. βi The min parameter can be set as an upper limit based on historical data or expert experience to prevent the contact rate from being revised to an unrealistically high value, thereby ensuring that the model remains biologically and epidemiologically sound during dynamic adjustments.

[0062] In some possible embodiments, the dynamic correction of the risk warning threshold of the target infectious disease dynamics model includes: Statistically analyze the number of false positives and false negatives in each region within the most recent validation window; Based on the number of false positives, the number of false negatives, and the business loss weights for false positives and false negatives, the threshold adjustment amount is calculated. The original risk warning threshold is updated based on the threshold adjustment amount to obtain the updated risk warning threshold. The update formula for the risk warning threshold is: ρinew = ρiold + κρ ( cFPFPi - cFNFNi ); in, ρinew This represents the updated risk warning threshold for region i. ρiold This represents the risk warning threshold for region i before the update. FPi This represents the number of false positives in region i within the most recent validation window. FNi This represents the number of false negatives in region i within the most recent validation window. cFP Indicates the business loss weighting for false positives.cFN The weight of business losses indicated by false negatives. κρ This indicates the threshold adjustment step size.

[0063] As one implementation method, the number of false positives and false negatives in each region within the most recent validation window is counted to quantify the actual warning performance of the early warning model within a specific time period. (False positive count) FPi This refers to a situation where the model issues a warning, but the actual epidemic situation in region i does not reach the warning threshold. This could lead to unnecessary resource investment and a "crying wolf" effect. (Number of false negatives) FNi This refers to a situation where the actual epidemic situation in region i reaches the warning threshold, but the model fails to issue a warning, which may lead to the spread of the epidemic due to a lack of timely intervention. These key indicators can be accurately calculated by comparing the warning results issued by the model within the most recent preset "validation window" (e.g., the past week, month, or a complete epidemic cycle) with the actual epidemic monitoring data during the same period. For example, a warning threshold can be set; when the number of cases predicted by the model exceeds this threshold, the system issues a warning. Subsequently, the accuracy of the warning is determined by checking whether the actual number of cases during the same period also exceeds the corresponding actual epidemic threshold. If the model issues a warning but the actual number does not exceed the threshold, it is counted as a false positive; if the model does not issue a warning but the actual number exceeds the threshold, it is counted as a false negative. This validation window can use a sliding window mechanism to ensure that the model is always evaluated and adjusted based on the latest operational performance data.

[0064] Based on this, and based on the number of false positives... FPI The number of false negatives FNi And the business loss weighting for false positives and false negatives cFP , cFN The next step is to calculate the threshold adjustment. This step combines the statistically recorded number of false positives / false negatives with the actual business impact (i.e., business loss weight) to quantify the direction and magnitude of the necessary adjustment to the warning threshold. Different business scenarios have different tolerances for false positives and false negatives. For example, for infectious diseases with high mortality rates or rapid transmission speeds, the business loss weight of false negatives is... cFN The threshold is typically set very high to prioritize avoiding underreporting; however, for areas with limited public health resources, the weighting for business losses from false positives is lower. cFP The weighting for business losses from false positives may also be higher to avoid wasting resources. cFP Weighting of business losses from false negatives cFN These are pre-set parameters that reflect the business's level of concern regarding the negative impact of these two error types. Through calculation... cFPFPi (Total loss due to false positives in region i) and cFNFNiThe difference between (the total loss caused by false negatives in region i) can provide a measure indicating the direction and magnitude of threshold adjustment. If the loss caused by false positives is much greater than the loss caused by false negatives, it indicates that the current warning threshold may be set too low and needs to be adjusted upwards; conversely, if the loss caused by false negatives is greater, it may mean that the threshold is too high and needs to be adjusted downwards.

[0065] Subsequently, the original risk warning threshold is adjusted according to the threshold adjustment amount. old Update the data to obtain the updated risk warning threshold. new This is an iterative update mechanism, with new risk warning thresholds. new It is at the old threshold old This is obtained by adding a term proportional to the threshold adjustment amount to the base value. Among them, kr (Threshold adjustment step size) is a parameter that controls the magnitude of each adjustment, designed to prevent over-adjustment from causing instability in the model's early warning performance. Each time parameter tuning is triggered, the system retrieves the old risk early warning threshold for the current region i. old And combined with the number of false positives in the most recent validation window kr Number of false negatives FNi and the preset false positive business loss weight. cFP Weight of false negative business losses cFN and threshold adjustment step size kr Substitute into the update formula new = old + kr ( cFPFPi - cFNFNi A new risk warning threshold ρinew is calculated. This new threshold will be used for subsequent warning judgments, thereby achieving dynamic adaptive adjustment of the warning threshold.

[0066] It should be noted that the above description describes some embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than that shown in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also proposes a device for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement, such as... Figure 2 As shown.

[0068] The acquisition module 100 is configured to acquire epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data. The construction module 200 is configured to construct a cross-regional coupled transmission model based on the target infectious disease dynamics model, using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data, to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window; wherein, the target infectious disease dynamics model is used to indicate the state transition relationship between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions, the multiple regions have different disease transmission risks, and the multiple regions are coupled through a population flow matrix; The optimization module 300 is configured to dynamically optimize the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, to obtain the optimized infectious disease early warning model. The prediction module 400 is configured to use the optimized infectious disease early warning model to predict the epidemic trend of infectious diseases in the multiple regions, output the early warning results, and generate a business application guide.

[0069] The apparatus described above is used to implement the corresponding methods in the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0070] Based on the same inventive concept, embodiments of this application also provide an electronic device, including: processor; Memory for storing the executable instructions of the processor; The processor is configured to execute the aforementioned method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement by executing the executable instructions.

[0071] Based on the same inventive concept, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, a method for dynamic optimization and operational verification of an infectious disease early warning model for areas with frequent cross-regional population movement is implemented.

[0072] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for dynamic optimization and operational validation of an infectious disease early warning model in areas with frequent cross-regional population movement, the method comprising: Acquire epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data; Based on the target infectious disease dynamics model, a cross-regional coupled transmission model is constructed using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window. The target infectious disease dynamics model is used to indicate the state transition relationships between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions. These multiple regions have different disease transmission risks, and a coupling relationship is established between them through a population flow matrix. Based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, the parameters of the target infectious disease dynamics model are dynamically optimized to obtain the optimized infectious disease early warning model. Using the optimized infectious disease early warning model, the epidemic trend of infectious diseases in the multiple regions is predicted, early warning results are output, and business application guidelines are generated.

2. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 1, characterized in that, The aforementioned model, based on the dynamics of the target infectious disease, utilizes the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environmental data, and intervention measure data to construct a cross-regional coupled transmission model, including: Based on the target infectious disease dynamics model, the epidemic state variables for each of the multiple regions are determined, including the size of susceptible individuals, the size of latent individuals, the size of infected individuals, and the size of removed individuals. Based on the population flow data, an inter-regional population flow matrix is ​​constructed, where the elements of the population flow matrix represent the population size or normalized flow intensity flowing from the first region to the second region. Based on the inter-regional population flow matrix, standardized flow weights are calculated, and external input infection pressure is calculated based on the standardized flow weights. The overall infection pressure is calculated based on the external input infection pressure, local transmission risk, and regional heterogeneity risk factors. Based on the overall infection pressure, a set of regional heterogeneous transmission differential equations with population mobility coupling terms is established. The set of transmission differential equations includes equations for the rate of change of susceptible individuals, the rate of change of latent individuals, the rate of change of infected individuals, and the rate of change of removed individuals.

3. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 1, characterized in that, The dynamic optimization of the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, includes: Based on the predicted number of cases and the actual number of monitored cases, the point prediction error index, peak prediction deviation index, early warning advance index, false positive rate index, and false negative rate index are calculated. Based on the business feedback data, extract the business confirmation results, intervention measures information, false positive labels, false negative labels, and expert consensus scores; Based on the point prediction error index, the peak prediction deviation index, the early warning advance index, the false positive rate index, the false negative rate index, and the business feedback data, determine whether the parameter tuning triggering conditions are met. Under the condition that the parameter tuning triggering condition is met, the population flow weight, local effective contact rate, external input risk amplification coefficient, and risk warning threshold of the target infectious disease dynamics model are dynamically corrected based on the verification dataset and business feedback dataset within the rolling time window, so as to obtain the tuned model parameter set.

4. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 3, characterized in that, The point prediction error indicators include mean absolute error, root mean square error, and mean absolute percentage error. The peak prediction deviation index includes peak time deviation and peak intensity deviation; The parameter tuning triggering conditions include at least one of the following: the overall prediction error exceeds a preset error tolerance threshold, the peak time deviation exceeds a preset peak deviation threshold, the number of false negatives exceeds a preset false negative tolerance threshold, the number of false positives exceeds a preset false positive tolerance threshold, and the number of continuous abnormal cycles reaches a preset cycle threshold.

5. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 3, characterized in that, The dynamic correction of the population movement weights in the target infectious disease dynamics model includes: The population mobility weight adjustment coefficient is calculated based on the ratio of the predicted residual to the actual number of cases. The original crowd flow weights are updated based on the crowd flow weight adjustment coefficients to obtain the updated crowd flow weights. The updated crowd flow weights are normalized to obtain normalized crowd flow weights.

6. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 3, characterized in that, The dynamic correction of the local effective contact rate of the target infectious disease dynamics model includes: The contact rate adjustment index is calculated based on the ratio of the predicted residuals to the actual number of cases. Based on the contact rate adjustment index, the original local effective contact rate is updated to obtain the updated local effective contact rate. The updated local effective contact rate meets the preset parameter boundary constraints.

7. The method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in claim 3, characterized in that, The dynamic correction of the risk warning threshold of the target infectious disease dynamics model includes: Statistically analyze the number of false positives and false negatives in each region within the most recent validation window; Based on the number of false positives, the number of false negatives, and the business loss weights for false positives and false negatives, the threshold adjustment amount is calculated. The original risk warning threshold is updated based on the threshold adjustment amount to obtain the updated risk warning threshold.

8. A device for dynamic optimization and operational verification of an infectious disease early warning model in areas with frequent cross-regional population movement, the device comprising: The acquisition module is configured to acquire epidemic monitoring data, population flow data, regional heterogeneity data, meteorological and environmental data, intervention measure data, and business feedback data; The construction module is configured to build a cross-regional coupled transmission model based on the target infectious disease dynamics model, using the epidemic monitoring data, population flow data, regional heterogeneity data, meteorological environment data, and intervention measure data, to obtain the predicted number of cases and comprehensive transmission risk index for each region within a predetermined time window; wherein, the target infectious disease dynamics model is used to indicate the state transition relationship between susceptible individuals, latent individuals, infected individuals, and removed individuals in multiple regions, the multiple regions have different disease transmission risks, and the multiple regions are coupled through a population flow matrix; The optimization module is configured to dynamically optimize the parameters of the target infectious disease dynamics model based on the predicted number of cases and the comprehensive transmission risk index, combined with the business feedback data, to obtain the optimized infectious disease early warning model. The prediction module is configured to use the optimized infectious disease early warning model to predict the epidemic trend of infectious diseases in the multiple regions, output the early warning results, and generate business application guidelines.

9. An electronic device, characterized in that, include: processor; Memory for storing the executable instructions of the processor; The processor is configured to execute the method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement, as described in any one of claims 1-8, by executing the executable instructions.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the method for dynamic optimization and operational verification of infectious disease early warning models in areas with frequent cross-regional population movement as described in any one of claims 1-8.

Images