A seasonal influenza epidemic risk early warning method and system

Abstract

The present application relates to the field of artificial intelligence technology, and discloses a seasonal influenza epidemic risk warning method and system, which includes: obtaining multi-dimensional influencing factors of seasonal influenza; determining the temporal composition ratio of antigen clusters of at least one virus subtype through pathogen sequence information of virus subtypes, and determining the infection rate of at least one virus subtype through case data of seasonal influenza; predicting the predicted infection rate sequence of at least one virus subtype in the first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; and weighted summing the predicted infection rate sequence through the warning contribution of multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype. The accuracy of the epidemic risk warning of seasonal influenza is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and system for early warning of epidemic risks of seasonal influenza. Background Art

[0002] Seasonal influenza has caused serious harm to human health, not only causing widespread seasonal outbreaks, but also potentially triggering global pandemics. Given the diversity of influenza virus subtypes and the seasonal characteristics of their epidemic trends, it is particularly important to build a set of accurate and forward-looking technical solutions for influenza epidemic risk prediction and early warning. This can not only provide guidance for seasonal influenza prevention measures for specific subtypes, but also help alleviate the health risks caused by large-scale infections.

[0003] However, there are still some limitations in the field of seasonal influenza forecasting and early warning technology. The existing technical framework often fails to fully capture the key factors that affect influenza infection, resulting in deviations in the early warning of seasonal influenza epidemic risks. Summary of the invention

[0004] The present application provides a seasonal influenza epidemic risk early warning method and system to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0005] A seasonal influenza epidemic risk warning method according to a technical solution of one aspect of the present application comprises the following steps: obtaining multidimensional influencing factors of seasonal influenza, wherein the multidimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; determining the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determining the infection rate of at least one virus subtype through the case data of seasonal influenza; predicting a predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; and weighted summing the predicted infection rate sequence according to the warning contribution of the multidimensional influencing factors to determine risk warning information of at least one virus subtype.

[0006] Some technical solutions of the present application, the method of determining the temporal composition ratio of antigen clusters of at least one viral subtype through the pathogen sequence information of the viral subtype, includes: determining the protein sequence mutation matrix through the antigen Euclidean distance between the full-fragment protein sequence data of at least one viral subtype; clustering the protein sequence mutation matrix according to the Markov clustering algorithm to generate multiple antigen clusters; and determining the temporal composition ratio of antigen clusters of at least one viral subtype based on the multiple antigen clusters.

[0007] Some technical solutions of the present application, the protein sequence mutation matrix is ​​clustered according to the Markov clustering algorithm to generate multiple antigen clusters, including: determining multiple antigen clusters through the formula: C = expand(M,2)⊙inflate(M,r); wherein C is a set of antigen clusters, expand and inflate are operations in the Markov clustering algorithm, M is the protein sequence mutation matrix, and r is the expansion parameter.

[0008] Some technical solutions of the present application, the time series composition ratio of antigen clusters of at least one virus subtype is determined based on multiple antigen clusters, including: determining the time series proportion of the corresponding antigen cluster by the ratio of any antigen cluster to the sum of all generated antigen clusters; and obtaining the time series composition ratio of antigen clusters of at least one virus subtype by resampling the time series proportions.

[0009] In some technical solutions of the present application, the infection rate of at least one virus subtype is determined through the case data of seasonal influenza, including: obtaining the infection rate of the corresponding virus subtype by multiplying the ratio of the number of seasonal influenza-like medical records to the number of visits within a second preset time period, the ratio of the number of positive cases to the amount of laboratory samples, and the proportion of any virus subtype in the number of positive cases.

[0010] Some technical solutions of the present application, based on the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics, predicting the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model, include: determining a time series data set, the time series data set includes the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of the corresponding virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; dividing the time series data set into a test set and a training set according to the epidemic time information of the seasonal influenza; inputting the training set into a Transformer regression model, and performing the Transformer regression model on the predicted infection rate sequence of at least one virus subtype in a first preset time period; The parameters are optimized to obtain target parameters; based on the test set, multiple rounds of retrospective predictions are performed on the Transformer regression model according to the epidemic time information, and the prediction performance of the Transformer regression model is evaluated to obtain a retrospective prediction evaluation of at least one virus subtype at the epidemic time information scale; the target parameters are used as the initial parameters of the Transformer prediction model, and based on a multi-step forward-looking prediction strategy of stepwise approximation, a predicted infection rate sequence of at least one virus subtype in a first preset time period is generated; wherein, evaluating the prediction performance of the Transformer regression model includes evaluating the prediction performance of the Transformer regression model through mean absolute error, root mean square error, and determination coefficient.

[0011] Some technical solutions of the present application, optimizing the parameters in the Transformer regression model to obtain the target parameters, include: optimizing the fitting error of the Transformer regression model based on a loss function with a regularization term added; updating the parameters in the Transformer regression model using a gradient descent method to obtain the target parameters; evaluating the predictive performance of the Transformer regression model, also includes: using a time series-based cross-validation strategy to evaluate the predictive performance of the Transformer regression model according to the popular time information, and optimizing the parameters in the Transformer regression model again according to the evaluation results of the predictive performance to obtain the target parameters.

[0012] In some technical solutions of the present application, the seasonal influenza epidemic risk warning method further includes: updating the time series data set according to a preset time period; analyzing the error accumulation trend to obtain an analysis result of the error accumulation trend; and according to the analysis result of the error accumulation trend, calculating by the formula: Adjust the parameters of the Transformer prediction model; where α is the learning rate, is the gradient calculated based on the error ∈, Θ is the model parameter, Θ update are the adjusted model parameters.

[0013] In some technical solutions of the present application, the predicted infection rate sequence is weighted and summed by the warning contribution of the multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype, including: quantifying the warning contribution of each of the multi-dimensional influencing factors; by the formula: Determine an early warning index for at least one virus subtype; wherein I t represents the early warning index, φ v represents the contribution of the vth factor to the change in model output, Represents the infection rate prediction value of the Transformer prediction model at the t+1 time point; determines the epidemic risk level of at least one virus subtype based on the statistical data of the warning index; wherein the risk warning information includes the warning index and the epidemic risk level.

[0014] Another aspect of the present application is a system of technical solutions, the system comprising: an acquisition module, used to acquire multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; a first determination module, used to determine the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determine the infection rate of at least one virus subtype through the case data of seasonal influenza; a prediction module, used to predict the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; a second determination module, used to perform weighted summation of the predicted infection rate sequence through the warning contribution of the multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype.

[0015] This application has at least the following beneficial effects:

[0016] The seasonal influenza epidemic risk warning method of the present application obtains multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; determines the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determines the infection rate of at least one virus subtype through the case data of seasonal influenza; predicts the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; and performs weighted summation of the predicted infection rate sequence through the warning contribution of the multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype. Since the time series composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics are introduced into the deep learning model to predict the infection rate sequence, and further consideration is given to the warning contribution of multi-dimensional influencing factors to perform weighted summation on the predicted infection rate sequence to finally obtain the risk warning information of at least one virus subtype, this can comprehensively capture the key factors affecting influenza infection and improve the accuracy of seasonal influenza epidemic risk warning.

[0017] It is understandable that the beneficial effects of the system disclosed in the present application are the same as the beneficial effects of the seasonal influenza epidemic risk early warning method, which will not be repeated here.

[0018] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0020] Figure 1 A schematic diagram of a process flow of a seasonal influenza epidemic risk early warning method according to an embodiment of the present application;

[0021] Figure 2 Another schematic diagram of the process of the seasonal influenza epidemic risk early warning method according to the embodiment of the present application;

[0022] Figure 3 This is a schematic block diagram of the modules of the seasonal influenza epidemic risk early warning system according to an embodiment of the present application;

[0023] Figure 4 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the reference to "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The appearance of this phrase in various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0025] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0026] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0028] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0029] In order to make the inventive concept of the present application easy to understand, before describing the embodiments of the present application in detail, the English abbreviations (terms) / related concepts involved in the embodiments of the present application are first described. The English abbreviations (terms) / related concepts involved in the embodiments of the present application are subject to the following explanations.

[0030] Virus subtype: A virus subtype refers to a branch with specific genetic characteristics formed due to genetic variation within the same virus species. These subtypes usually differ in the genetic sequence of the virus, which may affect the virus's ability to spread, pathogenicity, and immune escape. In influenza viruses, the classification of subtypes is mainly based on the differences in the amino acid sequences of the hemagglutinin (HA) and neuraminidase (NA) proteins on the surface of the virus.

[0031] Antigenic cluster: An antigenic cluster refers to a group of molecules with similar antigenicity. In immunology, antigens refer to molecules that can induce an immune response in the body, while antigenic clusters refer to groups of molecules with similar structures and antigenicity. These molecules may come from the same pathogen or partial structures of different pathogens, but they can be recognized by the body's immune system and stimulate an immune response.

[0032] Cross-validation: Cross-validation is a statistical analysis method that is mainly used to evaluate and verify the performance of machine learning models. It divides the data set into multiple parts, uses part of the data to train the model, and then uses the remaining part of the data to verify the performance of the model. This process is repeated many times, and different data sets are used for training and verification each time. The main purpose of cross-validation is to estimate the performance of the model more accurately and avoid overfitting and underfitting problems. Through multiple verifications, a more reliable average performance indicator can be obtained, thereby better evaluating the generalization ability of the model.

[0033] Transformer forecasting model with multi-step autoregression: The Transformer forecasting model with multi-step autoregression is a forecasting model that combines the Multistep Autoregressive and Transformer architectures. This model is often used for time series forecasting, especially when data for multiple time points in the future need to be forecasted. In time series analysis, the Multistep Autoregressive model (MSAR) is a common model that predicts future values ​​by considering the forward information of the time series. MSAR models can handle forecasts for multiple time points, but traditional AR (autoregressive) models may not be able to capture long-term dependencies and complex patterns in time series. The Transformer model, originally used for natural language processing tasks, has gradually been applied to time series forecasting due to its powerful sequence modeling capabilities. Transformer captures the dependencies between different positions in the sequence through the self-attention mechanism, which enables it to capture long-term dependencies in the time series and perform well even in long sequences. Applying the Transformer to the Multistep Autoregressive model creates a model that can handle multi-step forecasts and capture long-term dependencies. This model is often called the Transformer-based MSAR model. In this model, the Transformer architecture is used to learn an efficient representation of time series data, and the nature of multi-step autoregression enables the model to consider information from multiple future time points simultaneously to predict the value at the current time point.

[0034] SHAP: SHAP (SHapley Additive exPlanations) is a popular method for explaining predictions of machine learning models, especially when using gradient-based models such as deep learning. SHAP is based on the Shapley value in game theory, providing a quantitative and consistent explanation of the contribution of each feature in the model to the prediction. In machine learning, Shapley values ​​are used to fairly distribute the contribution of predictions, that is, how much each feature contributes to the prediction.

[0035] Seasonal influenza has caused serious harm to human health, not only causing widespread seasonal outbreaks, but also potentially triggering global pandemics. Given the diversity of influenza virus subtypes and the seasonal characteristics of their epidemic trends, it is particularly important to build a set of accurate and forward-looking technical solutions for influenza epidemic risk prediction and early warning. This can not only provide guidance for seasonal influenza prevention measures for specific subtypes, but also help alleviate the health risks caused by large-scale infections.

[0036] However, there are still some limitations in the field of seasonal influenza prediction and early warning technology. First, the existing technical framework often ignores the differences between influenza virus subtypes and lacks in-depth identification of their epidemic characteristics. Secondly, most studies have not fully utilized epidemiological time series big data, making it difficult to fully capture the key factors affecting influenza infection. Furthermore, related technologies focus more on retrospective analysis and evaluation and lack forward-looking prediction capabilities for the future. Finally, few studies have proposed a systematic influenza early warning program, which limits the practicality of prediction and early warning results.

[0037] In view of this, the present application proposes a seasonal influenza epidemic risk warning method and system, which obtains multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; determines the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determines the infection rate of at least one virus subtype through the case data of seasonal influenza; predicts the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; and performs weighted summation of the predicted infection rate sequence through the warning contribution of the multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype. Since the time series composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics are introduced into the deep learning model to predict the infection rate sequence, and further consideration is given to the warning contribution of multi-dimensional influencing factors to perform weighted summation on the predicted infection rate sequence to finally obtain the risk warning information of at least one virus subtype, this can comprehensively capture the key factors affecting influenza infection and improve the accuracy of seasonal influenza epidemic risk warning.

[0038] The method provided in the embodiment of the present application can be applied to the electronic device provided in the embodiment of the present application, wherein the electronic device can be a terminal or a server.

[0039] The terminal may be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto.

[0040] The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, as well as big data and artificial intelligence platforms.

[0041] The above is an introduction to the implementation environment that may be involved in the technical solution of this application.

[0042] The following will describe in detail the steps of implementing a seasonal influenza epidemic risk warning method provided by an embodiment of the present application in conjunction with the accompanying drawings.

[0043] Please refer to Figure 1 , Figure 1 The flowchart is a schematic diagram of the seasonal influenza epidemic risk warning method of the embodiment of the present application. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0044] The method of the embodiment of the present application comprises the following steps:

[0045] Step 1: First, the multi-dimensional influencing factors of seasonal influenza can be obtained, and the multi-dimensional influencing factors can include pathogen sequence information of at least one virus subtype and case data of seasonal influenza;

[0046] Step 2: Secondly, the temporal composition ratio of antigen clusters of at least one virus subtype can be determined through the pathogen sequence information of the virus subtype, and the infection rate of at least one virus subtype can also be determined through the case data of seasonal influenza;

[0047] Step 3: Again, the predicted infection rate sequence of at least one virus subtype in the first preset time period can be predicted by a deep learning model based on the temporal composition ratio of the antigen clusters of at least one virus subtype, population migration characteristics, the infection rate of at least one virus subtype, the intensity of non-drug intervention measures and climate characteristics;

[0048] Step 4: Finally, the predicted infection rate sequence can be weighted and summed by the warning contribution of multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype.

[0049] In the steps 1 to 4 shown in the embodiment of the present application, the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, the population migration characteristics, the intensity of non-drug intervention measures and the climate characteristics are input into the deep learning model to predict the predicted infection rate sequence of at least one virus subtype in the first preset time period; and the predicted infection rate sequence is weighted and summed by the early warning contribution of multi-dimensional influencing factors to determine the risk early warning information of at least one virus subtype. Since the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, the population migration characteristics, the intensity of non-drug intervention measures and the climate characteristics are introduced into the deep learning model to predict the infection rate sequence, it is further considered that the predicted infection rate sequence is weighted and summed by the early warning contribution of multi-dimensional influencing factors to finally obtain the risk early warning information of at least one virus subtype, thereby being able to fully capture the key factors affecting influenza infection and improve the accuracy of the epidemic risk early warning of seasonal influenza.

[0050] The specific implementation methods of the above steps are introduced below.

[0051] In step 1, multi-dimensional influencing factors of seasonal influenza may be obtained, and the multi-dimensional influencing factors may include pathogen sequence information of at least one virus subtype and case data of seasonal influenza.

[0052] Seasonal influenza, also known as seasonal flu, is a respiratory infectious disease caused by influenza virus. This virus causes periodic epidemics in a specific season every year. Seasonal influenza has obvious seasonal characteristics.

[0053] Multidimensional influencing factors may include time series data such as the number of cases, air and rail passenger traffic, climate change, the intensity of non-drug intervention measures, and the full-fragment protein sequence of pathogens.

[0054] For example, time series data such as the number of cases of four seasonal influenza subtypes (A / H1N1, A / H3N2, B / Victoria and B / Yamagata) in the influenza surveillance system of the study area, air and rail passenger population flow, climate change, the intensity of implementation of non-drug intervention measures, and full-fragment protein sequences of pathogens can be collected as multi-dimensional influencing factors of seasonal influenza.

[0055] The embodiment of the present application provides a data basis for subsequent infection rate prediction by obtaining multi-dimensional influencing factors of seasonal influenza.

[0056] In step 2, the temporal composition ratio of antigen clusters of at least one virus subtype can be determined through the pathogen sequence information of the virus subtype, and the infection rate of at least one virus subtype can be determined through the case data of seasonal influenza. The case data can include the number of hospital visits, the number of seasonal influenza-like medical records, the number of test samples, and the number of positive medical records of seasonal influenza.

[0057] In some embodiments, the mutation matrix of the protein sequence can be determined by the antigenic Euclidean distance between the full-fragment protein sequence data of at least one viral subtype.

[0058] For example, the full-segment protein sequence data of each subtype of seasonal influenza in the study area can be sorted in chronological order, and the genetic sequence is used as a template sequence to compare with the sorted full-segment protein sequences of each subtype of seasonal influenza, the change characteristics of the amino acid sites of all sequences are extracted, and the antigenic Euclidean distance D(i, j) between each sequence is calculated. D(i, j) can be calculated by formula (1).

[0059]

[0060] Where D(i,j) is the antigenic Euclidean distance between sequences s[i] and s[j], and a ik is the value of the kth amino acid site in sequence s[i] (i.e., the variation characteristics of the amino acid site), a jk It is the value of the kth amino acid position of the sequence s[j] (i.e., the variation characteristic of the amino acid position) k represents the amino acid position. It should be understood that for one influenza subtype, it may include multiple full-fragment protein sequences.

[0061] After calculating the antigenic Euclidean distances between the sequences, an N×N mutation matrix M (i.e., the mutation matrix of the protein sequence) can be constructed based on the calculated antigenic Euclidean distances between the sequences, where N is the number of sequences. ij =D(i,j), that is, element M ij is the antigenic Euclidean distance between sequences s[i] and s[j].

[0062] In some embodiments, the protein sequence mutation matrix is ​​clustered according to the method in the Markov clustering algorithm to generate multiple antigen clusters.

[0063] For example, the Markov clustering algorithm can be applied to cluster the protein sequence mutation matrix to generate multiple antigen clusters, which can be calculated by formula (2):

[0064] C=expand(M,2)⊙inflate(M,r) (2);

[0065] Multiple antigenic clusters were identified.

[0066] Among them, C is the set of antigen clusters, which can be expressed as: C(c1,c2,……,c n ); expand and inflate are operations in the Markov clustering algorithm, M is the protein sequence mutation matrix, and r is the expansion parameter.

[0067] In the Markov clustering algorithm (MCL), "expand" and "inflate" are two steps to operate on node data (i.e., protein sequence mutation matrix), which are mainly used to simulate random walks and introduce additional connections.

[0068] Expand: This step usually involves expanding the node so that new connections are created. In MCL, this is done by considering the degree of the node (the number of connections it has with other nodes).

[0069] Specifically, the algorithm might check the degree of a node and decide whether to connect the node to other nodes based on that degree.

[0070] Inflate: This step is to expand the existing connections to increase the connectivity between nodes. In MCL, this process may involve increasing the weight of existing edges or introducing new edges to strengthen the connection between certain nodes.

[0071] The "expand" and "inflate" operations are usually performed alternately until a certain stable state is reached or a specific stopping condition is met.

[0072] It should be noted that the specific implementation of the "expand" and "inflate" operations in the MCL algorithm may vary depending on the version and variant of the algorithm. In different application scenarios and network structures, the details of these operations may need to be adjusted according to actual conditions to obtain the best clustering effect.

[0073] In some embodiments, the temporal composition ratio of antigen clusters of at least one viral subtype can be determined based on multiple antigen clusters.

[0074] The temporal proportion of the corresponding antigen cluster can be determined by the ratio of any antigen cluster to the sum of all generated antigen clusters.

[0075] For example, it can be calculated by formula (3):

[0076]

[0077] The time series proportion of each antigen cluster in the matrix (i.e., the time series proportion of antigen clusters of at least one viral subtype) was calculated.

[0078] In formula (3), B(c i) indicates antigen cluster c i The time series proportion, |C| is the number of sequences contained in the antigen cluster set C, ∑|C| is the sum of the sizes of all antigen clusters, |c i | is antigen cluster c i sequence.

[0079] By resampling the time series proportions, the time series composition ratio of the antigen cluster of at least one viral subtype is obtained.

[0080] It should be understood that the temporal proportion of antigen clusters of virus subtypes is usually calculated based on the collected data, but the collected data is often not evenly distributed over time. It is possible that enough data will be collected within a month to determine the temporal proportion of antigen clusters of the corresponding virus subtypes, and the data collected in the following months is often insufficient to determine the temporal proportion of antigen clusters of the corresponding virus subtypes. Therefore, it is necessary to resample the determined temporal proportions to obtain the temporal composition ratio of antigen clusters of virus subtypes that can be processed by the deep learning model. It should be understood that resampling can be performed by means in the prior art, such as self-resampling and cross-resampling, and this application does not limit this.

[0081] For example, the timing ratio B(c i ) to obtain the temporal composition ratio of antigen clusters of virus subtype K Afterwards, As a research factor of seasonal influenza at the molecular evolution level. Here, t represents the time information included in the temporal composition ratio of the antigen cluster, sub K Represents virus subtype K. For example, K may be any subtype of A / H1N1, A / H3N2, B / Victoria and B / Yamagata.

[0082] In some embodiments, the infection rate of the corresponding virus subtype can be obtained by multiplying the ratio of the number of seasonal influenza-like medical records to the number of visits in the second preset time period, the ratio of the number of positive cases to the number of test samples, and the proportion of any virus subtype in the number of positive cases.

[0083] Exemplarily, the influenza infection rate of each subtype can be calculated based on case data reported by the influenza surveillance system.

[0084] For example, through formula (4):

[0085]

[0086] Calculate the infection rate of each influenza subtype

[0087] In formula (4), Patient visit represents the number of patients visiting the sentinel hospital per week, ILI represents the number of cases of influenza-like illness, Samples represents the number of samples collected for testing, Flu represents the number of cases confirmed as seasonal influenza positive by testing (including cases of four influenza subtypes), SP K It is the proportion of a specific subtype K (K is any of A / H1N1, A / H3N2, B / Victoria and B / Yamagata) in all positive cases. The infection rate can be statistically measured on a weekly basis.

[0088] The embodiment of the present application determines the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determines the infection rate of at least one virus subtype through the case data of seasonal influenza, which provides data support for the prediction of the infection rate sequence through the deep model. Since the data involved in the prediction include the temporal composition ratio of antigen clusters of at least one virus subtype and the infection rate of at least one virus subtype, the accuracy and pertinence of the prediction results are improved.

[0089] In step 3, the predicted infection rate sequence of at least one viral subtype in the first preset time period can be predicted through a deep learning model based on the temporal composition ratio of antigen clusters of at least one viral subtype, population migration characteristics, the infection rate of at least one viral subtype, the intensity of non-drug intervention measures and climate characteristics.

[0090] The deep learning model can be a Transformer model, that is, a model based on the self-attention mechanism, or a recurrent neural network model, or a long short-term memory network model. This application does not limit the type of deep learning model.

[0091] The first preset time period may be a week, a month or a quarter.

[0092] In some embodiments, it is first necessary to determine a time series data set, which can be determined by the time series composition ratio of antigen clusters of at least one viral subtype, population migration characteristics, infection rates of corresponding viral subtypes, the intensity of non-drug intervention measures, and climate characteristics.

[0093] Exemplarily, the acquired time series data of multi-dimensional influencing factors of seasonal influenza may be integrated to determine a time series data set.

[0094] For example, the population migration characteristics P in time period t can be m (t) (m represents different migration routes, such as aviation, railways, etc.); the intensity of non-drug intervention measures implemented in time period t NPI(t); climate characteristics E in time period t n(t) (n represents different climate factors, such as temperature and humidity) and the temporal composition ratio of antigen clusters of virus subtype K in time period t These data are preprocessed, such as data cleaning and missing value filling, to ensure the quality and integrity of the data. The scale of all time series data is taken as "week" as the statistical scale using the resampling method to form a standardized time series data set D, where: is the infection rate of virus subtype K calculated in the above example.

[0095] In some embodiments, the time series data set may be divided into a test set and a training set according to the epidemic time information of seasonal influenza, wherein the epidemic time information may be the epidemic season of seasonal influenza.

[0096] For example, in order to evaluate the prediction ability of the deep learning model and ensure its generalization performance, the time series dataset D can be divided into independent training and test sets: D = {D train ,D test The division can be based on the understanding of influenza seasonality, and the influenza season is defined as from the 40th week of each year to the 14th week of the following year. A historical time period T = [t0, t l ], and the first three influenza epidemic seasons were The time series are used as the test set, and the remaining time series are used as the training set.

[0097] This process can be expressed as formula (5):

[0098]

[0099] In formula (5), D train represents the training set, D test represents the test set, represents the pth influenza epidemic season, and l represents the variable taking the value of the influenza epidemic season.

[0100] In some implementations, the training set may be input into a Transformer regression model, and then the parameters in the Transformer regression model may be optimized to obtain target parameters.

[0101] Optionally, the fitting error of the Transformer regression model can be optimized based on a loss function with a regularization term added; the parameters in the Transformer regression model are updated using a gradient descent method to obtain target parameters.

[0102] For example, first, a Transformer regression model is constructed, and then the training set D in the above example is trainInput to the Transformer regression model, and use the Encoder-Decoder structure in the model to extract the corresponding data features and perform sequence-to-sequence prediction. The simplified input sequence is represented as x, and the output sequence is represented as Then the Transformer regression model f can be expressed as formula (6):

[0103]

[0104] In formula (6), Θ is the set of all trainable parameters of the Transformer regression model.

[0105] Afterwards, a loss function including a regularization term is used Optimize the fitting error of the Transformer regression model on the training set. Where y is the true sequence value, is the predicted sequence value, λ is the regularization coefficient. The loss function can be expressed as formula (7):

[0106]

[0107] In formula (7), N is the number of samples in the training set, s represents the sth sample, ranging from 1 to N, l is the loss function of a single sample, ||Θ|| 2 represents the L2 regularization term of the parameter, and λ is the regularization coefficient.

[0108] The Transformer regression model parameters Θ are optimized in each training cycle. The parameter update is performed using the gradient descent method. The parameter update can be expressed as formula (8):

[0109]

[0110] In formula (8), η is the learning rate, is the gradient of the loss function with respect to the parameter Θ, Θ t represents the Transformer regression model parameters at time t, Θ t+1 Represents the updated Transformer regression model parameters at time t+1. Dropout regularization technology can also be introduced into the model to improve the generalization ability of the model.

[0111] In some embodiments, multiple rounds of retrospective predictions can be performed on the Transformer regression model based on the test set according to the epidemic time information to evaluate the predictive performance of the Transformer regression model to obtain a retrospective predictive evaluation of at least one virus subtype at the epidemic time information scale.

[0112] For example, an independent test set method can be applied.test The mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R 2 ) and other evaluation indicators to evaluate the prediction performance of the Transformer model, and repeatedly execute the Transformer model parameter optimization method until the predicted sequence meets the requirements and obtains a retrospective prediction evaluation of at least one virus subtype at the epidemic time information scale.

[0113] In some embodiments, a time series-based cross-validation strategy may be used to evaluate the prediction performance of the Transformer regression model according to the popular time information, and the parameters in the Transformer regression model may be optimized again according to the evaluation results of the prediction performance to obtain the target parameters.

[0114] For example, in order to further optimize the model and measure its prediction accuracy and reliability in different infection rebound periods, a time series-based cross-validation strategy can be used to evaluate the prediction performance of the Transformer regression model in stages. The data of the above different influenza epidemic seasons can be selected for multiple rounds of training-testing cycles. For example, first, the time series dataset D can be divided into multiple continuous time windows W containing different influenza epidemic seasons. t , each W t The window corresponds to a season. Secondly, for each time window W t , using all previous windows (W1,…,W t-1 ) data for model training, and in the current window W t Repeat this process to evaluate the prediction error of the model in different seasons. Further, the infection rate prediction performance of the model on historical data can be evaluated, and the optimal hyperparameter Θ of the Transformer regression model can be obtained. * , where Θ * It can be expressed as formula (9):

[0115]

[0116] In formula (9), is a loss function that includes a regularization term.

[0117] For example, the optimized Transformer regression model can be used to obtain the infection rates of four seasonal influenza subtypes {y H1N1 ,y H3N2 ,y Victoria ,y Yamagata} Retrospective forecast evaluation at the epidemic season scale.

[0118] In some embodiments, the optimized target parameters can be used as the initial parameters of the Transformer prediction model, and based on a multi-step forward-looking prediction strategy of stepwise approximation, a predicted infection rate sequence of at least one virus subtype in a first preset time period is generated.

[0119] For example, a multi-step autoregressive Transformer prediction model can be constructed. First, the optimal hyperparameter Θ is used. * , construct the Transformer Encoder-Decoder architecture on the historical data set, equip it with a self-attention mechanism, and set parameters to obtain the prediction model: Secondly, set a suitable prediction time window T ′ , for example, T ′ It can be 1 week. Use the last T of the historical data set ′ Weekly time series data as input x t-T′+1:t To start the prediction process. A multi-step forward-looking prediction strategy of progressive approximation is adopted, that is, at each time t, the Transformer prediction model predicts the infection rate at the next time point t+1. This process can be expressed as formula (10), namely:

[0120]

[0121] Then, the new predicted infection rate is added to The time window tail is updated, and the oldest value is discarded to form a new input x t-T′+2:t+1 Finally, the above iterative process is followed until a complete sequence of predicted infection rates for the next 6 months (i.e., an example of the first preset time period) is generated.

[0122] In some embodiments, an error correction for the infection rate prediction value may also be included.

[0123] It may include updating the time series data set according to a preset time period, and then analyzing the error accumulation trend to obtain the analysis result of the error accumulation trend, and according to the analysis result of the error accumulation trend, using formula (11):

[0124]

[0125] Adjust the parameters of the Transformer prediction model;

[0126] In formula (11), α is the learning rate, is the gradient calculated based on the error ∈, Θ is the model parameter, Θ update are the adjusted model parameters.

[0127] The preset time period may be one month, one quarter or half a year, and this application does not impose any limitation on this.

[0128] For example, the autoregressive forecasting process may lead to the accumulation of forecast errors, so some measures must be taken to ensure the accuracy of the forecast trend and the overall forecast quality.

[0129] These measures may include:

[0130] 1. Continuously monitor and integrate updated epidemiological data. To address potential biases in predictions and maintain the timeliness of the model, plan to obtain the latest epidemiological datasets regularly (for example, once a month). new , update the time series dataset to form an updated time series dataset: D update (D update =D∪D new ).

[0131] 2. Error trend analysis. Monitoring (y is the calculated true value of infection rate, is the infection rate prediction value), and analyze the cumulative trend of errors. If systematic accumulation of errors is observed, then optimized model parameters or structures are required.

[0132] 3. Parameter adjustment and model correction. Based on the error trend, the model parameter θ can be adjusted and necessary model correction can be performed in a timely manner through formula (11).

[0133] For example, the optimized Transformer prediction model can be used to deduce the epidemic trends of infection rates of four seasonal influenza subtypes in the next six months:

[0134] The embodiment of the present application inputs the constructed data set into a deep learning model to predict the predicted infection rate sequence of at least one virus subtype in a first preset time period, thereby identifying the key variables that determine the risk of influenza infection from the influencing factors, thereby improving the effectiveness of influenza prediction.

[0135] In step 4, the predicted infection rate sequence can be weighted and summed according to the warning contribution of multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype.

[0136] Multidimensional influencing factors may include time series data such as the number of cases, air and rail passenger traffic, climate change, the intensity of non-drug intervention measures, and the full-fragment protein sequence of pathogens.

[0137] Risk warning information may include warning index and epidemic risk level.

[0138] In some implementations, the importance assessment of multi-dimensional influencing factors can be achieved by quantifying the warning contribution of each dimensional influencing factor.

[0139] For example, after obtaining the optimal hyperparameter θ of the Transformer regression model * After that, we can use the Transformer model The Shapley value is used to quantify the early warning contribution of each multi-dimensional influencing factor. v Describes the contribution of multi-dimensional influencing factors to the change of model output, which can be calculated by formula (12):

[0140]

[0141] In formula (12), x v represents the vth influencing factor.

[0142] In some embodiments, the formula: Determine the early warning index of at least one virus subtype. t represents the early warning index, φ v represents the contribution of the vth influencing factor to the change in model output, It represents the predicted infection rate of the Transformer prediction model at time point t+1. That is, the Shapley value of the influenza influencing factor is used as a weight factor and combined with the time series of the predicted infection rate to calculate the early warning indicator I t .

[0143] In some embodiments, the epidemic risk level of at least one virus subtype can be determined based on statistical data of the early warning index.

[0144] For example, the t Statistical analysis was performed to obtain I t The quartile q 0.25 ,q 0.5 ,q 0.75 , in order to classify the warning intensity of seasonal influenza. This process can be expressed as formula (13):

[0145]

[0146] Among them, Warning Level (I t) represents the warning intensity of seasonal influenza, and its value can be high risk (High Risk), moderate-high risk (Moderate-High Risk), moderate-low risk (Moderate-Low Risk) and low risk (Low Risk). Afterwards, the risk warning information of at least one virus subtype can be established respectively through the above method.

[0147] For example, early warning indices for four seasonal influenza subtypes can be established respectively: Warning K represents four seasonal influenza subtypes. It should be understood that this is warning information based on time t. In the time dimension, based on the change of time t, risk warning information of different time periods can be obtained. In the virus subtype dimension, risk warning information of different virus subtypes can also be obtained; further, combining virus subtype and time information, risk warning information of different virus subtypes at different times can be obtained. In this way, more accurate risk warning information of seasonal influenza can be obtained.

[0148] Please refer to Figure 2 , Figure 2 Another schematic diagram of the seasonal influenza epidemic risk early warning method according to an embodiment of the present application is shown in FIG. Figure 2 In the paper, the epidemic risk warning method of seasonal influenza is divided into four parts, namely, the part of obtaining time series data, the part of retrospective prediction model, the part of prospective prediction model and the part of epidemic risk warning.

[0149] In the part of obtaining time series data, population migration, climate, and non-drug intervention measures can be regarded as part of the multi-dimensional influencing factors of seasonal influenza. Another part can be obtained by calculating the subtype antigen cluster composition ratio through the pathogen protein sequence. Another part can be obtained by calculating the influenza subtype infection rate through the number of influenza subtype cases.

[0150] The obtained time series data of multi-dimensional influencing factors of seasonal influenza were preprocessed to obtain subtype-specific datasets, and then the subtype-specific datasets were input into Figure 2The retrospective prediction model part in this part. In this part, the subtype-specific dataset is divided into a training set and a test set, and the training set is input into the Transformer model (i.e., the Transformer regression model), and the structure and parameters of the Transformer model are optimized to obtain the optimal parameters. Afterwards, the optimal parameters are respectively input into the Transformer model in the prospective prediction model part (i.e., the Transformer prediction model) and the retrospective prediction part. In the retrospective prediction part, the predictive performance of the Transformer model is evaluated by retrospectively predicting the test set and combining the optimal parameters. The retrospective prediction model part also includes dividing the subtype-specific dataset into epidemic season time series, and then inputting it into the Transformer regression model for cross-validation to achieve the evaluation of the predictive performance of the Transformer model.

[0151] In the forward-looking prediction model, the subtype-specific dataset is input into the Transformer model (i.e., the Transformer prediction model), and then the autoregression, subtype infection rate prediction, and prediction error correction steps are performed, and the parameters of the Transformer prediction model are adjusted according to the prediction error correction. Among them, the updated subtype-specific dataset is also taken into account in the prediction error correction.

[0152] In the epidemic risk warning part, by inputting the subtype-specific dataset into the Transformer model, based on the optimal parameters obtained in the retrospective prediction model part, the SHAP method is used to obtain the importance of each factor in the multi-dimensional influencing factors, and the warning index is calculated in combination with the subtype predicted infection rate to further obtain the subtype risk level.

[0153] The above is an introduction to the embodiment of the method of the present application.

[0154] For the seasonal influenza epidemic risk warning method provided above, the present application embodiment also provides a seasonal influenza epidemic risk warning system for implementing the above method, such as Figure 3 As shown, Figure 3 This is a schematic block diagram of the modules of the seasonal influenza epidemic risk warning system of the embodiment of the present application. The seasonal influenza epidemic risk warning system includes:

[0155] An acquisition module, used to acquire multi-dimensional influencing factors of seasonal influenza, where the multi-dimensional influencing factors may include pathogen sequence information of at least one virus subtype and case data of seasonal influenza;

[0156] A first determination module is used to determine the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and can also determine the infection rate of at least one virus subtype through seasonal influenza case data;

[0157] A prediction module, for predicting a predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, population migration characteristics, infection rate of at least one virus subtype, intensity of non-drug intervention measures and climate characteristics;

[0158] The second determination module is used to perform weighted summation on the predicted infection rate sequence through the warning contribution of multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype.

[0159] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0160] The present application also provides an electronic device, such as Figure 4 As shown, the electronic device includes a memory, one or more processors ( Figure 4 Only one is shown) and a computer program stored in the memory and executable on the processor. Wherein: the memory is used to store software programs and units, and the processor executes various functional applications and data processing by running the software programs and units stored in the memory to obtain the resources corresponding to the above-mentioned preset events. Specifically, the processor implements the above-mentioned seasonal influenza epidemic risk warning method by running the above-mentioned computer program stored in the memory. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0161] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0162] The embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned seasonal influenza epidemic risk warning method is implemented.

[0163] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0164] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0165] The embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by one or more processors, the steps of the seasonal influenza epidemic risk warning method can be implemented.

[0166] It can be understood that the contents of the above method embodiments are all applicable to this computer program product, the functions specifically implemented by this computer program product embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0167] The seasonal influenza epidemic risk warning method, system, electronic device and computer program product provided in the embodiments of the present application obtain multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; determine the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determine the infection rate of at least one virus subtype through the case data of seasonal influenza; predict the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; and perform weighted summation of the predicted infection rate sequence through the warning contribution of the multi-dimensional influencing factors to determine the risk warning information of at least one virus subtype. Since the time series composition ratio of antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics are introduced into the deep learning model to predict the infection rate sequence, and further consideration is given to the warning contribution of multi-dimensional influencing factors to perform weighted summation on the predicted infection rate sequence to finally obtain the risk warning information of at least one virus subtype, this can comprehensively capture the key factors affecting influenza infection and improve the accuracy of seasonal influenza epidemic risk warning.

[0168] In the embodiment of the present application, by introducing the Transformer architecture and SHAP explanatory framework, a new solution is provided for the prediction and early warning of the epidemic risk of multi-subtype seasonal influenza. The Transformer architecture performs well in grasping the long-term dependencies in time series data, enabling it to build a multi-step prediction model with high accuracy. The SHAP explanatory method can provide a quantitative importance assessment for features with high contribution to model prediction, helping to identify and explain important factors in epidemiological data. The embodiment of the present application combines Transformer and SHAP, based on epidemiological time series big data, to accurately predict the infection risk of global and regional influenza virus subtypes, not only to identify the key variables that determine the risk of influenza infection from the influencing factors, but also to establish a set of systematic influenza epidemic risk early warning strategies based on these key variables, thereby improving the effectiveness of influenza prediction and early warning. Provide a scientific basis for formulating accurate seasonal influenza prevention strategies.

[0169] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0170] Although specific embodiments are described herein, those of ordinary skill in the art will recognize that many other modifications or alternative embodiments are also within the scope of the present disclosure. For example, any of the functions and / or processing capabilities described in conjunction with a particular device or component may be performed by any other device or component. In addition, although various exemplary implementations and architectures have been described according to embodiments of the present disclosure, those of ordinary skill in the art will recognize that many other modifications to the exemplary implementations and architectures described herein are also within the scope of the present disclosure.

[0171] Some aspects of the present disclosure are described above with reference to the block diagram and flow chart of the system, method, system and / or computer program product according to the exemplary embodiments.It should be understood that the combination of one or more blocks in the block diagram and the flow chart and the blocks in the block diagram and the flow chart can be realized by executing computer executable program instructions respectively.Equally, according to some embodiments, some blocks in the block diagram and the flow chart may not need to be executed in the order shown, or may not need to be executed in full.In addition, additional components and / or operations beyond those components and / or operations shown in the blocks in the block diagram and the flow chart may be present in certain embodiments.

[0172] Therefore, the blocks in the block diagrams and flow charts support the combination of devices for performing the specified functions, the combination of elements or steps for performing the specified functions, and the program instruction devices for performing the specified functions. It should also be understood that each block in the block diagrams and flow charts and the combination of blocks in the block diagrams and flow charts can be implemented by a special-purpose hardware computer system or a combination of special-purpose hardware and computer instructions that performs a specific function, element or step.

[0173] The program modules, applications, etc. described herein may include one or more software components, including, for example, software objects, methods, data structures, etc. Each such software component may include computer-executable instructions that, in response to execution, cause at least a portion of the functionality described herein (e.g., one or more operations of the illustrative methods described herein) to be performed.

[0174] Software components can be encoded with any of various programming languages. An exemplary programming language can be a low-level programming language, such as an assembly language associated with a specific hardware architecture and / or an operating system platform. The software component comprising assembly language instructions may need to be converted to executable machine code by an assembler before being executed by hardware architecture and / or platform. Another exemplary programming language can be a higher-level programming language, which can be transplanted across multiple architectures. The software component comprising a higher-level programming language may need to be converted to an intermediate representation by an interpreter or a compiler before execution. Other examples of programming languages ​​include but are not limited to macro languages, shells or command languages, job control languages, scripting languages, database queries or search languages ​​or report writing languages. In one or more exemplary embodiments, the software component comprising the instruction of one of the above-mentioned programming language examples can be directly executed by an operating system or other software components, without first being converted into another form.

[0175] Software components may be stored as files or other data storage structures. Software components of similar type or related functions may be stored together, such as in a specific directory, folder, or library. Software components may be static (e.g., preset or fixed) or dynamic (e.g., created or modified at execution time).

[0176] The embodiments of the present application are described in detail above in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in the relevant technical field without departing from the purpose of the present application.

Claims

1. A seasonal influenza epidemic risk early warning method, characterized in that: The following steps are involved: Obtaining multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; Determine the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determine the infection rate of at least one virus subtype through the case data of seasonal influenza; Predicting a predicted infection rate sequence of at least one viral subtype in a first preset time period through a deep learning model based on the temporal composition ratio of the antigen clusters of at least one viral subtype, the infection rate of at least one viral subtype, population migration characteristics, the intensity of non-drug intervention measures, and climate characteristics; Performing a weighted summation of the predicted infection rate sequence according to the warning contribution of the multi-dimensional influencing factors to determine risk warning information of at least one virus subtype; Wherein, determining the temporal composition ratio of antigen clusters of at least one viral subtype through the pathogen sequence information of the viral subtype comprises: Determining a protein sequence mutation matrix by using the antigenic Euclidean distance between the full-segment protein sequence data of at least one viral subtype; Clustering the protein sequence mutation matrix according to a Markov clustering algorithm to generate multiple antigen clusters; The temporal composition ratio of antigen clusters of at least one viral subtype is determined based on the plurality of antigen clusters.

2. The seasonal influenza epidemic risk early warning method according to claim 1, characterized in that: The step of clustering the protein sequence mutation matrix according to the Markov clustering algorithm to generate a plurality of antigen clusters comprises: Multiple antigen clusters are determined by the formula: C = expand(M,2)⊙inflate(M,r); Among them, C is the set of antigen clusters, expand and inflate are operations in the Markov clustering algorithm, M is the protein sequence mutation matrix, and r is the expansion parameter.

3. The seasonal influenza epidemic risk early warning method according to claim 1, characterized in that: The step of determining the temporal composition ratio of antigen clusters of at least one virus subtype based on the plurality of antigen clusters comprises: The time series proportion of the corresponding antigen cluster is determined by the ratio of any antigen cluster to the sum of all generated antigen clusters; By resampling the time series proportions, the time series composition ratio of the antigen cluster of at least one virus subtype is obtained.

4. The seasonal influenza epidemic risk early warning method according to claim 1, characterized in that: Determining the infection rate of at least one virus subtype through the seasonal influenza case data comprises: The infection rate of the corresponding virus subtype is obtained by multiplying the ratio of the number of seasonal influenza-like medical records to the number of visits in the second preset time period, the ratio of the number of positive cases to the number of test samples, and the proportion of any virus subtype in the number of positive cases.

5. The seasonal influenza epidemic risk early warning method according to claim 1, characterized in that: The method predicts the predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics, including: Determining a time series data set, wherein the time series data set includes a time series composition ratio of antigen clusters of at least one virus subtype, an infection rate of the corresponding virus subtype, population migration characteristics, intensity of non-drug intervention measures, and climate characteristics; Dividing the time series data set into a test set and a training set according to the epidemic time information of the seasonal influenza; Inputting the training set into a Transformer regression model, and optimizing parameters in the Transformer regression model to obtain target parameters; Based on the test set, the Transformer regression model is retrospectively predicted multiple times according to the epidemic time information, and the prediction performance of the Transformer regression model is evaluated to obtain a retrospective prediction evaluation of at least one virus subtype at the epidemic time information scale; Using the target parameter as the initial parameter of the Transformer prediction model, and generating a predicted infection rate sequence of at least one virus subtype in a first preset time period based on a multi-step forward-looking prediction strategy of stepwise approximation; The step of evaluating the prediction performance of the Transformer regression model includes evaluating the prediction performance of the Transformer regression model through mean absolute error, root mean square error, and determination coefficient.

6. The seasonal influenza epidemic risk early warning method according to claim 5, characterized in that: The step of optimizing the parameters in the Transformer regression model to obtain target parameters includes: Optimizing the fitting error of the Transformer regression model based on a loss function to which a regularization term is added; Using the gradient descent method to update the parameters in the Transformer regression model to obtain the target parameters; The evaluating the prediction performance of the Transformer regression model further includes: A time series-based cross-validation strategy is used to evaluate the prediction performance of the Transformer regression model according to the popular time information, and the parameters in the Transformer regression model are optimized again according to the evaluation results of the prediction performance to obtain the target parameters.

7. The seasonal influenza epidemic risk early warning method according to claim 5, characterized in that: Also includes: Updating the time series data set according to a preset time period; Analyze the error accumulation trend and obtain the analysis result of the error accumulation trend; According to the analysis results of the error accumulation trend, the formula is: Adjusting parameters of the Transformer prediction model; Among them, α is the learning rate, is the gradient calculated based on the error ε, Θ is the model parameter, Θ update are the adjusted model parameters.

8. The seasonal influenza epidemic risk early warning method according to claim 1, characterized in that: The step of weighting and summing the predicted infection rate sequence by the warning contribution of the multi-dimensional influencing factors to determine risk warning information of at least one virus subtype includes: Quantify the early warning contribution of each of the multi-dimensional influencing factors; By formula: Determine an early warning index for at least one viral subtype; Among them, I t represents the early warning index, φ v represents the contribution of the vth factor to the change in model output, Represents the infection rate prediction value of the Transformer prediction model at time point t+1; Determining the epidemic risk level of at least one virus subtype based on the statistical data of the early warning index; Wherein, the risk warning information includes the warning index and the epidemic risk level.

9. A seasonal influenza epidemic risk early warning system, characterized in that: The system comprises: An acquisition module, used to acquire multi-dimensional influencing factors of seasonal influenza, wherein the multi-dimensional influencing factors include pathogen sequence information of at least one virus subtype and case data of seasonal influenza; A first determination module is used to determine the temporal composition ratio of antigen clusters of at least one virus subtype through the pathogen sequence information of the virus subtype, and determine the infection rate of at least one virus subtype through the case data of seasonal influenza; A prediction module, for predicting a predicted infection rate sequence of at least one virus subtype in a first preset time period through a deep learning model based on the temporal composition ratio of the antigen clusters of at least one virus subtype, the infection rate of at least one virus subtype, population migration characteristics, the intensity of non-drug intervention measures and climate characteristics; A second determination module is used to perform weighted summation on the predicted infection rate sequence according to the warning contribution of the multi-dimensional influencing factors to determine risk warning information of at least one virus subtype; Wherein, determining the temporal composition ratio of antigen clusters of at least one viral subtype through the pathogen sequence information of the viral subtype comprises: Determining a protein sequence mutation matrix by using the antigenic Euclidean distance between the full-segment protein sequence data of at least one viral subtype; Clustering the protein sequence mutation matrix according to a Markov clustering algorithm to generate multiple antigen clusters; The temporal composition ratio of antigen clusters of at least one viral subtype is determined based on the plurality of antigen clusters.

Images