Public health event traceability management method and device and storage medium

Abstract

The invention provides a public health event traceability management method and device and a storage medium, and the method comprises the steps: obtaining a first infectious disease diffusion information atlas which represents one or more infectious disease outbreak activities, and the first infectious disease diffusion information atlas comprises a plurality of primitives and connection edges among the primitives, each infectious disease outbreak activity is represented based on a plurality of primitives in the plurality of primitives and connection edges among the plurality of primitives, and each primitive in each infectious disease outbreak activity comprises information of disease monitoring elements forming the infectious disease outbreak activity, loading the first infectious disease diffusion information atlas to an infectious disease traceability identification network, the traceability identification result represents whether the one or more infectious disease outbreak activities include the source infectious disease outbreak activity or not. The accuracy of infectious disease traceability can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method, device and storage medium for tracing the source of public health events. Background Art

[0002] In the field of public health, the outbreak and spread of infectious diseases pose a serious threat to human health and social development. Rapid and accurate tracing of the source of infectious diseases is crucial for taking effective measures in a timely manner and curbing the spread of infectious diseases. However, the existing methods for tracing the source of public health events face many challenges in practical applications. Traditional methods for tracing the source of infectious diseases mainly rely on manual investigation and empirical judgment. For example, the possible source of infection and transmission path are inferred by checking the travel trajectory and contact history of the cases one by one. This method not only consumes a lot of manpower, material resources and time, but also often finds it difficult to obtain information comprehensively and accurately when facing complex infectious disease transmission scenarios, and is prone to omissions and misjudgments, resulting in a significant reduction in the accuracy and timeliness of the tracing results. With the development of information technology, some tracing methods based on data analysis have gradually emerged. However, most of these methods are limited to single-dimensional data processing, such as focusing only on one aspect of information such as the temporal distribution or spatial distribution of cases, and cannot comprehensively consider the multiple factors involved in the outbreak of infectious diseases and their interrelationships. The spread of infectious diseases is a complex dynamic process, which is affected by the combined effects of multiple factors such as disease monitoring elements (such as patient symptoms, geographical location, infection time, etc.), infectious disease types, and the interaction between different outbreak activities. Single-dimensional analysis methods cannot fully explore the potential information behind these complex factors, making it difficult to accurately identify the source of infectious disease outbreaks. Summary of the invention

[0003] In view of this, the embodiment of the present application provides a public health event tracing management method, device and storage medium. The technical solution of the embodiment of the present application is implemented as follows: In a first aspect, the present application provides a method for tracing the source of a public health event, comprising: obtaining a first infectious disease spread information map, the first infectious disease spread information map representing one or more infectious disease outbreak activities, the first infectious disease spread information map comprising a plurality of graph elements and connecting edges between the plurality of graph elements, each infectious disease outbreak activity being represented based on a number of graph elements among the plurality of graph elements and the connecting edges between the plurality of graph elements, each graph element in each infectious disease outbreak activity comprising information on disease monitoring elements constituting the infectious disease outbreak activity; loading the first infectious disease spread information map into an infectious disease traceability identification network to obtain a traceability identification result, the traceability identification result representing whether one or more infectious disease outbreak activities include a source infectious disease outbreak activity, the source infectious disease outbreak activity being determined based on a commonality measurement between infectious disease outbreak activities or a commonality measurement between graph elements in infectious disease outbreak activities.

[0004] Based on this application, the infectious disease source tracing and identification network can determine the deviation of infectious disease outbreak activities by measuring the commonalities between graph elements in infectious disease outbreak activities or the commonalities between infectious disease outbreak activities, so as to more accurately identify the source infectious disease outbreak activities in multiple dimensions.

[0005] Optionally, the infectious disease traceability and identification network includes one or more of the following components: a graph element group commonality measurement component, a multidimensional association analysis component or an activity association analysis component. The graph element group commonality measurement component is used to obtain the commonality measurement between graph elements in infectious disease outbreak activities, the multidimensional association analysis component is used to obtain the commonality measurement between graph elements and infectious disease types in infectious disease outbreak activities, and the activity association analysis component is used to obtain the commonality measurement between infectious disease outbreak activities.

[0006] Based on the present application, the infectious disease source tracing and identification network may include a component for the deviation between graph elements in infectious disease outbreak activities or a component for the deviation between infectious disease outbreak activities, which can accurately identify the source infectious disease outbreak activities in multiple dimensions.

[0007] Optionally, if the infectious disease source tracing and identification network includes a graph element group commonality measurement component, loading the first infectious disease diffusion information map into the infectious disease source tracing and identification network to obtain the source tracing and identification result may include: outputting a first deviation of each infectious disease outbreak activity based on the graph element group commonality measurement component, wherein the graph element group commonality measurement component is used to obtain the commonality measurement between graph element groups in each infectious disease outbreak activity, and obtaining the first deviation based on the commonality measurement between graph element groups in each infectious disease outbreak activity; based on the first deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain the source tracing and identification result.

[0008] Based on this application, the commonality measurement between the graph elements in the infectious disease outbreak activities can be determined based on the graph element group commonality measurement component, so that the infectious disease source tracing and identification network can identify the deviation of the infectious disease outbreak activities based on the commonality measurement between the graph elements in the infectious disease outbreak activities, and accurately identify the source infectious disease outbreak activities.

[0009] Optionally, if the infectious disease source tracing and identification network includes a multidimensional association analysis component, the first infectious disease diffusion information map is loaded into the infectious disease source tracing and identification network to obtain the source tracing and identification result, and may also include: outputting the second deviation of each infectious disease outbreak activity based on the multidimensional association analysis component, wherein the multidimensional association analysis component is used to integrate multiple graph elements in one or more infectious disease outbreak activities to obtain representative graph elements, or to use the core elements of each infectious disease outbreak activity as representative graph elements, and to obtain the second deviation of each infectious disease outbreak activity based on the commonality measurement between one or more graph elements and the representative graph elements; based on the second deviation of each infectious disease outbreak activity, evaluate whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain the source tracing and identification result.

[0010] Generally speaking, infectious disease outbreaks may involve interactions between a variety of graph elements. We can generate a multidimensional correlation analysis component to identify the degree of interaction deviation between the graph elements in infectious disease outbreaks and the overall infectious disease outbreak, so as to accurately identify the deviation of infectious disease outbreaks.

[0011] Optionally, if the infectious disease source tracing and identification network includes an activity association analysis component, the first infectious disease diffusion information map is loaded into the infectious disease source tracing and identification network to obtain the source tracing and identification result, and may also include: outputting the third deviation of each infectious disease outbreak activity based on the activity association analysis component, wherein the activity association analysis component is used to obtain the commonality measurement between pairs of infectious disease outbreak activities, and calculating the third deviation of each infectious disease outbreak activity based on the commonality measurement between pairs of infectious disease outbreak activities; based on the third deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain the source tracing and identification result.

[0012] Based on this application, the infectious disease source tracing and identification network can also deploy activity correlation analysis components to identify the deviations between infectious disease outbreak activities and identify the source infectious disease outbreak activities using infectious disease outbreak activities as the smallest unit.

[0013] Optionally, the activity association analysis component is used to select similar activity sets corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities; based on the commonality measurement between each infectious disease outbreak activity and the corresponding infectious disease outbreak activities in the similar activity set, obtain the third deviation degree of each infectious disease outbreak activity.

[0014] Based on the present application, a similar activity set of each infectious disease outbreak activity can be determined, and whether the infectious disease outbreak activity is a source activity can be identified based on the commonality measurement between adjacent activities.

[0015] Optionally, the activity association analysis component is specifically used to extract contextual features of each infectious disease outbreak activity to obtain a coding vector for each infectious disease outbreak activity; and based on the coding vector of each infectious disease outbreak activity, calculate a commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in a corresponding similar activity set.

[0016] Based on this, the present application can accurately determine the commonality measure between infectious disease outbreak activities based on the extracted vectors.

[0017] Optionally, if the infectious disease source tracing and identification network includes a graph group commonality measurement component, a multidimensional association analysis component and an activity association analysis component, then the first infectious disease diffusion information map is loaded into the infectious disease source tracing and identification network to obtain a source tracing and identification result, and also includes: using the first infectious disease diffusion information map as the input of the graph group commonality measurement component, the multidimensional association analysis component and the activity association analysis component respectively; integrating the first deviation of each infectious disease outbreak activity output by the graph group commonality measurement component, the second deviation of each infectious disease outbreak activity output by the multidimensional association analysis component and the third deviation of each infectious disease outbreak activity output by the activity association analysis component to obtain a fourth deviation of each infectious disease outbreak activity; evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity based on the fourth deviation of each infectious disease outbreak activity to obtain a source tracing and identification result.

[0018] Based on this application, when the infectious disease source tracing and identification network contains multiple components, the results output by multiple components can be integrated to identify source infectious disease outbreak activities at multiple scales, with more accurate results.

[0019] Optionally, the first infectious disease spread information map is loaded into the infectious disease source tracing and identification network, including: projecting each graphic element in each infectious disease outbreak activity in the first infectious disease spread information map into the same feature domain to obtain a second representation vector of each infectious disease outbreak activity in the same feature domain; loading the second representation vector into the infectious disease source tracing and identification network to obtain a tracing and identification result.

[0020] Generally speaking, the graph elements in the infectious disease spread information map may have different dimensions. Projecting each graph element into the same feature domain and obtaining the encoding vectors of each graph element with consistent dimensions can help to perform analysis based on the encoding vectors of each graph element with the same dimension.

[0021] In this application, the training process of the infectious disease source tracing identification network includes: A second infectious disease diffusion information map is obtained, the second infectious disease diffusion information map represents multiple infectious disease outbreak activities, the second infectious disease diffusion information map includes multiple graph elements and connecting edges between the multiple graph elements, each infectious disease outbreak activity is represented based on several graph elements among the multiple graph elements and the connecting edges between the several graph elements, and the graph elements in each infectious disease outbreak activity include information on disease monitoring elements that constitute the infectious disease outbreak activity; an infectious disease source tracing identification network is constructed based on the second infectious disease diffusion information map, the infectious disease source tracing identification network is used to identify source infectious disease outbreak activities among multiple infectious disease outbreak activities, and the source infectious disease outbreak activity is determined based on commonality measurements between infectious disease outbreak activities or commonality measurements between graph elements in infectious disease outbreak activities.

[0022] The infectious disease diffusion information map sets multiple graph elements, and the edges between multiple graph elements represent the association relationship, which is used to represent the complex infectious disease outbreak activities. Modeling is carried out based on the infectious disease diffusion information map to obtain a network that can identify the source of infectious disease outbreak activities, namely the infectious disease source tracing identification network, to trace the source of infectious disease outbreak activities.

[0023] Optionally, the infectious disease source tracing and identification network includes a graph element group commonality measurement component, which can be used to obtain the commonality measurement of the graph element group, that is, the commonality measurement between graph elements, and every several graph elements form a pair of graph element groups; generating an infectious disease source tracing and identification network based on the second infectious disease diffusion information map, including: grouping multiple graph elements in each infectious disease outbreak activity into at least one group of graph element groups, each group of graph element groups may include several graph elements; based on the graph element group commonality measurement component, obtaining the first commonality measurement of each group of graph element groups in at least one group of graph element groups, that is, the commonality measurement between every several graph elements; based on the first commonality measurement of each group of graph element groups, obtaining the graph element group commonality measurement error value of each graph element in multiple graph elements, and iterating the infectious disease source tracing and identification network based on the graph element group commonality measurement error value of each graph element group to obtain the iterated infectious disease source tracing and identification network.

[0024] Based on this application, the commonality measurement between the graph elements in the infectious disease outbreak activity can be determined based on the graph element group commonality measurement component, and metric learning can be performed based on the commonality measurement between the graph elements, so that the infectious disease source tracing and identification network can identify the deviation of the infectious disease outbreak activity based on the commonality measurement between the graph elements in the infectious disease outbreak activity, and accurately identify the source infectious disease outbreak activity.

[0025] Optionally, the infectious disease source tracing and identification network is iterated based on the graph group common measurement error value of each graph group to obtain the iterated infectious disease source tracing and identification network, including: integrating the graph group common measurement error values ​​of multiple graph groups in each infectious disease outbreak activity to obtain a first training error value; iterating the infectious disease source tracing and identification network based on the first training error value to obtain the iterated infectious disease source tracing and identification network.

[0026] Based on the present application, in the iterative infectious disease source tracing and identification network, the global graph group metric loss of the infectious disease outbreak activity is determined by combining the metric loss of each graph group, so as to learn based on the global infectious disease outbreak activity and obtain an infectious disease source tracing and identification network that can identify the source infectious disease outbreak activity based on the common metrics between graphs.

[0027] Optionally, the pixel group commonality metric error value of each pixel is obtained based on the first commonality metric of each group of pixel groups, including: obtaining a similar pixel sample set of the first pixel from multiple pixels, and constructing a dissimilar pixel sample set, the first commonality metric between the pixels in the similar pixel sample set and the first pixel is greater than the first commonality metric between the pixels in the dissimilar pixel sample set and the first pixel, and the first pixel is a random one of multiple pixels in each infectious disease outbreak activity; based on the first commonality metric between the first pixel and the pixels in the similar pixel sample set, and the commonality metric between the first pixel and the pixels in the dissimilar pixel sample set, determine the pixel group commonality metric error value corresponding to the first pixel group.

[0028] Based on the present application, when determining samples, there is no need to determine negative samples separately. A set of similar graph element samples can be determined from the graph elements in the infectious disease outbreak activities, and a set of dissimilar graph element samples can be generated based on the graph elements in the infectious disease outbreak activities. By performing comparative learning based on the similar graph element sample set and the dissimilar graph element sample set, the obtained graph element group commonality measurement component can determine the deviated graph elements in the infectious disease outbreak activities.

[0029] Optionally, based on a first commonality measure between the first primitive and primitives in a sample set of similar primitives, and a commonality measure between the first primitive and primitives in a sample set of dissimilar primitives, determining a primitive group commonality measure error value corresponding to the first primitive group includes: obtaining a sensitivity adjustment factor, the sensitivity adjustment factor being related to the commonality measure between the primitives in a sample set of dissimilar primitives and the first primitive; based on the sensitivity adjustment factor, based on the first commonality measure between the first primitive and primitives in a sample set of similar primitives, and a commonality measure between the first primitive and primitives in a sample set of dissimilar primitives, determining a primitive group commonality measure error value corresponding to the first primitive group.

[0030] Based on this application, a sensitivity adjustment factor is set to adjust the focus weight of difficult samples based on the sensitivity adjustment factor, reduce the impact of difficult samples on the tuning results, and improve the tuning quality.

[0031] Optionally, the infectious disease source tracing identification network further includes a multidimensional association analysis component, which is used to perform cluster analysis on the graph elements in the infectious disease outbreak activities to obtain no less than one category, and obtain the commonality measurement between each graph element in the infectious disease outbreak activities and no less than one category, and the commonality measurement represents the deviation degree of the corresponding infectious disease outbreak activities; Constructing an infectious disease traceability identification network based on a second infectious disease diffusion information map also includes: obtaining a second commonality measure between one or more of the multiple graphs in each infectious disease outbreak activity and a representative graph in a multidimensional association analysis component, where the representative graph may include the central graph of each infectious disease outbreak activity or an graph obtained by integrating multiple graphs; determining a second training error value based on the second commonality measure between one or more graphs and the representative graph; iterating the infectious disease traceability identification network based on the second training error value to obtain an iterated infectious disease traceability identification network.

[0032] Generally speaking, infectious disease outbreak activities may involve interactions between multiple graph elements. By generating a multidimensional correlation analysis component to identify the interaction deviation between the graph elements in the infectious disease outbreak activities and the global infectious disease outbreak activities, the deviation of the infectious disease outbreak activities can be accurately identified.

[0033] Optionally, the multidimensional association analysis component can also perform cluster analysis on multiple graph elements in each infectious disease outbreak activity to obtain no less than one category; change the first graph element into a second graph element, the first graph element is an element in the first infectious disease outbreak activity, and the second graph element has the same attributes as the first graph element but a different category; obtain a third commonality measure between the second graph element and the representative graph element; determine a training error value based on the second commonality measure and the third commonality measure to obtain a second training error value.

[0034] Based on this application, dissimilar samples can be generated by changing the graph elements in the infectious disease outbreak activities into target graph elements to help unsupervised metric learning.

[0035] Optionally, the infectious disease source tracing and identification network also includes an activity association analysis component, which is used to obtain commonality metrics between infectious disease outbreak activities; generating an infectious disease source tracing and identification network based on the second infectious disease diffusion information map may also include: selecting similar activity sets and dissimilar activity sets corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities; determining a third training error value based on a fourth commonality metric between each infectious disease outbreak activity and the infectious disease outbreak activities in the similar activity set and a fifth commonality metric between each infectious disease outbreak activity and the infectious disease outbreak activities in the dissimilar activity set; and iterating the infectious disease source tracing and identification network based on the third training error value to obtain an iterated infectious disease source tracing and identification network.

[0036] Based on this application, an activity association analysis component is constructed to identify common metrics between infectious disease outbreak activities, and metric learning is completed by determining similar activity sets and dissimilar activity sets of infectious disease outbreak activities.

[0037] Optionally, similar activity sets and dissimilar activity sets corresponding to each infectious disease outbreak activity are selected from multiple infectious disease outbreak activities, including: obtaining the number of cross-graph elements between each group of infectious disease outbreak activities based on the activity association analysis component; obtaining one or more infectious disease outbreak activities whose number of cross-graph elements with a second infectious disease outbreak activity is greater than a preset threshold, to obtain a similar activity set, the second infectious disease outbreak activity being a random infectious disease outbreak activity among the multiple infectious disease outbreak activities; obtaining one or more infectious disease outbreak activities whose number of cross-graph elements with the second infectious disease outbreak activity is not greater than a preset threshold, to obtain a dissimilar activity set.

[0038] Based on the present application, similar activity sets and dissimilar activity sets of each infectious disease outbreak activity can be determined based on the number of graph elements intersecting between infectious disease outbreak activities, so as to select samples with high commonality measures as similar samples and samples with low commonality measures as dissimilar samples.

[0039] Optionally, obtaining the fourth commonality measure between each group of infectious disease outbreak activities among multiple infectious disease outbreak activities based on the activity association analysis component may include: performing context feature extraction on each infectious disease outbreak activity based on the activity association analysis component to obtain a representation of each infectious disease outbreak activity; the activity association analysis component may calculate the fourth commonality measure between the infectious disease outbreak activities based on each infectious disease outbreak activity representation.

[0040] Based on the present application, the encoding vector of each infectious disease outbreak activity can be obtained based on context feature extraction to accurately calculate the commonality metric based on the encoding vector.

[0041] Optionally, constructing an infectious disease traceability and identification network based on the second infectious disease diffusion information map may also include: projecting the data corresponding to each graph element in each infectious disease outbreak activity in the second infectious disease diffusion information map into the same feature domain to obtain a first representation vector for each infectious disease outbreak activity in the same feature domain; and constructing an infectious disease traceability and identification network based on the first representation vector.

[0042] In a second aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor implements the steps in the above-described method when executing the program.

[0043] In a third aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the above-described method when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the implementation flow of a public health event traceability management method provided in an embodiment of the present application.

[0045] Figure 2 A hardware entity schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The embodiment of the present application provides a method for tracing the source of public health events, which can be executed by a processor of a computer device. The computer device may refer to a device with data processing capabilities such as a server, a laptop, a tablet computer, and a desktop computer.

[0047] Figure 1 A schematic diagram of the implementation process of a public health event tracing management method provided in an embodiment of the present application is as follows: Figure 1 As shown, the method includes: Step S100: Obtain a first infectious disease diffusion information map, the first infectious disease diffusion information map represents one or more infectious disease outbreak activities, the first infectious disease diffusion information map includes multiple graph elements and connecting edges between the multiple graph elements, each infectious disease outbreak activity is represented based on several graph elements among the multiple graph elements and connecting edges between the several graph elements, and each graph element in each infectious disease outbreak activity includes information constituting disease monitoring elements in each infectious disease outbreak activity.

[0048] In step S100, it is necessary to first collect data related to infectious diseases. These data come from a wide range of sources, including but not limited to case reports from medical institutions, data from disease monitoring computer devices, and survey records from public health departments. For example, during an influenza epidemic, a computer device collects patient visit information from various hospitals, including the patient's basic information (name, age, gender, etc.), symptoms (fever, cough, fatigue, etc.), visit time, visit location, etc., and obtains statistical data on influenza cases in different regions from a disease monitoring computer device. These data will serve as the basis for constructing the first infectious disease diffusion information map. It is understandable that obtaining relevant data needs to be within the scope permitted by laws and regulations and on the premise that the parties are informed and authorized.

[0049] The first infectious disease diffusion information map consists of multiple graph elements and the connecting edges between the graph elements. Graph elements are the basic elements that constitute the map, and each infectious disease outbreak activity is represented based on a number of graph elements and the connecting edges between them. Each graph element in each infectious disease outbreak activity contains information that constitutes the disease monitoring elements in the infectious disease outbreak activity. These disease monitoring element information is key information for describing and analyzing infectious disease outbreak activities.

[0050] Taking the influenza epidemic as an example, the disease monitoring element information contained in the graph element can be of various types. For example, the patient graph element contains the patient's personal information (such as age, gender, occupation, etc.), disease information (disease type, symptoms, onset time, etc.); the location graph element contains information about the location of the epidemic, such as cities, streets, specific places (hospitals, schools, shopping malls, etc.); the time graph element records the time nodes related to the outbreak of infectious diseases, such as the first time of the case, the peak period of transmission, etc.; the transmission path graph element indicates the mode of transmission of infectious diseases, such as droplet transmission, contact transmission, etc. These different types of graph elements are interconnected through connecting edges to form a complete infectious disease diffusion information map. The connecting edges reflect the relationship between the graph elements. For example, the connecting edge between the patient graph element and the location graph element can indicate that the patient is infected or seeks medical treatment at a certain location; the connecting edge between the patient graph element and the time graph element can record the time when the patient becomes ill or seeks medical treatment; the connecting edge between the transmission path graph element and other related graph elements can explain how infectious diseases spread between entities represented by different graph elements.

[0051] The technical means for the computer device to construct the first infectious disease diffusion information map can adopt data modeling and graph database technology. First, according to the collected infectious disease data, data cleaning and preprocessing are performed to remove noise data and incomplete data. Then, according to the predefined primitive types and connection edge rules, the data is converted into graph structured data. For example, using a graph database (such as Neo4j), by writing corresponding query statements and data import scripts, various primitives and their relationships are stored in the database to form the first infectious disease diffusion information map.

[0052] During the construction process, for each infectious disease outbreak, the computer device determines the relevant primitives and establishes connecting edges between them. For example, in an influenza outbreak, the computer device identifies key information such as the patients involved, the hospitals where the patients were treated, the time of onset, and the route of transmission, and represents this information as corresponding primitives, and connects them through connecting edges, thereby fully representing the infectious disease outbreak.

[0053] Through the above method, the computer device obtains the first infectious disease spread information map, which comprehensively and intuitively displays the spread of the infectious disease and provides basic data support for subsequent loading it into the infectious disease traceability identification network for traceability analysis.

[0054] Step S200: Load the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result, which characterizes whether one or more infectious disease outbreak activities include source infectious disease outbreak activities, and the source infectious disease outbreak activities are determined based on the commonality measurement between infectious disease outbreak activities or the commonality measurement between graph elements in the infectious disease outbreak activities.

[0055] In step S200, the computer device first ensures that the infectious disease traceability identification network has been correctly constructed and is in an operational state. This network may include at least one of a graph element group commonality measurement component, a multidimensional association analysis component, or an activity association analysis component.

[0056] If the infectious disease traceability identification network includes a graph group commonality measurement component, the computer device outputs the first deviation of each infectious disease outbreak activity based on the component. The graph group commonality measurement component is used to obtain the commonality measurement between the graph groups in each infectious disease outbreak activity, and obtain the first deviation accordingly. For example, in an information graph of an influenza epidemic, for a certain infectious disease outbreak activity, the graph group may be a binary group consisting of a patient graph and a location graph. The computer device calculates the commonality measurement between graph groups in different infectious disease outbreak activities, such as calculating the similarity between different patient graphs and location graph combinations. The commonality measurement can be calculated by the cosine similarity formula. Assuming that the graph group A and the graph group B have characteristic vectors a and b, respectively, the cosine similarity sim(A,B)=(a·b) / (||a|| ||b||), where a·b is the vector dot product, and ||a|| and ||b|| are the moduli of vectors a and b, respectively. Based on these commonality measurement results, the first deviation of each infectious disease outbreak activity is calculated. For example, the commonality measurement result is directly used as the first deviation, or the average or standard deviation of the commonality measurement values ​​of the infectious disease outbreak activity and all activities in the similar activity set is calculated as the first deviation. The second deviation and the third deviation mentioned later refer to this determination method. If the commonality measurement between the graph element groups in a certain infectious disease outbreak activity is significantly different from the commonality measurement of the graph element groups in most other infectious disease outbreak activities, its first deviation will be higher.

[0057] If the infectious disease source tracing identification network includes a multidimensional association analysis component, the computer device outputs the second deviation of each infectious disease outbreak activity based on the component. The multidimensional association analysis component integrates multiple graphs in one or more infectious disease outbreak activities to obtain a representative graph, or uses the core elements of each infectious disease outbreak activity as a representative graph, and obtains the second deviation of each infectious disease outbreak activity based on the commonality measurement between one or more graphs and the representative graph. For example, in the malaria epidemic map, for each infectious disease outbreak activity, multiple graphs such as patient symptom graphs, infected area graphs, and transmission medium graphs are integrated to obtain a representative graph. The computer device calculates the commonality measurement between each graph and the representative graph. If the commonality measurement between the graph in a certain infectious disease outbreak activity and the representative graph is low, its second deviation is high. This may mean that the infectious disease outbreak activity differs from the overall situation in some key features.

[0058] If the infectious disease traceability identification network includes an activity association analysis component, the computer device outputs the third deviation of each infectious disease outbreak activity based on the component. The activity association analysis component obtains the commonality measure between pairs of infectious disease outbreak activities, and determines the third deviation of each infectious disease outbreak activity based on this. For example, in a dengue fever epidemic, the computer device first selects a similar activity set corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities. By extracting contextual features for each infectious disease outbreak activity, an encoding vector for each infectious disease outbreak activity is obtained. Based on these encoding vectors, the commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set is determined. If the commonality measure between an infectious disease outbreak activity and a similar activity set is low, its third deviation is high.

[0059] When the infectious disease source tracing identification network includes a graph element group commonality measurement component, a multidimensional association analysis component, and an activity association analysis component at the same time, the computer device loads the first infectious disease diffusion information map into these three components respectively. Then, the first deviation, the second deviation, and the third deviation of each infectious disease outbreak activity output by these three components are integrated to obtain the fourth deviation of each infectious disease outbreak activity. The weighted average method can be used for integration. Assuming that the first deviation is d1 and the weight is w1; the second deviation is d2 and the weight is w2; the third deviation is d3 and the weight is w3, and w1+w2+w3=1, then the fourth deviation d4=w1×d1+w2×d2+w3×d3. Finally, based on the fourth deviation of each infectious disease outbreak activity, it is evaluated whether it is a source infectious disease outbreak activity. If the fourth deviation of an infectious disease outbreak activity is lower than a certain set threshold, the computer device determines that the infectious disease outbreak activity may be a source infectious disease outbreak activity, otherwise it is not.

[0060] The computer device loads the first infectious disease diffusion information map into the infectious disease source tracing identification network, and uses different components to analyze and integrate, and finally obtains the source tracing identification result.

[0061] As an embodiment, the infectious disease traceability and identification network includes at least one of the following components: a graph element group commonality measurement component, a multidimensional association analysis component, or an activity association analysis component; wherein the graph element group commonality measurement component is used to obtain the commonality measurement between graph elements in infectious disease outbreak activities, the multidimensional association analysis component is used to obtain the commonality measurement between graph elements and infectious disease types in infectious disease outbreak activities, and the activity association analysis component is used to obtain the commonality measurement between infectious disease outbreak activities.

[0062] Specifically, if the infectious disease source tracing identification network includes a graph element group commonality measurement component, then step S200 loads the first infectious disease diffusion information graph into the infectious disease source tracing identification network to obtain a source tracing identification result, including: Step S210: outputting a first deviation of each infectious disease outbreak activity based on the graph element group commonality measurement component, wherein the graph element group commonality measurement component is used to obtain the commonality measurement between the graph element groups in each infectious disease outbreak activity, and obtain the first deviation based on the commonality measurement between the graph element groups in each infectious disease outbreak activity; Step S220: Based on the first deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity, and obtaining a source tracing identification result; If the infectious disease source tracing identification network includes a multi-dimensional association analysis component, step S200, loading the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result, also includes: Step S230: outputting the second deviation of each infectious disease outbreak activity based on the multidimensional association analysis component, wherein the multidimensional association analysis component is used to integrate multiple graph elements in one or more infectious disease outbreak activities to obtain a representative graph element, or to use the core elements of each infectious disease outbreak activity as a representative graph element, and to obtain the second deviation of each infectious disease outbreak activity based on a commonality measurement between the one or more graph elements and the representative graph element; Step S240: Based on the second deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity, and obtaining a source tracing identification result; If the infectious disease source tracing identification network includes an activity association analysis component, then step S200, loading the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result, also includes: Step S250: outputting a third deviation of each infectious disease outbreak activity based on the activity association analysis component, wherein the activity association analysis component is used to obtain a commonality metric between pairs of infectious disease outbreak activities, and determining the third deviation of each infectious disease outbreak activity based on the commonality metric between pairs of infectious disease outbreak activities; Step S260: Based on the third deviation of each infectious disease outbreak activity, evaluate whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain a source tracing identification result.

[0063] In step S210, the computer device outputs the first deviation of each infectious disease outbreak activity based on the graph element group commonality measurement component. The core task of the graph element group commonality measurement component is to obtain the commonality measurement between the graph element groups in each infectious disease outbreak activity, and obtain the first deviation accordingly. In actual operation, taking an influenza epidemic as an example, it is assumed that multiple influenza infectious disease outbreak activities occur in a certain area. Each infectious disease outbreak activity contains multiple graph elements, such as patient graph elements, location graph elements, time graph elements, etc., and the graph element group is a binary group composed of these graph elements. The computer device analyzes the graph element groups in each infectious disease outbreak activity. For example, for a graph element group consisting of a patient graph element A and a location graph element B in an infectious disease outbreak activity, and a graph element group consisting of a similar patient graph element C and a location graph element D in another infectious disease outbreak activity, the computer device calculates the commonality measurement between them. A variety of methods can be used to calculate the commonality measurement, such as using the Euclidean distance formula. Assume that the eigenvector of the graph group A is a=[a1, a2, …, an], and the eigenvector of the graph group B is b=[b1, b2, …, bn], and the Euclidean distance . The smaller the distance, the higher the commonality measure between the groups of graph elements. The computer device performs such calculations on all groups of graph elements in each infectious disease outbreak activity, and obtains the first deviation based on these commonality measure results. If there is a large difference between the commonality measure of the graph element group in a certain infectious disease outbreak activity and the commonality measure of the graph element group in most other infectious disease outbreak activities, then the first deviation of the infectious disease outbreak activity will be higher. In other words, the deviation represents the difference between the commonality measure of each infectious disease outbreak activity and other infectious disease outbreak activities. In terms of technical implementation means, the computer device will first digitally encode the various attribute information of the graph element and convert it into a vector form that can be processed by the computer. Then, using the corresponding calculation library, the graph element group is calculated according to the set commonality measure calculation method, and finally the first deviation of each infectious disease outbreak activity is obtained.

[0064] In step S220, the computer device evaluates whether each infectious disease outbreak activity is a source infectious disease outbreak activity based on the first deviation of each infectious disease outbreak activity, thereby obtaining a traceability identification result. The computer device sets a threshold value, which is determined based on a large amount of historical data and actual experience. Taking the influenza epidemic just now as an example, if the first deviation of a certain infectious disease outbreak activity is lower than the set threshold, the computer device will determine that the infectious disease outbreak activity is likely to be a source infectious disease outbreak activity; conversely, if the first deviation is higher than the threshold, it is determined that the infectious disease outbreak activity is unlikely to be a source infectious disease outbreak activity. For example, if the set threshold is 0.5, and the first deviation of a certain infectious disease outbreak activity is calculated to be 0.3, then the computer device will mark it as a possible source infectious disease outbreak activity. In terms of technical implementation, the computer device only needs to compare the calculated first deviation of each infectious disease outbreak activity with a preset threshold value, and mark it accordingly according to the comparison result, thereby completing a preliminary evaluation of whether each infectious disease outbreak activity is a source infectious disease outbreak activity and obtaining a traceability identification result.

[0065] When the infectious disease source tracing identification network includes a multidimensional association analysis component, step S230 is entered. The computer device outputs the second deviation of each infectious disease outbreak activity based on the multidimensional association analysis component. The multidimensional association analysis component integrates multiple graphs in one or more infectious disease outbreak activities to obtain representative graphs, or uses the core elements of each infectious disease outbreak activity as representative graphs, and obtains the second deviation of each infectious disease outbreak activity based on the commonality measurement between one or more graphs and the representative graphs. Continuing with the influenza epidemic as an example, in a specific infectious disease outbreak activity, multiple graphs such as patient symptom graphs, infection time graphs, and infection location graphs are included. The computer device can integrate these graphs through a clustering algorithm to obtain a representative graph. For example, using the K-means clustering algorithm, these graphs are clustered according to their feature vectors, and the graphs at the cluster center can be used as representative graphs. Then, the computer device calculates the commonality measurement between other graphs and this representative graph. Assuming that cosine similarity is used for calculation, if there is a graph element E and a representative graph element F, whose characteristic vectors are e and f respectively, cosine similarity sim(E, F)=(e·f) / (||e|| ||f||), where e·f is the vector dot product, and ||e|| and ||f|| are the moduli of vectors e and f respectively. By performing such calculations on all graph elements and representative graph elements in an infectious disease outbreak activity, the computer device can obtain the second deviation of the infectious disease outbreak activity. If the commonality measure between the graph element and the representative graph element in an infectious disease outbreak activity is low, then its second deviation will be higher. In technical implementation, the computer device first uses a clustering algorithm to integrate the graph elements to obtain a representative graph element, and then calculates each graph element and the representative graph element according to the selected commonality measure calculation method, and finally determines the second deviation of each infectious disease outbreak activity.

[0066] In step S240, the computer device evaluates whether each infectious disease outbreak activity is a source infectious disease outbreak activity based on the second deviation of each infectious disease outbreak activity, and then obtains the traceability identification result. Similar to step S220, the computer device will also set a threshold. Taking the influenza epidemic as an example, if the second deviation of an infectious disease outbreak activity is lower than the set threshold, the computer device determines that the infectious disease outbreak activity is more likely to be a source infectious disease outbreak activity; if it is higher than the threshold, it is determined that it is unlikely to be a source infectious disease outbreak activity. For example, if the threshold is set to 0.6, and the second deviation of an infectious disease outbreak activity is calculated to be 0.4, then the computer device will regard it as a possible source infectious disease outbreak activity. In terms of technical implementation, the computer device compares the second deviation of each infectious disease outbreak activity with a preset threshold, and marks each infectious disease outbreak activity according to the comparison result, so as to obtain a traceability identification result based on the second deviation evaluation.

[0067] When the infectious disease source tracing identification network includes an activity association analysis component, step S250 is entered. The computer device outputs the third deviation of each infectious disease outbreak activity based on the activity association analysis component. The activity association analysis component must first select and obtain the similar activity set corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities. Taking a hand, foot and mouth disease epidemic as an example, the computer device traverses all infectious disease outbreak activities and determines the similar activity set by analyzing various characteristics of each infectious disease outbreak activity, such as the time of onset, the place of onset, the age of the patient, etc. For example, for a specific infectious disease outbreak activity, the computer device finds that there are several other infectious disease outbreak activities with similar onset times, concentrated in the same area, and similar age groups of the patients, then these infectious disease outbreak activities will be classified as the similar activity set of the specific infectious disease outbreak activity. Then, based on the commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set, the third deviation of each infectious disease outbreak activity is obtained. The computer device can determine the commonality measure by calculating the distance between the feature vectors of two infectious disease outbreak activities, such as using the Manhattan distance formula. Assume that the characteristic vector of infectious disease outbreak activity G is g=[g1, g2, …, gn], the characteristic vector of infectious disease outbreak activity H in the similar activity set is h=[h1, h2, …, hn], and the Manhattan distance d(G, H)=|g1-h1|+|g2-h2|+…+|gn-hn|. By performing such calculations on each infectious disease outbreak activity and all infectious disease outbreak activities in the similar activity set, the computer device can obtain the third deviation of the infectious disease outbreak activity. If the commonality measure between an infectious disease outbreak activity and the similar activity set is low, its third deviation will be high. In terms of technical implementation means, the computer device first determines the similar activity set by analyzing the characteristics of the infectious disease outbreak activity, and then uses the selected commonality measure calculation method to calculate each infectious disease outbreak activity and the activities in the similar activity set, thereby obtaining the third deviation of each infectious disease outbreak activity.

[0068] In step S260, the computer device evaluates whether each infectious disease outbreak activity is a source infectious disease outbreak activity based on the third deviation of each infectious disease outbreak activity, and obtains a traceability identification result. Similarly, the computer device sets a threshold. Taking the hand, foot and mouth disease epidemic as an example, if the third deviation of a certain infectious disease outbreak activity is lower than the set threshold, the computer device determines that the infectious disease outbreak activity may be a source infectious disease outbreak activity; if it is higher than the threshold, it is determined that it is unlikely to be a source infectious disease outbreak activity. For example, if the threshold is set to 0.7, and the third deviation of an infectious disease outbreak activity is calculated to be 0.5, then the computer device will mark it as a possible source infectious disease outbreak activity. In technical implementation, the computer device compares the calculated third deviation of each infectious disease outbreak activity with a preset threshold, and marks each infectious disease outbreak activity according to the comparison result, thereby obtaining a traceability identification result based on the third deviation evaluation.

[0069] The computer device uses different components in the infectious disease source tracing identification network to perform analysis from different angles, such as graph element groups, graph elements and infectious disease types, and infectious disease outbreak activities, to obtain the first deviation, second deviation, and third deviation respectively, and based on these deviations, evaluates whether each infectious disease outbreak activity is a source infectious disease outbreak activity, providing comprehensive and detailed analysis results for the tracing of public health events.

[0070] As an implementation method, the activity association analysis component is specifically used to: select similar activity sets corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities; and obtain a third deviation degree of each infectious disease outbreak activity based on a commonality measurement between each infectious disease outbreak activity and the corresponding infectious disease outbreak activities in the similar activity set.

[0071] Specifically, the above process can be implemented in the following ways: perform context feature extraction on each infectious disease outbreak activity to obtain a coding vector for each infectious disease outbreak activity; determine the commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set based on the coding vector of each infectious disease outbreak activity, the commonality measure being the third deviation.

[0072] The activity association analysis component in the computer device plays an important role in the infectious disease tracing process. It is specifically used for two key tasks: selecting similar activity sets corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities; and obtaining the third deviation degree of each infectious disease outbreak activity based on the commonality measurement between each infectious disease outbreak activity and the corresponding infectious disease outbreak activities in the similar activity set.

[0073] When selecting the similar activity sets corresponding to each infectious disease outbreak, the computer device comprehensively analyzes the various characteristics of multiple infectious disease outbreaks. Taking the multiple local outbreaks that occurred in different regions and time periods during the influenza epidemic as an example, the computer device will first perform detailed feature extraction for each infectious disease outbreak. These features cover multiple dimensions, including time features, such as the start time and duration of the outbreak; location features, such as specific geographical locations (accurate to streets, communities, and even buildings); population characteristics, such as the age and occupational distribution of patients; and disease characteristics, such as symptoms and virus strain types.

[0074] Based on these rich features, the computer device uses a specific algorithm to determine the similarity between different infectious disease outbreaks. One of the commonly used methods is to calculate the distance between feature vectors. For example, the feature of each infectious disease outbreak activity is represented as a feature vector. Let the feature vector of infectious disease outbreak activity A be a=[a1, a2, a3, …, an], where a1 represents the quantitative value of the time feature, a2 represents the quantitative value of the location feature, etc. The feature vector of infectious disease outbreak activity B is b=[b1, b2, b3, …, bn]. The cosine similarity formula can be used to measure the similarity between them, and the formula will not be repeated here.

[0075] The computer device traverses all infectious disease outbreak activities, and for each infectious disease outbreak activity, performs the above-mentioned similarity calculation with all other outbreak activities. Then, according to the set similarity threshold, infectious disease outbreak activities with similarity higher than the threshold are selected to form a similar activity set for the specific infectious disease outbreak activity. For example, the similarity threshold is set to 0.8. For infectious disease outbreak activity C, the computer device finds through calculation that the similarities between infectious disease outbreak activities D, E, and F and C are 0.85, 0.9, and 0.82, respectively, which are all higher than the threshold of 0.8. Then infectious disease outbreak activities D, E, and F constitute the similar activity set of infectious disease outbreak activity C.

[0076] Next, based on the commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set, a third deviation degree of each infectious disease outbreak activity is obtained. The computer device further analyzes the relationship between each infectious disease outbreak activity and the activities in its similar activity set. In this process, context features are first extracted for each infectious disease outbreak activity to obtain a coding vector for each infectious disease outbreak activity.

[0077] Continuing with the influenza epidemic as an example, the computer device extracts contextual features from various aspects of the infectious disease outbreak. In addition to the basic features mentioned above, the characteristics of the transmission path are also considered, such as whether there is clustered transmission, the proportion of family transmission, etc. After these features are comprehensively extracted and sorted, they are converted into a coding vector through a specific coding algorithm. For example, a neural network model in deep learning, such as an autoencoder, can be used to input the characteristics of the infectious disease outbreak into the autoencoder. After being processed by the encoder part, a low-dimensional coding vector is obtained, which can effectively summarize the key information of the infectious disease outbreak.

[0078] After obtaining the coding vectors, the computer device determines the commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set based on the coding vectors. The commonality measure can also be calculated using the cosine similarity formula mentioned above.

[0079] The computer device performs such a commonality measurement calculation for each infectious disease outbreak activity and all activities in the similar activity set. Then, the third deviation of the infectious disease outbreak activity is obtained by a specific calculation method. A feasible method is to calculate the average value or standard deviation of the commonality measurement values ​​of the infectious disease outbreak activity and all activities in the similar activity set. For example, if the commonality measurement values ​​of infectious disease outbreak activity I and infectious disease outbreak activities J, K, and L in the similar activity set are 0.7, 0.6, and 0.8, respectively, then the average value (0.7+0.6+0.8) / 3=0.7 can be calculated as a reference value, and then combined with other factors (such as the degree of dispersion of data, etc.) to calculate the third deviation through a specific formula. Assume that the formula Deviation=|mean-specific_similarity|+k×std is used, where mean is the average value of all commonality measurement values, specific_similarity is the commonality measurement value of the infectious disease outbreak activity and a specific reference activity (such as the most representative activity in the similar activity set), k is a weight coefficient, and std is the standard deviation of all commonality measurement values. Through this formula, multiple factors can be comprehensively considered to obtain a more accurate third deviation.

[0080] Through the two main functions of the activity association analysis component, the computer device can comprehensively and deeply analyze the relationship between infectious disease outbreak activities and obtain the third deviation of each infectious disease outbreak activity, providing strong data support for the subsequent assessment of whether the infectious disease outbreak activity is a source infectious disease outbreak activity based on the deviation.

[0081] As an implementation mode, if the infectious disease source tracing identification network includes a graph element group commonality measurement component, a multidimensional association analysis component, and an activity association analysis component, then step S200, loading the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result, also includes: Step S270: loading the first infectious disease diffusion information graph into the graph element group commonality measurement component, the multidimensional association analysis component and the activity association analysis component respectively; Step S280: Integrate the first deviation of each infectious disease outbreak activity output by the graph element group commonality measurement component, the second deviation of each infectious disease outbreak activity output by the multidimensional association analysis component, and the third deviation of each infectious disease outbreak activity output by the activity association analysis component to obtain a fourth deviation of each infectious disease outbreak activity; Step S290: Based on the fourth deviation degree of each infectious disease outbreak activity, evaluate whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain a source tracing identification result.

[0082] In step S270, the computer device loads the first infectious disease diffusion information map into the commonality measurement component, multidimensional association analysis component and activity association analysis component of the graph element group. Taking an unexpected infectious disease event as an example, assuming that this is the spread of a new respiratory infectious disease, the constructed first infectious disease diffusion information map contains a lot of graph element information about patients, locations, time and transmission routes. The computer device adjusts the format and adapts the data contained in this map according to the input requirements of each component, and transmits it to the commonality measurement component, multidimensional association analysis component and activity association analysis component respectively.

[0083] For the element group commonality measurement component, the computer device accurately loads the element group data in each infectious disease outbreak activity in the atlas. For example, in this new respiratory infectious disease case, each infectious disease outbreak activity may involve an element group consisting of patient elements and location elements. The computer device passes the relevant feature data of these element groups, such as the patient's age, symptom manifestations, the population density of the location, the type of venue, etc., to the element group commonality measurement component so that the component can subsequently calculate the commonality measurement.

[0084] For the multi-dimensional association analysis component, the computer device integrates the information of multiple graph elements in the graph and inputs them. For example, in this case, the graph elements of multiple dimensions, such as the patient's basic information, symptom information, infection time information, and location information, are sorted and loaded into the component in a form suitable for processing. In this way, the multi-dimensional association analysis component can perform cluster analysis on the graph elements, determine representative graph elements, and other operations based on these multi-dimensional information.

[0085] For the activity association analysis component, the computer device loads the comprehensive information of the entire infectious disease outbreak, including the overall characteristics of each activity, the relationship between the involved graph elements, etc. For example, in different outbreaks of new respiratory infectious diseases, the overall scale, spread trend and potential connection between each outbreak and other activities are passed to the activity association analysis component, so that it can select similar activity sets and calculate commonality metrics.

[0086] In step S280, the computer device integrates the first deviation of each infectious disease outbreak activity output by the graph element group commonality measurement component, the second deviation of each infectious disease outbreak activity output by the multidimensional association analysis component, and the third deviation of each infectious disease outbreak activity output by the activity association analysis component to obtain a fourth deviation of each infectious disease outbreak activity.

[0087] In order to realize this integration process, the computer device can adopt a variety of methods, and the feasible one is the weighted average method. Assume that the first deviation is D1, and its weight is W1; the second deviation is D2, and its weight is W2; the third deviation is D3, and its weight is W3, and W1+W2+W3=1. Then the calculation formula of the fourth deviation D4 is: D4=W1×D1+W2×D2+W3×D3.

[0088] Continuing with the example of a new respiratory infectious disease, the computer device obtains the first deviation D1=0.3 of a certain infectious disease outbreak activity A from the graph element group commonality measurement component, obtains the second deviation D2=0.4 from the multidimensional association analysis component, and obtains the third deviation D3=0.5 from the activity association analysis component. Assume that after comprehensive consideration, the weights set for these three deviations are W1=0.3, W2=0.3, and W3=0.4 respectively. Then, according to the above formula, the fourth deviation D4 of the infectious disease outbreak activity A=0.3×0.3+0.3×0.4+0.4×0.5=0.09+0.12+0.2=0.41.

[0089] In step S290, the computer device evaluates whether each infectious disease outbreak activity is a source infectious disease outbreak activity based on the fourth deviation of each infectious disease outbreak activity to obtain a source tracing identification result. The computer device sets a suitable threshold based on a large amount of historical data and actual experience.

[0090] Still taking the new respiratory infectious disease as an example, assuming that the threshold is set to 0.5. For infectious disease outbreak activity A, its fourth deviation is 0.41, which is less than the set threshold of 0.5. Based on this, the computer device determines that infectious disease outbreak activity A is more likely to be the source infectious disease outbreak activity. For another infectious disease outbreak activity B, if its fourth deviation is calculated to be 0.6, which is greater than the threshold of 0.5, then the computer device determines that infectious disease outbreak activity B is unlikely to be the source infectious disease outbreak activity.

[0091] Through steps S270-S290, the computer device makes full use of the deviation information obtained from different components in the infectious disease traceability identification network from different perspectives. The graph element group commonality measurement component provides a first deviation from the perspective of commonality measurement between graph element groups, reflecting the uniqueness of the graph element combination within the infectious disease outbreak activity; the multidimensional association analysis component provides a second deviation from the perspective of the integration of graph elements, infectious disease types and multidimensional features, reflecting the deviation of infectious disease outbreak activities in multidimensional features; the activity association analysis component provides a third deviation from the perspective of the association between infectious disease outbreak activities, showing the degree of difference between a certain infectious disease outbreak activity and other similar activities.

[0092] The fourth deviation obtained by integrating these three deviations combines information from multiple levels and can more comprehensively and accurately characterize the degree of deviation of each infectious disease outbreak activity from other activities and the overall pattern. The assessment of source infectious disease outbreak activities based on the fourth deviation is more accurate and reliable than the analysis results based on a single component.

[0093] As an implementation mode, in step S200, loading the first infectious disease diffusion information map into the infectious disease source tracing identification network includes: Step S201: Projecting each graphic element in each infectious disease outbreak activity in the first infectious disease diffusion information graph into the same feature domain to obtain a second representation vector of each infectious disease outbreak activity in the same feature domain; Step S202: Load the second representation vector into the infectious disease source tracing and identification network to obtain a source tracing and identification result.

[0094] In step S201, the computer device projects each graphic element in each infectious disease outbreak activity in the first infectious disease diffusion information map into the same feature domain to obtain a second representation vector of each infectious disease outbreak activity in the same feature domain. To better understand this process, an outbreak of dengue fever infectious disease is used as an example for explanation.

[0095] Assume that the first infectious disease diffusion information map records multiple dengue fever infectious disease outbreaks, and each outbreak contains patient primitives, location primitives, time primitives, etc. The patient primitive contains information such as the patient's age, gender, and symptoms; the location primitive contains information such as the geographical location and environmental characteristics of the outbreak location; and the time primitive contains information such as the onset time and the duration of the outbreak. These different types of primitives have different feature representations. In order to process and analyze them under a unified framework, the computer device needs to project them into the same feature domain.

[0096] The feature domain is a mathematical space in which different data can be represented and compared in a unified way. The computer device uses a specific mapping function to realize the projection of the primitives to the same feature domain. For example, the linear transformation method can be used. Assuming that the original feature vector of a certain primitive is x=[x1, x2, …, xn], the mapping function can be expressed as y=Wx+b, where W is a weight matrix, b is a bias vector, and y is the vector after projection to the feature domain. The purpose of this mapping process is to transform the features of different types of primitives into comparable forms.

[0097] For the patient graph in the dengue fever epidemic, assuming that its original feature vector x1 contains information such as age, gender, and symptoms, it is projected into the feature domain through the mapping function y1=W1x1+b1 to obtain a new vector y1. For the location graph, its original feature vector x2 contains information such as geographic location and environmental characteristics, and is projected to obtain vector y2 through another mapping function y2=W2x2+b2. Here, W1, W2, b1, and b2 are parameters that are adjusted according to the characteristics of the graph and the requirements of the target feature domain.

[0098] In this way, the computer device projects all the primitives in each infectious disease outbreak activity into the same feature domain. For a specific dengue fever infectious disease outbreak activity, after the patient primitives, location primitives, time primitives, etc. are projected, the representation of these primitives in the feature domain has a unified format and scale, which is convenient for subsequent analysis and processing.

[0099] After step S201, the computer device obtains the primitive vector representation of each infectious disease outbreak activity in the same feature domain. Next, in step S202, the computer device loads these second representation vectors into the infectious disease source tracing identification network to obtain the source tracing identification result.

[0100] Continuing with the dengue fever epidemic as an example, the infectious disease source tracing identification network is a trained model that can analyze and judge the input feature vectors of infectious disease outbreak activities. The computer device sequentially inputs the second representation vectors of each infectious disease outbreak activity in the same feature domain obtained in step S201 into the infectious disease source tracing identification network.

[0101] The infectious disease traceability identification network contains multiple processing layers and computing modules, such as a neural network layer. When the second representation vector is input into the network, it will first pass through the input layer, and then perform a series of calculations and feature extraction in the hidden layer. In the hidden layer, the input vector will be weighted summed, nonlinear transformation and other operations. Assuming that the input vector is z, the weight matrix of the hidden layer is V, the bias vector is c, and the nonlinear function is f, the output of the hidden layer is h=f(Vz+c). Through multiple layers of such calculations and feature extraction, the network can gradually dig out the deep characteristics of infectious disease outbreaks.

[0102] Finally, after the calculation of the output layer, the network will output a result, which is the traceability identification result. For example, the output result may be a probability value, indicating the possibility that the infectious disease outbreak activity is a source infectious disease outbreak activity. If the output probability value is greater than a certain set threshold, the computer device determines that the infectious disease outbreak activity is a source infectious disease outbreak activity; conversely, if the probability value is less than the threshold, it is determined that it is not a source infectious disease outbreak activity.

[0103] Through steps S201-S202, the computer device effectively loads the first infectious disease diffusion information map into the infectious disease source tracing identification network. Step S201 solves the compatibility and comparability of data by projecting different graph elements into the same feature domain, so that different types of information can be processed under a unified framework. Step S202 uses the trained infectious disease source tracing identification network to analyze and judge the processed data, and finally obtains accurate tracing identification results.

[0104] As an implementation method, the training process of the infectious disease source tracing identification network includes the following steps: Step S10: obtaining a second infectious disease diffusion information map, the second infectious disease diffusion information map representing multiple infectious disease outbreak activities, the second infectious disease diffusion information map including multiple graph elements and connecting edges between the multiple graph elements, each graph element in each infectious disease outbreak activity including information constituting a disease monitoring element in each infectious disease outbreak activity; Step S20: Generate an infectious disease source tracing identification network based on the second infectious disease diffusion information map, the infectious disease source tracing identification network is used to identify the source infectious disease outbreak activities among multiple infectious disease outbreak activities, the source infectious disease outbreak activities are determined based on the commonality measurement between infectious disease outbreak activities or the commonality measurement between graph elements in infectious disease outbreak activities.

[0105] In step S10, the computer device obtains a second infectious disease diffusion information map. The map represents multiple infectious disease outbreaks, and also includes multiple graph elements and connecting edges between these graph elements, and each graph element in each infectious disease outbreak contains information constituting the disease monitoring elements in the infectious disease outbreak. Taking a measles outbreak as an example, the computer device collects relevant information from various data sources, which may include hospital case record computer devices, monitoring reports from disease control centers, and statistical data from community health institutions. In these data, each measles infectious disease outbreak is recorded in detail, such as patient information (age, gender, symptoms, etc. as patient graph element information), outbreak location (specific to school, community, etc. as location graph element information), onset time (accurate to date or even hour as time graph element information) and transmission route (air transmission, contact transmission, etc. as transmission route graph element information). The computer device integrates this information to construct a second infectious disease diffusion information map, in which graph elements are represented by connecting edges. For example, the patient graph element is connected to the location graph element through the connecting edge of "appearing at this location", and the patient graph element is connected to the time graph element through the connecting edge of "onset time". In terms of technical implementation, computer devices use data collection tools to extract data from different data sources, and perform data cleaning to remove erroneous and incomplete data. Then, they use graph data structures to organize this information into a graph form and store it in a graph database for subsequent operations.

[0106] In step S20, the computer device generates an infectious disease source tracing identification network based on the second infectious disease diffusion information map, and the network is used to identify the source infectious disease outbreak activity among multiple infectious disease outbreak activities, and the source infectious disease outbreak activity is determined based on the commonality measurement between infectious disease outbreak activities or the commonality measurement between graph elements in infectious disease outbreak activities. The computer device begins to build a network architecture, which may include multiple components to analyze infectious disease outbreak activities from different angles.

[0107] Assume that the infectious disease traceability identification network includes a graph element group commonality measurement component. Taking the measles epidemic map as an example, the computer device first executes step S21 to group multiple graph elements in each infectious disease outbreak activity into at least one group of graph element groups. For example, in a specific measles outbreak activity, the patient graph element and the location graph element are grouped into a graph element group, and the patient graph element and the time graph element may be grouped into another group of graph element groups. In terms of technical implementation, the computer device traverses the graph elements in each infectious disease outbreak activity and generates all possible graph element groups according to preset combination rules.

[0108] Then in step S22, based on the primitive group commonality measurement component, the computer device obtains a first commonality measurement of each primitive group in at least one group of primitive groups. The computer device uses a suitable algorithm to calculate the commonality measurement between primitive groups, such as using a cosine similarity formula.

[0109] In step S23, the commonality metric error value of the image group of each image element in the plurality of images is obtained based on the first commonality metric of each group of image element groups. The computer device obtains a similar image element sample set of the first image element in the plurality of images, and constructs a dissimilar image element sample set. For example, for a patient image element as the first image element, the computer device compares the commonality metric of other patient images elements with the first image element, and classifies the patient images elements with a commonality metric greater than a set threshold value as a similar image element sample set, and classifies the patient images with a commonality metric less than the threshold value as a dissimilar image element sample set. Then, based on the first commonality metric between the first image element and the images in the similar image element sample set, and the commonality metric between the first image element and the images in the dissimilar image element sample set, the commonality metric error value of the image element group corresponding to the first image element is determined. Here we can use the formula Error=Σ(similar-dissimilar)², where similar is the commonality measure between the first primitive and the primitives in the similar primitive sample set, and dissimilar is the commonality measure between the first primitive and the primitives in the dissimilar primitive sample set. The error value is obtained by summing the squares of all such difference values.

[0110] In step S24, the computer device integrates the error values ​​of the commonality metric of the graph groups of multiple graph groups in each infectious disease outbreak activity to obtain a first training error value. The computer device can integrate in a simple summation method or a weighted summation method. Assuming that there are n graph groups in a certain infectious disease outbreak activity, the error value of each graph group is ei, and the weighted summation formula can be expressed as First_error=Σ(wi×ei), where wi is the weight of the error value of each graph group, and different weights can be set for different graph groups according to actual conditions. In this way, the computer device integrates the error values ​​of multiple graph groups into a first training error value, which reflects the degree of difference between the current infectious disease traceability identification network and the expected degree in processing the commonality metric of the graph group. In terms of technical means, the computer device calculates the error value of the graph group in each infectious disease outbreak activity according to the set integration method to obtain a first training error value.

[0111] In step S25, the infectious disease source tracing and identification network is iterated based on the first training error value to obtain an iterated infectious disease source tracing and identification network. The computer device uses an optimization algorithm, such as a stochastic gradient descent algorithm, to adjust the parameters of the network. Assume that the network parameter is θ and the loss function is L(θ), where the loss function is the first training error value. The update formula of the stochastic gradient descent algorithm is θ=θ-η∇L(θ), where η is the learning rate and ∇L(θ) is the gradient of the loss function with respect to the parameter θ. The computer device continuously calculates the gradient and adjusts the parameters according to the learning rate, so that the error of the network in processing the commonality measurement of the graph element group is gradually reduced, thereby obtaining an iterated infectious disease source tracing and identification network, so that it can more accurately identify the source infectious disease outbreak activities.

[0112] When the infectious disease traceability identification network also includes a multidimensional association analysis component, the computer device will also perform related operations on the basis of step S20. In step S26, based on the multidimensional association analysis component, a second commonality measure between one or more of the multiple primitives in each infectious disease outbreak activity and a representative primitive is obtained, and the representative primitive includes the central primitive of each infectious disease outbreak activity or the primitive obtained by integrating multiple primitives. For example, in a certain outbreak activity of a measles epidemic, the computer device analyzes multiple primitives through a clustering algorithm and determines the central primitive as a representative primitive. Then calculate the second commonality measure between other primitives and this representative primitive, and methods such as cosine similarity can also be used. In technical implementation, the computer device uses a clustering algorithm to determine the representative primitive, and then calculates the similarity between the primitives to obtain the second commonality measure.

[0113] In step S27, a second training error value is determined based on a second commonality measure between one or more primitives and a representative primitive. The computer device changes the first primitive into a second primitive, where the first primitive is a primitive in a first infectious disease outbreak activity, and the second primitive has the same attributes as the first primitive, but different categories. For example, a patient primitive (first primitive) with milder symptoms is changed to a patient primitive (second primitive) with more severe symptoms. They are both patient primitive attributes, but belong to different symptom categories. Then the third commonality measure between the second primitive and the representative primitive is obtained, and the training error value is determined based on the second commonality measure and the third commonality measure to obtain the second training error value. The formula Second_error=|second_similarity-third_similarity| can be used, where second_similarity is the second commonality measure between the first primitive and the representative primitive, and third_similarity is the third commonality measure between the second primitive and the representative primitive. In terms of technical means, the computer device performs primitive changes and similarity calculations according to the above steps, and then obtains the second training error value according to the formula.

[0114] In step S28, the infectious disease source tracing identification network is iterated based on the second training error value to obtain an iterated infectious disease source tracing identification network. The computer device also uses an optimization algorithm, such as a stochastic gradient descent algorithm, to adjust network parameters in a manner similar to step S25, so that the network is more accurate in processing the commonality measurement between the primitive and the representative primitive, thereby improving the ability to identify the source infectious disease outbreak activities.

[0115] When the infectious disease source tracing identification network also includes an activity association analysis component, on the basis of step S20, the computer device continues to perform relevant operations. In step S29, similar activity sets and dissimilar activity sets corresponding to each infectious disease outbreak activity are selected from multiple infectious disease outbreak activities. The computer device obtains the number of cross-graphic elements between each group of infectious disease outbreak activities based on the activity association analysis component. For example, for two infectious disease outbreak activities in a measles epidemic, the number of cross-graphic elements such as patient graphics, location graphics, etc. shared by them is calculated. Then one or more infectious disease outbreak activities whose number of cross-graphic elements between the second infectious disease outbreak activity is greater than a preset threshold are obtained to obtain a similar activity set, and the second infectious disease outbreak activity is a random one of the multiple infectious disease outbreak activities. One or more infectious disease outbreak activities whose number of cross-graphic elements between the second infectious disease outbreak activity is not greater than a preset threshold are obtained to obtain a dissimilar activity set. In technical implementation, the computer device traverses all pairs of infectious disease outbreak activities, calculates the number of cross-graphic elements, and divides similar and dissimilar activity sets according to a preset threshold. It can be understood that obtaining relevant data needs to be within the scope permitted by laws and regulations and is carried out on the premise that the parties are informed and authorized.

[0116] In step S2101, a third training error value is determined based on the fourth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the similar activity set and the fifth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the dissimilar activity set. The computer device calculates the similarity between each infectious disease outbreak activity and the activities in the similar activity set and the dissimilar activity set, for example, using the cosine similarity formula. Then, by the formula Third_error=Σ(similar_fourth-dissimilar_fifth)², where similar_fourth is the fourth commonality measure between each infectious disease outbreak activity and the activities in the similar activity set, and dissimilar_fifth is the fifth commonality measure between each infectious disease outbreak activity and the activities in the dissimilar activity set, the squares of these difference values ​​are summed to obtain the third training error value. In terms of technical means, the computer device calculates the similarity between all infectious disease outbreak activities and similar and dissimilar activity sets, and obtains the third training error value according to the formula.

[0117] Finally, in step S2102, the infectious disease source tracing identification network is iterated based on the third training error value to obtain the iterated infectious disease source tracing identification network. The computer device again uses an optimization algorithm, such as a stochastic gradient descent algorithm, to adjust the network parameters so that the network is more accurate in identifying the relationship between infectious disease outbreak activities, and further improves the ability to identify source infectious disease outbreak activities. Through this series of steps, the computer device gradually constructs and optimizes the infectious disease source tracing identification network, so that it can accurately identify the source events in infectious disease outbreak activities, and provide strong support for the tracing of public health events.

[0118] As an implementation method, the infectious disease source tracing identification network includes a graph element group commonality measurement component, which is used to obtain the commonality measurement of the graph element group. Step S20, generating an infectious disease source tracing identification network based on the second infectious disease diffusion information graph, includes: Step S21: Grouping multiple graphic elements in each infectious disease outbreak activity into at least one graphic element group; Step S22: based on the primitive group commonality measurement component, obtaining a first commonality measurement of each primitive group in at least one group of primitive groups; Step S23: obtaining a primitive group commonality metric error value of each primitive in the plurality of primitives based on the first commonality metric of each primitive group; Step S24: integrating the commonality metric error values ​​of the multiple graph groups in each infectious disease outbreak activity to obtain a first training error value; Step S25: Iterating the infectious disease source tracing and identification network based on the first training error value to obtain an iterated infectious disease source tracing and identification network.

[0119] In step S21, the computer device groups multiple graphs in each infectious disease outbreak activity into at least one graph group. Taking an influenza infectious disease outbreak event as an example, it is assumed that there are multiple infectious disease outbreak activities in the influenza epidemic map of a certain area. Each outbreak activity contains multiple graphs, such as patient graphs, which record the patient's age, gender, symptoms and other information; location graphs, which cover information such as whether the place of onset is a school, a shopping mall or a hospital; time graphs, which mark the time of onset, etc. The computer device combines these graphs according to specific rules to form a graph group. For example, the patient graph and the location graph are combined into a group of graph groups, because the association that the patient gets sick at a specific place has important analytical value; the patient graph and the time graph may also be combined into another group of graph groups to analyze the relationship between the onset time and the patient's characteristics. In technical implementation, the computer device traverses all the graphs in each infectious disease outbreak activity, and generates different types of graph groups according to the pre-set combination logic. These combination logics can be based on the field knowledge of infectious disease transmission, such as considering which associations between graphs are important for infectious disease tracing.

[0120] In step S22, the computer device obtains the first commonality measure of each group of graphic elements in at least one group of graphic element groups based on the graphic element group commonality measure component. The computer device adopts a suitable method to calculate the commonality measure between graphic element groups. Taking the influenza epidemic as an example, for the graphic element group composed of patient graphic elements and place graphic elements, the computer device digitally represents the attribute information of these graphic elements to form a feature vector. Assume that the feature vector a of the patient graphic element contains the quantized values ​​of features such as age, gender, and symptoms, and the feature vector b of the place graphic element contains the quantized values ​​of features such as place type and population density. The cosine similarity formula can be used to calculate the first commonality measure between them, which will not be repeated here.

[0121] In step S23, the computer device obtains the commonality metric error value of the primitive group of each primitive in the multiple primitives based on the first commonality metric of each group of primitive groups. First, the computer device obtains a similar primitive sample set of the first primitive in the multiple primitives, and constructs a dissimilar primitive sample set. For example, for a specific patient primitive as the first primitive, the computer device divides the set by comparing the commonality metric of other patient primitives with the first primitive. Assuming that a commonality metric threshold is set to 0.6, other patient primitives with a commonality metric greater than 0.6 with the first primitive are classified as similar primitive sample sets, and patient primitives with a commonality metric less than or equal to 0.6 are classified as dissimilar primitive sample sets. Then, based on the first commonality metric between the first primitive and the primitives in the similar primitive sample set, and the commonality metric between the first primitive and the primitives in the dissimilar primitive sample set, the primitive group commonality metric error value corresponding to the first primitive is determined. Here we can use the formula Error = Σ(similar-dissimilar)², where similar is the commonality measure between the first primitive and the primitives in the sample set of similar primitives, and dissimilar is the commonality measure between the first primitive and the primitives in the sample set of dissimilar primitives. The error value is obtained by summing the squares of all such difference values.

[0122] In step S24, the computer device integrates the commonality metric error values ​​of the graph element groups of multiple graph element groups in each infectious disease outbreak activity to obtain a first training error value. The computer device can use different integration methods, such as simple summation or weighted summation. Assume that there are m graph element groups in a certain infectious disease outbreak activity, and the error value of each graph element group is ei (i=1, 2, …, m). If a simple summation method is used, the first training error value First_error=Σei (i from 1 to m); if a weighted summation method is used, the weight of the error value of each graph element group is wi (i=1, 2, …, m), and Σwi=1, then the weighted summation formula is First_error=Σ(wi×ei) (i from 1 to m). The setting of weights can be determined according to the importance of the graph element or the criticality to the traceability of infectious diseases. For example, a higher weight is given to a graph element group related to the core link of infectious disease transmission. In terms of technical means, the computer device calculates the error values ​​of all graph element groups in each infectious disease outbreak activity according to the set integration method, so as to obtain the first training error value. This error value comprehensively reflects the degree of deviation of the current network from the expected level in processing the commonality measurement of graph element groups.

[0123] In step S25, the computer device iterates the infectious disease source tracing identification network based on the first training error value to obtain the iterated infectious disease source tracing identification network. The computer device uses an optimization algorithm to adjust the parameters of the network to reduce the first training error value. A feasible optimization algorithm such as the stochastic gradient descent algorithm assumes that the network parameter is θ and the loss function is L(θ), where the loss function is the first training error value. The update formula of the stochastic gradient descent algorithm is θ=θ-η∇L(θ), where η is the learning rate, which controls the step size of each parameter update, and ∇L(θ) is the gradient of the loss function to the parameter θ. The computer device continuously calculates the gradient and adjusts the parameters according to the learning rate, so that the error of the network in processing the commonality measurement of the graph element group is gradually reduced. For example, in each iteration, the computer device calculates the first training error value under the current parameters, and then calculates the update direction and step size of the parameters according to the gradient, and updates the network parameters according to the formula. After multiple iterations, the parameters of the network are continuously optimized, thus obtaining the iterative infectious disease source tracing and identification network, which enables it to more accurately analyze and process the commonality measurements of graph element groups, thereby improving the ability to identify source infectious disease outbreaks.

[0124] Through this series of steps, from combining graph elements into graph element groups, to calculating the commonality measurement of graph element groups, determining the error value, and then iterating the network based on the error value, the computer device gradually builds and optimizes the graph element group commonality measurement component in the infectious disease source tracing identification network, laying a solid foundation for accurately identifying the source of infectious disease outbreaks. This process not only involves complex data processing and calculations, but also requires the reasonable setting of parameters and rules based on the knowledge and characteristics of infectious disease transmission to ensure that the network can effectively mine key information from the infectious disease spread information map.

[0125] As an implementation manner, step S23, obtaining a primitive group commonality metric error value of each primitive in a plurality of primitives based on the first commonality metric of each primitive group, includes: Step S231: obtaining a similar primitive sample set of the first primitive from multiple primitives, and constructing a dissimilar primitive sample set, wherein a first commonality measure between the primitives in the similar primitive sample set and the first primitive is greater than a first commonality measure between the primitives in the dissimilar primitive sample set and the first primitive, and the first primitive is a random one of the multiple primitives in each infectious disease outbreak activity; Step S232: Determine a commonality metric error value of a first primitive group based on a first commonality metric between the first primitive and primitives in a similar primitive sample set and a commonality metric between the first primitive and primitives in a dissimilar primitive sample set.

[0126] In step S231, the computer device obtains a similar graphic element sample set of the first graphic element from multiple graphic elements, and constructs a dissimilar graphic element sample set, wherein the first commonality measure between the graphic elements in the similar graphic element sample set and the first graphic element is greater than the first commonality measure between the graphic elements in the dissimilar graphic element sample set and the first graphic element, and the first graphic element is a random one of multiple graphic elements in each infectious disease outbreak activity.

[0127] Taking the infectious disease spread information map of influenza as an example, assuming that in a specific infectious disease outbreak, there are multiple types of graphs, such as patient basic information graphs (including age, gender, occupation, etc.), symptom graphs (fever, cough, fatigue, etc.), infection location graphs (specific hospitals, communities, workplaces, etc.) and infection time graphs (diagnosis time, onset time, etc.). The computer device randomly selects a patient basic information graph as the first graph.

[0128] In order to obtain a similar primitive sample set, the computer device calculates the first commonality measure between the first primitive and all other patient basic information primitives. The cosine similarity formula mentioned above can be used here to calculate the commonality measure. The computer device sets a similarity threshold, such as 0.7. For all other patient basic information primitives, if the cosine similarity with the first primitive is greater than 0.7, they are included in the similar primitive sample set. For example, there is a patient basic information primitive A, whose feature vector is calculated to have a cosine similarity of 0.8 with the first primitive, then primitive A will be added to the similar primitive sample set.

[0129] When constructing the dissimilar primitive sample set, the computer device includes the patient basic information primitives whose cosine similarity with the first primitive is less than or equal to 0.7. For example, if the cosine similarity between the patient basic information primitive B and the first primitive is 0.6, then primitive B belongs to the dissimilar primitive sample set. In this way, the computer device accurately divides the primitives into similar and dissimilar primitive sample sets based on the commonality metric.

[0130] In step S232, the computer device determines a commonality metric error value of a primitive group corresponding to the first primitive based on a first commonality metric between the first primitive and primitives in a similar primitive sample set and a commonality metric between the first primitive and primitives in a dissimilar primitive sample set.

[0131] First, the computer device obtains a sensitivity adjustment factor, which is related to the commonality measure between the first image element and the image element in the sample set of dissimilar images. In actual situations, there may be some images that are essentially dissimilar to the first image element, but due to the noise of the data or the limitations of feature extraction, their commonality measure with the first image element is relatively high. These images are called difficult examples. Difficult examples may interfere with the accurate identification of the source infectious disease outbreak activities, so it is necessary to adjust the attention to them through the sensitivity adjustment factor.

[0132] The calculation method of the sensitivity adjustment factor can be designed according to the actual situation. For example, the following formula can be used: sensitivity_factor=1 / (1+average_dissimilarity), where average_dissimilarity is the average commonality measure between the primitives in the dissimilar primitive sample set and the first primitive. Assuming that there are 5 primitives in the dissimilar primitive sample set, and their commonality measures with the first primitive are 0.5, 0.6, 0.4, 0.55, and 0.65, respectively, then the average commonality measure average_dissimilarity=(0.5+0.6+0.4+0.55+0.65) / 5=0.55, and the sensitivity adjustment factor sensitivity_factor=1 / (1+0.55)≈0.645. It can be seen that the higher the average commonality measure between the dissimilar primitive sample set and the first primitive, the smaller the sensitivity adjustment factor, which means that the computer device pays more attention to these difficult samples and tries to distinguish them from similar primitives.

[0133] Based on the sensitivity adjustment factor, the computer device determines the commonality measurement error value of the image element group corresponding to the first image element. The formula Error=Σ((similar-dissimilar) / sensitivity_factor)² can be used, where similar is the commonality measurement value between the first image element and the image elements in the similar image element sample set, and dissimilar is the commonality measurement value between the first image element and the image elements in the dissimilar image element sample set. For example, the commonality measurement between the first image element and an image element in the similar image element sample set is 0.8, and the commonality measurement with an image element in the dissimilar image element sample set is 0.5, and the sensitivity adjustment factor is 0.645, then the error value of this item is ((0.8-0.5) / 0.645)²≈2.17. The computer device performs such calculations on all image element pairs in the similar image element sample set and the dissimilar image element sample set, and sums all the results to obtain the commonality measurement error value of the image element group corresponding to the first image element.

[0134] Specifically, the computer device first calculates the average commonality measure of the dissimilar primitive sample set and the first primitive, and calculates the factor according to the sensitivity adjustment factor formula. Then, the similar primitive sample set and the dissimilar primitive sample set are traversed, and for each pair of primitives (a similar primitive and a dissimilar primitive), calculations are performed according to the error value formula, and all calculation results are accumulated to finally obtain the primitive group commonality measure error value corresponding to the first primitive.

[0135] Through steps S231-S232, the computer device constructs a sample set of similar and dissimilar graph elements for a random graph element in each infectious disease outbreak activity, and combines the sensitivity adjustment factor to accurately calculate the graph element group commonality metric error value corresponding to the graph element. This process fully considers the complex situation in the data, especially the impact of difficult samples on error calculation, so that the calculated error value more accurately reflects the deviation of the network in processing the commonality metric of the graph element group, and provides reliable data support for the subsequent iterative optimization of the infectious disease source tracing identification network based on these error values, which helps to improve the network's recognition accuracy of the source infectious disease outbreak activities.

[0136] As an implementation manner, step S232, based on the first commonality metric between the first primitive and the primitives in the similar primitive sample set and the commonality metric between the first primitive and the primitives in the dissimilar primitive sample set, determines the primitive group commonality metric error value corresponding to the first primitive group, including: Step S2321: obtaining a sensitivity adjustment factor, where the sensitivity adjustment factor is related to a commonality measure between the first image element and the image elements in the dissimilar image element sample set; Step S2322: Based on the sensitivity adjustment factor, based on the first commonality measurement between the first primitive and the primitives in the similar primitive sample set, and the commonality measurement between the first primitive and the primitives in the dissimilar primitive sample set, determine the primitive group commonality measurement error value corresponding to the first primitive group.

[0137] In step S2321, the computer device obtains a sensitivity adjustment factor, which is related to the commonality measure between the image elements in the dissimilar image element sample set and the first image element. As mentioned above, in some scenarios, there may be a high commonality measure between the dissimilar image element sample and the first image element. Such images are regarded as difficult examples and will interfere with the identification of the source infectious disease outbreak activities. Therefore, the computer device needs to calculate the sensitivity adjustment factor to adjust the attention to the difficult examples.

[0138] Taking a chickenpox outbreak as an example, in the graph element analysis of a specific infectious disease outbreak activity, the computer device has divided the graph elements into a similar graph element sample set and a dissimilar graph element sample set according to step S231. Assume that the first graph element is a patient graph element containing characteristics such as the patient's age and onset time. The graph elements in the dissimilar graph element sample set are also patient graph elements, but under normal circumstances, they are significantly different from the first graph element. However, due to the inaccuracy of data recording or the limitations of feature extraction, some dissimilar graph element samples have a higher commonality measure with the first graph element.

[0139] There are many ways for a computer device to calculate the sensitivity adjustment factor. One feasible way is to determine it based on the average commonality measure between the dissimilar primitive sample set and the first primitive. Assume that there are n primitives in the dissimilar primitive sample set, namely, primitive 1, primitive 2, ..., primitive n, and their commonality measures with the first primitive are d1, d2, ..., dn, respectively. First, calculate the average commonality measure average_dissimilarity=(d1+d2+...+dn) / n. Then, the sensitivity adjustment factor sensitivity_factor=1 / (1+average_dissimilarity).

[0140] For example, there are 5 graph elements in the sample set of dissimilar graph elements, and their commonality measures with the first graph element are 0.4, 0.5, 0.35, 0.45, and 0.55 respectively. Then the average commonality measure average_dissimilarity=(0.4+0.5+0.35+0.45+0.55) / 5=0.45. According to the formula, the sensitivity adjustment factor sensitivity_factor=1 / (1+0.45) ≈0.69. It can be seen that the higher the average commonality measure between the sample set of dissimilar graph elements and the first graph element, the smaller the sensitivity adjustment factor. A smaller sensitivity adjustment factor means that the computer device will pay more attention to these difficult samples when subsequently calculating the commonality measure error of the graph element group, and distinguish the current sample from similar negative samples to improve the accuracy of identifying the source infectious disease outbreak activities.

[0141] Specifically, the computer device first traverses the sample set of dissimilar primitives and calculates the commonality measure between each primitive and the first primitive. The calculation can be performed using the previously introduced method such as cosine similarity. Then, the average commonality measure is calculated according to the above formula, and the sensitivity adjustment factor is obtained. This process can be performed by writing a loop statement to traverse the primitives in the set, and using the mathematical calculation library to perform similarity calculation and subsequent formula operations.

[0142] In step S2322, the computer device determines the commonality metric error value of the graphic element group corresponding to the first graphic element based on the sensitivity adjustment factor, the first commonality metric between the first graphic element and the graphic elements in the similar graphic element sample set, and the commonality metric between the first graphic element and the graphic elements in the dissimilar graphic element sample set.

[0143] Still taking the outbreak of chickenpox as an example, the computer device has obtained the sensitivity adjustment factor. For the first image element, its first commonality measure with image element A in the similar image element sample set is 0.8, and its commonality measure with image element B in the dissimilar image element sample set is 0.4. The computer device uses the formula Error=Σ((similar-dissimilar) / sensitivity_factor)² to calculate the error value. Here, similar represents the commonality measure between the first image element and the image elements in the similar image element sample set, and dissimilar represents the commonality measure between the first image element and the image elements in the dissimilar image element sample set.

[0144] In this example, the calculation is ((0.8-0.4) / 0.69)²≈0.33. The computer device performs such calculations for all pairs of primitives in the similar primitive sample set and the dissimilar primitive sample set. For example, if there are m primitives in the similar primitive sample set and p primitives in the dissimilar primitive sample set, there will be a total of m×p such calculations. The computer device accumulates all these calculation results and finally obtains the primitive group commonality metric error value corresponding to the first primitive.

[0145] Through steps S2321-S2322, the computer device can effectively handle the impact of difficult samples on the calculation of the commonality metric error of the graph element group. Step S2321 quantifies the degree of attention to difficult samples by calculating the sensitivity adjustment factor, so that the calculation strategy can be reasonably adjusted under different data conditions. Based on this factor, step S2322 combines the commonality metric of the set of similar and dissimilar graph element samples with the first graph element to accurately calculate the error value of the commonality metric of the graph element group. These two steps work closely together to provide more accurate error data for the training of the infectious disease source tracing and identification network, which helps the network to better learn and optimize in subsequent iterations, and improve the ability to identify source infectious disease outbreaks.

[0146] As an implementation method, the infectious disease source tracing identification network also includes a multidimensional association analysis component, which is used to perform cluster analysis on the graph elements in the infectious disease outbreak activities to obtain at least one category, and obtain the commonality measurement between each graph element in the infectious disease outbreak activities and at least one category, and the commonality measurement represents the deviation of the corresponding infectious disease outbreak activities. Step S20, generating an infectious disease source tracing identification network based on the second infectious disease diffusion information map, also includes: Step S26: obtaining a second commonality measure between one or more of the multiple graph elements in each infectious disease outbreak activity and a representative graph element based on the multidimensional association analysis component, wherein the representative graph element includes a central graph element of each infectious disease outbreak activity or a graph element obtained by integrating multiple graph elements; Step S27: determining a second training error value based on a second commonality measure between the one or more primitives and the representative primitive; Step S28: Iterating the infectious disease source tracing and identification network based on the second training error value to obtain an iterated infectious disease source tracing and identification network.

[0147] In step S26, the computer device obtains a second commonality measure between one or more of the multiple graphs in each infectious disease outbreak activity and a representative graph based on the multidimensional association analysis component, wherein the representative graph includes the central graph of each infectious disease outbreak activity or a graph obtained by integrating multiple graphs.

[0148] Take a hand, foot and mouth disease outbreak as an example. Assume that there are multiple outbreaks of infectious diseases in the hand, foot and mouth disease epidemic map in a certain area. Each outbreak contains multiple elements, such as patient elements (including age, gender, symptoms, etc.), location elements (such as kindergartens, schools, home addresses, etc.), and time elements (onset time, first report time, etc.). The computer device uses a multidimensional association analysis component to determine the representative element for each infectious disease outbreak.

[0149] For a specific hand, foot and mouth disease outbreak, the computer device may analyze multiple primitives through a clustering algorithm to determine the central primitive as a representative primitive. For example, the K-means clustering algorithm is used to cluster all primitives in the outbreak according to their feature vectors. Assuming that the feature vectors of patient primitives, location primitives, time primitives, etc. are comprehensively considered, after multiple iterations of clustering, the primitive at the cluster center is found. This cluster center primitive combines the main features of multiple primitives in the outbreak and can be used as a representative primitive.

[0150] The computer device then calculates a second commonality measure between one or more primitives and the representative primitive. Still taking the hand, foot and mouth disease outbreak activity as an example, a patient primitive is selected to calculate the second commonality measure between it and the representative primitive. The cosine similarity formula can be used for calculation.

[0151] In step S27, the computer device determines a second training error value based on a second commonality measure between the one or more primitives and the representative primitive. Specifically, the computer device changes the first primitive into a second primitive, the first primitive being a primitive in the first infectious disease outbreak activity, and the second primitive having the same attributes as the first primitive but different categories.

[0152] Continuing with the example of the hand, foot and mouth disease outbreak, suppose the first primitive is a patient primitive with mild symptoms, which contains information such as the patient's age, gender, symptoms, etc. The computer device changes it into a second primitive, which is also a patient primitive, but with more severe symptoms, that is, it has the same patient primitive attributes as the first primitive, but belongs to a different symptom category.

[0153] Then, the computer device obtains a third commonality measure between the second primitive and the representative primitive. The cosine similarity formula is also used for calculation. Assuming that the feature vector of the second primitive is c=[c1, c2, …, cn], and the feature vector of the representative primitive is still b=[b1, b2, …, bn], the third commonality measure sim(c, b)=(c·b) / (||c|| ||b||).

[0154] The training error value is determined based on the second commonality metric and the third commonality metric to obtain the second training error value. The formula Second_error=|second_similarity-third_similarity| can be used, where second_similarity is the second commonality metric between the first primitive and the representative primitive, and third_similarity is the third commonality metric between the second primitive and the representative primitive. For example, if the second commonality metric between the first primitive and the representative primitive is 0.7, and the third commonality metric between the second primitive and the representative primitive is 0.4, then the second training error value Second_error=|0.7-0.4|=0.3.

[0155] In step S28, the computer device iterates the infectious disease source tracing identification network based on the second training error value to obtain an iterated infectious disease source tracing identification network. The computer device uses an optimization algorithm, such as a stochastic gradient descent algorithm, to adjust the parameters of the network.

[0156] Assume that the parameter of the infectious disease traceability network is θ and the loss function is L(θ). The loss function here is the second training error value. The update formula of the stochastic gradient descent algorithm is θ=θ-η∇L(θ), where η is the learning rate, which controls the step size of each parameter update, and ∇L(θ) is the gradient of the loss function with respect to the parameter θ.

[0157] The computer device continuously calculates the gradient and adjusts the parameters according to the learning rate, making the network more accurate in processing the commonality measurement between the primitive and the representative primitive. For example, in each iteration, the computer device calculates the second training error value under the current parameters, calculates the update direction and step size of the parameters according to the formula of the stochastic gradient descent algorithm, and then updates the parameters of the network. After multiple iterations, the parameters of the network are continuously optimized, and the iterated infectious disease tracing and identification network is obtained.

[0158] Through steps S26-S28, the computer device uses a multidimensional association analysis component to determine representative graph elements, calculate the commonality metrics between graph elements and representative graph elements, determine the second training error value based on these metrics, and finally iteratively optimize the network, so that the infectious disease source tracing and identification network can better learn the relationship and characteristics between graph elements and improve the ability to identify source infectious disease outbreak activities.

[0159] As an implementation mode, the multidimensional association analysis component is further used to perform cluster analysis on multiple graph elements in each infectious disease outbreak activity to obtain no less than one category; step S27, determining a second training error value based on a second commonality measure between one or more graph elements and a representative graph element, includes: Step S271: changing the first graphic element into a second graphic element, the first graphic element being a graphic element in the first infectious disease outbreak activity, the second graphic element having the same attribute as the first graphic element but different categories; Step S272: Obtain a third commonality measure between the second primitive and the representative primitive; Step S273: Determine a training error value based on the second commonality metric and the third commonality metric to obtain a second training error value.

[0160] In step S271, the computer device changes the first primitive into a second primitive, the first primitive being a primitive in the first infectious disease outbreak activity, and the second primitive having the same attributes as the first primitive, but different categories. Taking an influenza infectious disease outbreak activity as an example, assuming that in the information map of a certain influenza outbreak event, there are multiple types of primitives, such as patient basic information primitives, symptom manifestation primitives, infection location primitives, etc. The computer device selects a patient basic information primitive as the first primitive, which contains attribute information such as the patient's age, gender, and occupation.

[0161] The computer device changes the first graph element into the second graph element according to specific rules. The formulation of the rules is based on the domain knowledge of infectious disease transmission and a deep understanding of the characteristics of the graph elements. For example, in this influenza case, if the patient's occupation corresponding to the first graph element is a teacher, the computer device may change it to the second graph element whose occupation is a student. Both belong to the attribute category of the patient's basic information graph element, but the specific occupations are different, that is, the categories are inconsistent. This change is intended to examine the changes in the commonality metrics between the graph element and the representative graph element from different angles, so as to more comprehensively evaluate the network's ability to handle different feature situations.

[0162] For example, the computer device first classifies and features the graphic elements in each infectious disease outbreak activity. After the first graphic element is selected, the preset rule library is used to search for graphic element information with the same attributes as the first graphic element but different categories, thereby generating a second graphic element. The rule library contains change rules for different types of graphic elements. For example, for the patient basic information graphic element, the change relationship between different occupations, age groups, etc. may be specified. This process can be achieved by writing conditional judgments and data retrieval statements to ensure that the generated second graphic element meets the requirements.

[0163] In step S272, the computer device obtains the third commonality measure between the second primitive and the representative primitive. In the above influenza case, the computer device has determined the second primitive (the basic information primitive of the patient whose occupation is student). Next, the third commonality measure between the second primitive and the previously determined representative primitive is obtained.

[0164] The representative graph is obtained by integrating multiple graphs in the infectious disease outbreak or determining the central graph. Assume that in this influenza outbreak, a representative graph that integrates multiple key features such as patient information, infection location, and time is determined by a clustering algorithm. The computer device digitally encodes the feature information of the second graph into a feature vector, set as vector c=[c1, c2, …, cn], and the feature vector of the representative graph is set as vector b=[b1, b2, …, bn].

[0165] The computer device uses a suitable algorithm to calculate the commonality measure between them. Here, the cosine similarity formula is still used as an example. The third commonality measure sim(c, b)=(c·b) / (||c|| ||b||), where c·b=c1×b1+c2×b2+…+cn×bn, ||c||=√(c1²+c2²+…+cn²), ||b||=√(b1²+b2²+…+bn²). Through this formula, the computer device can accurately calculate the third commonality measure between the second primitive and the representative primitive.

[0166] In step S273, the computer device determines the training error value based on the second commonality metric and the third commonality metric to obtain a second training error value. Assume that in the influenza case, the second commonality metric between the first element (the basic information element of the patient whose occupation is a teacher) and the representative element is second_similarity=0.6, and the third commonality metric between the second element (the basic information element of the patient whose occupation is a student) and the representative element is third_similarity=0.4.

[0167] The computer device uses the formula Second_error=|second_similarity-third_similarity| to calculate the second training error value. Substituting the above value into the formula, it can be obtained that Second_error=|0.6-0.4|=0.2. This second training error value reflects the difference in the network when processing the commonality measurement between the primitives of different categories but the same attributes and the representative primitives.

[0168] Through steps S271-S273, the computer device deeply explores the changes in the commonality measure between the primitive and the representative primitive under different primitive categories. By changing the first primitive to a second primitive with the same attributes but a different category, the scope of examination of primitive features is expanded. The third commonality measure between the second primitive and the representative primitive is calculated and compared with the second commonality measure between the first primitive and the representative primitive. The second training error value obtained can more comprehensively and meticulously reflect the accuracy of the infectious disease source tracing and identification network in processing primitive relationships. This error value provides an important basis for subsequent network iterations, which helps the computer device adjust network parameters so that the network can better learn and distinguish the relationship between different primitives and representative primitives, thereby improving the ability to identify source infectious disease outbreaks.

[0169] As an implementation method, the infectious disease source tracing identification network further includes an activity association analysis component, which is used to obtain commonality metrics between infectious disease outbreak activities; step S20, generating an infectious disease source tracing identification network based on the second infectious disease diffusion information map, also includes: Step S29: selecting from a plurality of infectious disease outbreak activities a similar activity set and a dissimilar activity set corresponding to each infectious disease outbreak activity; Step S2101: determining a third training error value based on a fourth commonality measure between each infectious disease outbreak activity and infectious disease outbreak activities in a similar activity set and a fifth commonality measure between each infectious disease outbreak activity and infectious disease outbreak activities in a dissimilar activity set; Step S2102: Iterate the infectious disease source tracing and identification network based on the third training error value to obtain an iterated infectious disease source tracing and identification network.

[0170] In step S29, the computer device selects similar activity sets and dissimilar activity sets corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities. Taking the outbreak of different regions during an influenza outbreak as an example, it is assumed that there are multiple infectious disease outbreak activities distributed in different communities, offices, etc. The computer device first obtains the number of cross-graphs between each group of infectious disease outbreak activities based on the activity association analysis component.

[0171] For every two infectious disease outbreak activities, the computer device carefully compares the primitives they contain and counts the number of shared primitives. For example, in infectious disease outbreak activities A and B, activity A involves patient primitives P1, P2, P3, location primitive L1, and time primitive T1; activity B involves patient primitives P2, P4, location primitive L2, and time primitive T1. By comparison, the computer device finds that they share patient primitives P2 and time primitive T1, that is, the number of cross primitives is 2.

[0172] Next, the computer device obtains one or more infectious disease outbreak activities whose number of cross-graph elements with the second infectious disease outbreak activity is greater than a preset threshold, and obtains a similar activity set. Assuming the preset threshold is 1, an infectious disease outbreak activity C is randomly selected as the second infectious disease outbreak activity. The computer device traverses all other infectious disease outbreak activities and calculates the number of cross-graph elements between them and activity C. If the number of cross-graph elements between infectious disease outbreak activity D and activity C is 3, which is greater than the threshold 1, then activity D is included in the similar activity set of activity C.

[0173] At the same time, the computer device obtains one or more infectious disease outbreak activities whose number of cross-graph elements with the second infectious disease outbreak activity is not greater than a preset threshold, and obtains a dissimilar activity set. For example, the number of cross-graph elements between infectious disease outbreak activity E and activity C is 1, which is equal to the threshold. According to the rule, activity E is included in the dissimilar activity set of activity C.

[0174] For example, a computer device can use data traversal and matching algorithms to implement this process. First, a data structure is constructed to store each infectious disease outbreak activity and the information of the graph elements it contains. Then, all pairs of infectious disease outbreak activities are traversed through nested loops, and the graph elements of the two activities are compared in the inner loop to count the number of intersecting graph elements. According to the preset threshold, the qualified activities are classified into similar activity sets and dissimilar activity sets respectively. This process can be implemented through conditional judgment statements.

[0175] In step S2101, the computer device determines a third training error value based on a fourth commonality measure between each infectious disease outbreak activity and infectious disease outbreak activities in a similar activity set and a fifth commonality measure between each infectious disease outbreak activity and infectious disease outbreak activities in a dissimilar activity set.

[0176] The computer device calculates the fourth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the similar activity set. Still taking influenza as an example, for the infectious disease outbreak activity F and its similar activity set {activity G, activity H}, the computer device represents the characteristics of each activity as a feature vector. Assume that the feature vector of activity F is f=[f1, f2, …,fn], and the feature vector of activity G is g=[g1, g2, …, gn].

[0177] The computer device uses a suitable algorithm to calculate the fourth commonality measure between them, such as the cosine similarity formula, the fourth commonality measure sim(f, g)=(f·g) / (||f|| ||g||), where f·g=f1×g1+f2×g2+…+fn×gn, ||f||=√(f1²+f2²+…+fn²), ||g||=√(g1²+g2²+…+gn²). The computer device performs such calculations on activity F and each activity in the similar activity set to obtain a set of fourth commonality measure values.

[0178] Next, the computer device calculates the fifth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the dissimilar activity set. For example, for activity F and its dissimilar activity set {activity I, activity J}, the feature vector of activity I is also set to i=[i1, i2, …, in], and the fifth commonality measure sim(f, i) between activity F and activity I is calculated according to the above cosine similarity formula. The calculation is performed for activity F and each activity in the dissimilar activity set to obtain a set of fifth commonality measure values.

[0179] Based on these commonality metrics, the computer device determines a third training error value. The formula Third_error=Σ(similar_fourth-dissimilar_fifth)² can be used, where similar_fourth is the fourth commonality metric between each infectious disease outbreak activity and the activities in the similar activity set, and dissimilar_fifth is the fifth commonality metric between each infectious disease outbreak activity and the activities in the dissimilar activity set. For example, the fourth commonality metric between activity F and activity G is 0.8, and the fifth commonality metric with activity I is 0.3, and the calculation of this term is (0.8-0.3)²=0.25. The computer device performs such calculations on all combinations of infectious disease outbreak activities and similar and dissimilar activity sets, and accumulates the results to obtain the third training error value.

[0180] In step S2102, the computer device iterates the infectious disease source tracing identification network based on the third training error value to obtain an iterated infectious disease source tracing identification network. The computer device uses an optimization algorithm, such as a stochastic gradient descent algorithm, to adjust the parameters of the network.

[0181] By continuously calculating the gradient and adjusting the parameters according to the learning rate, the network can be more accurate in processing the relationship between infectious disease outbreak activities. For example, in each iteration, the computer device calculates the third training error value under the current parameters, calculates the update direction and step size of the parameters according to the formula of the stochastic gradient descent algorithm, and then updates the parameters of the network. After multiple iterations, the parameters of the network are continuously optimized, and the iterated infectious disease traceability identification network is obtained.

[0182] Through steps S29-S2102, the computer device comprehensively analyzes the similarities and differences between infectious disease outbreak activities. By determining similar activity sets and dissimilar activity sets, the computer device grasps the degree of association between different activities from a macro level. The fourth and fifth commonality metrics are calculated, and the third training error value is determined based on this, providing the network with quantitative feedback on the accuracy of activity relationship processing. Finally, by iterating the network, the network can better capture the intrinsic connection between infectious disease outbreak activities and improve the ability to identify source infectious disease outbreak activities.

[0183] As an implementation mode, step S29, selecting a similar activity set and a dissimilar activity set corresponding to each infectious disease outbreak activity from multiple infectious disease outbreak activities, includes: Step S291: obtaining the number of cross-element graphs between each group of infectious disease outbreak activities based on the activity association analysis component; Step S292: obtaining one or more infectious disease outbreak activities whose number of intersection graph elements with the second infectious disease outbreak activity is greater than a preset threshold, and obtaining a similar activity set, wherein the second infectious disease outbreak activity is a random infectious disease outbreak activity among the multiple infectious disease outbreak activities; Step S293: Acquire one or more infectious disease outbreak activities whose number of crossover elements with the second infectious disease outbreak activity is not greater than a preset threshold, and obtain a dissimilar activity set.

[0184] In step S291, the computer device obtains the number of cross-element graphs between each group of infectious disease outbreak activities based on the activity association analysis component. Taking a hepatitis B infectious disease outbreak event as an example, there are multiple infectious disease outbreak activities in the relevant infectious disease diffusion information map. These activities involve different patient groups, infection locations, infection times and other factors, which are presented in the form of graphs. For example, there is an infectious disease outbreak activity A, involving patient graphs P1, P2, and P3, location graph L1 (a liver disease department in a hospital), and time graph T1 (a specific one-week period); infectious disease outbreak activity B, involving patient graphs P2 and P4, location graph L2 (a community health service center), and time graph T2 (a time period that partially overlaps with T1); infectious disease outbreak activity C, involving patient graphs P5 and P6, location graph L3 (a factory medical room), and time graph T3 (a time period that does not overlap with both T1 and T2).

[0185] The computer device uses the activity association analysis component to carefully compare the various types of elements contained in each group of infectious disease outbreak activities. For activities A and B, the computer device checks the patient element, location element, and time element. It is found that they share the patient element P2. Although the location elements L1 and L2 are different, the time elements T1 and T2 overlap partially. Based on this, it is determined that the two activities have a certain association, and the number of cross-element statistics is 2 (the overlapping part of the patient element P2 and the time element is regarded as a connection point). For activities A and C, after comparison, it is found that their patient elements, location elements, and time elements have no direct sharing or obvious association, and the number of cross-element statistics is 0.

[0186] The computer device can construct a data structure to store the primitive information contained in each infectious disease outbreak activity. For example, a hash table or linked list structure can be used to associate and store the activity identifier with the corresponding primitive set. Then, all pairs of infectious disease outbreak activities are traversed through nested loops. In the inner loop, for each pair of activities, the different types of primitives contained in them are compared one by one. For patient primitives, they can be matched through the patient's unique identifier; for location primitives, they can be matched through geographic location information or place identifiers; for time primitives, they can be matched through overlapping time ranges. By writing detailed matching algorithms and conditional judgment statements, the number of intersecting primitives between each group of infectious disease outbreak activities can be accurately counted, and the results are stored in the corresponding data structure for use in subsequent steps.

[0187] In step S292, the computer device obtains one or more infectious disease outbreak activities whose number of crossover elements with the second infectious disease outbreak activity is greater than a preset threshold, and obtains a similar activity set. Still taking the hepatitis B infectious disease outbreak event as an example, it is assumed that the computer device randomly selects infectious disease outbreak activity B as the second infectious disease outbreak activity, and sets the preset threshold to 1.

[0188] The computer device traverses all other infectious disease outbreak activities and compares them with activity B in terms of the number of cross-element graphs. As mentioned above, the number of cross-element graphs between activity A and activity B is 2, which is greater than the preset threshold of 1, so activity A is identified by the computer device as an activity similar to activity B. However, the number of cross-element graphs between activity C and activity B is 0, which is less than the preset threshold and does not meet the condition. After comparing all infectious disease outbreak activities one by one, the computer device collects activities A whose number of cross-element graphs with activity B is greater than 1 to form a similar activity set of activity B.

[0189] After completing step S291, the computer device has obtained the data on the number of cross-element numbers between all activity pairs. At this time, the computer device traverses these data again, and uses the selected second infectious disease outbreak activity as a benchmark to filter out activities with a number of cross-element numbers greater than a preset threshold through conditional judgment statements. A loop structure can be used to traverse the records of the number of cross-element numbers of all activities. When it is found that the number of cross-element numbers between an activity and the second infectious disease outbreak activity meets the condition, the identifier of the activity is added to a list or set data structure specifically used to store similar activity sets. This process can be achieved by writing simple loops and conditional judgment codes to ensure accurate acquisition of similar activity sets.

[0190] In step S293, the computer device obtains one or more infectious disease outbreak activities whose number of crossover elements with the second infectious disease outbreak activity is not greater than a preset threshold, and obtains a dissimilar activity set. Continuing with the hepatitis B infectious disease outbreak event as an example, activity B is still used as the second infectious disease outbreak activity, and the preset threshold is 1.

[0191] When the computer device traverses the number of intersections between all infectious disease outbreak activities and activity B, it finds that the number of intersections between activity C and activity B is 0, which is not greater than the preset threshold. Similarly, if there are other activities whose number of intersections with activity B is 1 (equal to the preset threshold) or less than 1, these activities meet the conditions. The computer device collects these activities, such as activity C and other activities that meet the conditions, to form a dissimilar activity set of activity B.

[0192] Similar to step S292, the computer device traverses the existing cross-element number data again. With the second infectious disease outbreak activity as a reference, the conditional judgment statement is used to filter out activities whose cross-element number is not greater than the preset threshold. The loop structure is used to traverse the cross-element number records of all activities. When the cross-element number of an activity meets the condition with the second infectious disease outbreak activity, the identifier of the activity is added to the list or set data structure used to store dissimilar activity sets. This process ensures the accurate construction of dissimilar activity sets by writing corresponding code logic.

[0193] Through steps S291-S293, the computer device accurately divides each infectious disease outbreak activity into similar activity sets and dissimilar activity sets. Step S291 provides basic data for subsequent screening through detailed primitive comparison and statistics; steps S292 and S293 accurately classify activities into similar and dissimilar activity sets based on preset thresholds. This process not only takes into account the direct correlation between infectious disease outbreak activities (reflected by the number of cross-element graphs), but also through the flexible parameter setting of preset thresholds, it can be adjusted according to the characteristics and actual needs of different infectious diseases, so as to classify activities more scientifically. These similar activity sets and dissimilar activity sets provide an important basis for the subsequent calculation of commonality metrics between infectious disease outbreak activities and the determination of training error values, which helps the infectious disease source tracing and identification network to better learn and distinguish the relationship between different activities and improve the ability to identify source infectious disease outbreak activities.

[0194] An embodiment of the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and when the processor executes the program, some or all of the steps in the above method are implemented.

[0195] The embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, some or all of the steps in the above method are implemented. The computer-readable storage medium can be transient or non-transient.

[0196] Figure 2 A hardware entity diagram of a computer device provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the hardware entity of the computer device 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program that can be run on the processor 1001, and the processor 1001 implements the steps in the method of any of the above embodiments when executing the program.

Claims

1. A method for tracing the source of public health events, characterized in that: include: Obtaining a first infectious disease diffusion information map, wherein the first infectious disease diffusion information map represents one or more infectious disease outbreak activities, the first infectious disease diffusion information map comprises a plurality of graph elements and connecting edges between the plurality of graph elements, each infectious disease outbreak activity is represented based on a plurality of graph elements among the plurality of graph elements and the connecting edges between the plurality of graph elements, and each graph element in each infectious disease outbreak activity comprises information constituting a disease monitoring element in each infectious disease outbreak activity; The first infectious disease diffusion information map is loaded into the infectious disease source tracing identification network to obtain a source tracing identification result, wherein the source tracing identification result characterizes whether the one or more infectious disease outbreak activities include source infectious disease outbreak activities, and the source infectious disease outbreak activities are determined based on the commonality measurement between infectious disease outbreak activities or the commonality measurement between graph elements in infectious disease outbreak activities.

2. The method according to claim 1, characterized in that The infectious disease source tracing identification network includes at least one of the following components: a graph element group commonality measurement component, a multidimensional association analysis component, or an activity association analysis component; wherein the graph element group commonality measurement component is used to obtain the commonality measurement between graph elements in infectious disease outbreak activities, the multidimensional association analysis component is used to obtain the commonality measurement between graph elements and infectious disease types in infectious disease outbreak activities, and the activity association analysis component is used to obtain the commonality measurement between infectious disease outbreak activities; If the infectious disease source tracing identification network includes the graph element group commonality measurement component, the step of loading the first infectious disease diffusion information graph into the infectious disease source tracing identification network to obtain a source tracing identification result includes: Outputting a first deviation of each infectious disease outbreak activity based on the graph element group commonality measurement component, wherein the graph element group commonality measurement component is used to obtain a commonality measurement between graph element groups in each infectious disease outbreak activity, and obtaining the first deviation based on the commonality measurement between graph element groups in each infectious disease outbreak activity; Based on the first deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity, and obtaining the source tracing identification result; If the infectious disease source tracing identification network includes the multi-dimensional association analysis component, the step of loading the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result further includes: Outputting the second deviation of each infectious disease outbreak activity based on the multidimensional association analysis component, wherein the multidimensional association analysis component is used to integrate multiple graph elements in the one or more infectious disease outbreak activities to obtain a representative graph element, or to use the core elements of each infectious disease outbreak activity as the representative graph element, and to obtain the second deviation of each infectious disease outbreak activity based on the commonality measurement between the one or more graph elements and the representative graph element; Based on the second deviation of each infectious disease outbreak activity, evaluating whether each infectious disease outbreak activity is a source infectious disease outbreak activity, and obtaining the source tracing identification result; If the infectious disease source tracing identification network includes the activity association analysis component, then the step of loading the first infectious disease diffusion information map into the infectious disease source tracing identification network to obtain a source tracing identification result further includes: Outputting a third degree of deviation of each infectious disease outbreak activity based on the activity association analysis component, wherein the activity association analysis component is used to obtain a commonality metric between pairs of infectious disease outbreak activities, and determining the third degree of deviation of each infectious disease outbreak activity based on the commonality metric between the pairs of infectious disease outbreak activities; Based on the third deviation degree of each infectious disease outbreak activity, it is evaluated whether each infectious disease outbreak activity is a source infectious disease outbreak activity to obtain the source tracing identification result.

3. The method according to claim 2, characterized in that The activity association analysis component is specifically used to: select from the multiple infectious disease outbreak activities to obtain a similar activity set corresponding to each infectious disease outbreak activity; based on the commonality measurement between each infectious disease outbreak activity and the infectious disease outbreak activities in the corresponding similar activity set, obtain a third deviation degree of each infectious disease outbreak activity; Specifically, it includes: extracting context features of each infectious disease outbreak activity to obtain a coding vector for each infectious disease outbreak activity; and determining a commonality measure between each infectious disease outbreak activity and infectious disease outbreak activities in a corresponding similar activity set based on the coding vector of each infectious disease outbreak activity, wherein the commonality measure is the third deviation.

4. The method according to claim 3, characterized in that If the infectious disease source tracing identification network includes the graph element group commonality measurement component, the multidimensional association analysis component and the activity association analysis component, then the first infectious disease diffusion information map is loaded into the infectious disease source tracing identification network to obtain the source tracing identification result, and further includes: Loading the first infectious disease diffusion information graph into the graph element group commonality measurement component, the multidimensional association analysis component and the activity association analysis component respectively; Integrate the first deviation of each infectious disease outbreak activity output by the graph element group commonality measurement component, the second deviation of each infectious disease outbreak activity output by the multidimensional association analysis component, and the third deviation of each infectious disease outbreak activity output by the activity association analysis component to obtain a fourth deviation of each infectious disease outbreak activity; Based on the fourth deviation degree of each infectious disease outbreak activity, assess whether each infectious disease outbreak activity is a source infectious disease outbreak activity, and obtain the source tracing identification result; The step of loading the first infectious disease diffusion information map into the infectious disease source tracing identification network includes: Projecting each graphic element in each infectious disease outbreak activity in the first infectious disease diffusion information graph to the same feature domain to obtain a second representation vector of each infectious disease outbreak activity in the same feature domain; The second characterization vector is loaded into the infectious disease source tracing and identification network to obtain the source tracing and identification result.

5. The method according to any one of claims 2 to 4, characterized in that: The training process of the infectious disease source tracing identification network includes: Acquire a second infectious disease diffusion information map, wherein the second infectious disease diffusion information map represents a plurality of infectious disease outbreak activities, the second infectious disease diffusion information map comprises a plurality of graph elements and connecting edges between the plurality of graph elements, and each graph element in each infectious disease outbreak activity comprises information constituting a disease monitoring element in each infectious disease outbreak activity; An infectious disease source tracing and identification network is generated based on the second infectious disease diffusion information map, and the infectious disease source tracing and identification network is used to identify the source infectious disease outbreak activity among the multiple infectious disease outbreak activities, and the source infectious disease outbreak activity is determined based on the commonality measurement between the infectious disease outbreak activities or the commonality measurement between the graph elements in the infectious disease outbreak activities.

6. The method according to claim 5, characterized in that The infectious disease source tracing identification network includes a graph element group commonality measurement component, and the graph element group commonality measurement component is used to obtain the commonality measurement of the graph element group; The generating of the infectious disease source tracing identification network based on the second infectious disease diffusion information graph includes: Grouping the plurality of graph elements in each infectious disease outbreak activity into at least one graph element group; Based on the primitive group commonality measurement component, obtaining a first commonality measurement of each group of primitive groups in the at least one group of primitive groups; Obtaining a primitive group commonality metric error value of each primitive in the plurality of primitives based on the first commonality metric of each group of primitive groups; Integrating the commonality metric error values ​​of the plurality of graph groups in each infectious disease outbreak activity to obtain a first training error value; The infectious disease source tracing and identification network is iterated based on the first training error value to obtain an iterated infectious disease source tracing and identification network.

7. The method according to claim 6, characterized in that The obtaining the commonality metric error value of the graphic element group of each graphic element based on the first commonality metric of each group of graphic element groups comprises: Acquire a similar primitive sample set of a first primitive from the plurality of primitives, and construct a dissimilar primitive sample set, wherein a first commonality measure between a primitive in the similar primitive sample set and the first primitive is greater than a first commonality measure between a primitive in the dissimilar primitive sample set and the first primitive, and the first primitive is a random one of the plurality of primitives in each infectious disease outbreak activity; Based on a first commonality metric between the first primitive and primitives in the similar primitive sample set and a commonality metric between the first primitive and primitives in the dissimilar primitive sample set, a primitive group commonality metric error value corresponding to the first primitive group is determined.

8. The method according to claim 7, characterized in that The determining, based on a first commonality metric between the first primitive and primitives in the similar primitive sample set and a commonality metric between the first primitive and primitives in the dissimilar primitive sample set, a primitive group commonality metric error value corresponding to the first primitive group includes: Obtaining a sensitivity adjustment factor, wherein the sensitivity adjustment factor is related to a commonality measure between the first image element and the image element in the dissimilar image element sample set; Based on the sensitivity adjustment factor, based on the first commonality measure between the first primitive and the primitives in the similar primitive sample set, and based on the commonality measure between the first primitive and the primitives in the dissimilar primitive sample set, determine the primitive group commonality measure error value of the first primitive group.

9. The method according to claim 5, characterized in that The infectious disease source tracing identification network also includes a multidimensional association analysis component, which is used to perform cluster analysis on the graph elements in the infectious disease outbreak activities to obtain at least one category, and obtain the commonality measurement between each graph element in the infectious disease outbreak activities and the at least one category, wherein the commonality measurement represents the deviation degree of the corresponding infectious disease outbreak activities; The generating of the infectious disease source tracing identification network based on the second infectious disease diffusion information map also includes: Acquire a second commonality measure between one or more of the multiple graph elements in each infectious disease outbreak activity and a representative graph element based on the multidimensional association analysis component, wherein the representative graph element includes a central graph element of each infectious disease outbreak activity or a graph element obtained by integrating the multiple graph elements; determining a second training error value based on a second commonality measure between the one or more primitives and the representative primitive; The infectious disease source tracing and identification network is iterated based on the second training error value to obtain an iterated infectious disease source tracing and identification network.

10. The method according to claim 9, characterized in that The multidimensional association analysis component is also used to perform cluster analysis on multiple graph elements in each infectious disease outbreak activity to obtain no less than one category; The determining a second training error value based on a second commonality measure between the one or more primitives and the representative primitive comprises: Changing a first graphic element into a second graphic element, wherein the first graphic element is a graphic element in a first infectious disease outbreak activity, and the second graphic element has the same attributes as the first graphic element, but different categories; Obtaining a third commonality measure between the second primitive and the representative primitive; Determine a training error value based on the second commonality metric and the third commonality metric to obtain the second training error value; The infectious disease source tracing identification network also includes an activity association analysis component, which is used to obtain commonality metrics between infectious disease outbreak activities; The generating of the infectious disease source tracing identification network based on the second infectious disease diffusion information map also includes: Selecting from the multiple infectious disease outbreak activities to obtain a similar activity set and a dissimilar activity set corresponding to each infectious disease outbreak activity; determining a third training error value based on a fourth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the similar activity set and a fifth commonality measure between each infectious disease outbreak activity and the infectious disease outbreak activities in the dissimilar activity set; The infectious disease source tracing and identification network is iterated based on the third training error value to obtain an iterated infectious disease source tracing and identification network.

11. The method according to claim 10, characterized in that The selecting of similar activity sets and dissimilar activity sets corresponding to each infectious disease outbreak activity from the plurality of infectious disease outbreak activities comprises: Acquire the number of cross-graphic elements between each group of infectious disease outbreak activities based on the activity association analysis component; Acquire one or more infectious disease outbreak activities whose number of intersection graph elements with the second infectious disease outbreak activity is greater than a preset threshold, to obtain the similar activity set, wherein the second infectious disease outbreak activity is a random infectious disease outbreak activity among the multiple infectious disease outbreak activities; One or more infectious disease outbreak activities whose number of intersection graph elements with the second infectious disease outbreak activity is not greater than a preset threshold are acquired to obtain the dissimilar activity set.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, wherein: When the processor executes the program, the steps in the method according to any one of claims 1 to 11 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images