Infectious disease infection prediction method and device based on macrograph-micrograph fusion, and storage medium

Abstract

An infectious disease infection prediction method and device based on macrograph-micrograph fusion, and a storage medium. The method comprises: acquiring macrographs of a plurality of first areas and micrographs of second areas in a certain set time period; respectively inputting the macrographs and the micrographs into a first graph convolutional neural network and a second graph convolutional neural network to respectively obtain a first hidden layer vector and a second hidden layer vector, and fusing the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each first area; performing temporal calculation on the fused hidden layer information to obtain temporal hidden layer information of each first area, and separately inputting the temporal hidden layer information into two prediction networks to obtain two prediction results; and performing fusion calculation on the two prediction results to obtain a final infectious disease prediction result for each first area.

Description

Macro-micro image fusion infectious disease infection prediction method, device and storage medium

[0001] Related applications

[0002] This application claims priority to Chinese patent application number 202311366722.4 filed on October 20, 2023, entitled “Infectious disease infection prediction method, device and storage medium based on macro-micro image fusion”, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the technical field of infectious disease infection prediction, and in particular to a macro-micro graph fusion infectious disease infection prediction method, device and storage medium. Background Art

[0004] Accurately predicting the future number of infections from infectious diseases has always been a crucial issue in infectious disease research. Traditionally, this approach is based on the SIR (Severe Influenza Reduction) mechanistic model. However, this approach requires manual setting of numerous parameters based on empirical experience, resulting in significant uncertainty and unsatisfactory predictions.

[0005] Some neural network and machine learning methods have also been applied to infectious disease infection prediction, but these methods only use the number of infectious disease infections for simple population predictions and are unable to explore the implicit internal relationships between infectious disease data. The prediction accuracy is not high, and the prediction results cannot be explained.

[0006] Summary of the Invention

[0007] According to various embodiments of the present application, a method, device and storage medium for predicting infectious disease infection using a macro-micro graph fusion method are provided, which can combine multiple infectious disease factors, comprehensively mine the relationships between various infectious disease factors through a macro-micro graph structure, and improve the interpretability of prediction results.

[0008] In a first aspect, the present application provides a method for predicting infectious disease infection by integrating macro- and micro-graphs. The method comprises:

[0009] Obtaining macroscopic infection data of a plurality of first regions within a set time period to generate a macroscopic map, and obtaining microscopic infection data of a plurality of second regions within the same set time period to generate a microscopic map corresponding to each of the first regions, wherein the first region includes the plurality of the second regions;

[0010] Inputting the macroscopic image into a first graph convolutional neural network to obtain a first hidden layer vector, inputting the microscopic image into a second graph convolutional neural network to obtain a second hidden layer vector, and performing a fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions;

[0011] Performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and inputting each of the time series hidden layer information into the first prediction network and the second prediction network to obtain corresponding first results and second results;

[0012] The first result and the second result are fused and calculated to obtain infectious disease prediction results for each of the first areas.

[0013] In one embodiment, the macro infection data includes macro personnel data and macro geographic data of each of the first areas, and obtaining the macro infection data of multiple first areas within a set time period to generate a macro map includes:

[0014] Each of the first regions is taken as the first node of the macro graph, the macro personnel data of each of the first regions within a set time period is obtained as the node feature of each of the first nodes, and the first connection edge of each of the first nodes is determined based on the macro geographical data of each of the first regions to generate the macro graph; the macro personnel data includes: the total population, population density, number of infected people and number of recovered people of each of the first regions within a set time period.

[0015] In one embodiment, determining the first edge of each first node based on the macro-geographic data of each first area includes:

[0016] Based on the macro-geographic data of each first region, the first connection probability between each first node is determined, and two first nodes whose first connection probability is greater than a set threshold are connected to obtain the first connection; the macro-geographic data includes: within a certain set time period, the number of infected people flowing from the first region to another first region, the total number of infected people flowing out of the first region, the historical number of infected people in the first region, the historical number of infected people in another first region, and the geographical index between the first region and another first region.

[0017] In one embodiment, the microscopic infection data includes microscopic personnel data and microscopic geographic data of each second area, and obtaining microscopic infection data of multiple second areas within the same set time period to generate microscopic maps corresponding to each first area includes:

[0018] Each second region is used as a second node of the micrograph, micro-personnel data of each second region is obtained as the node feature of each second node, and the second connection edge of each second node is determined based on the micro-geographic data of each second region to generate the micrograph; the micro-personnel data includes the total population, population density and number of hospitals of each second region within a set time period.

[0019] In one embodiment, determining the second edge of each second node based on the micro-geographic data of each second area includes:

[0020] Based on the micro-geographical data of each second region, the second connection probability of each second node is determined, and two second regions whose second connection probability is greater than a set threshold are connected to obtain the second connection; the micro-geographical data includes: within a certain set time period, the number of infected people flowing from the second region to another second region, the total number of infected people flowing out of the second region, and the geographical index between the second region and another second region.

[0021] In one embodiment, performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each first region includes:

[0022] Obtain the temporal hidden layer information of each of the first regions in the previous time period, input each of the fused hidden layer information and the corresponding temporal hidden layer information of the previous time period into a recurrent neural network, and obtain the temporal hidden layer information of each of the first regions in the current time period.

[0023] In one embodiment, the first result includes a first predicted number of infected people and a first predicted number of recovered people, the second result includes a second predicted number of infected people and a second predicted number of recovered people, the second prediction network includes a parameter prediction network and an infection prediction network, and the inputting of each of the time series hidden layer information into the first prediction network and the second prediction network to obtain the corresponding first result and second result includes:

[0024] Inputting the time series hidden layer information into the first prediction network, and outputting a first predicted number of infected people and a first predicted number of recovered people in the first area;

[0025] The time series hidden layer information is input into the parameter prediction network, and the first prediction parameter and the second prediction parameter of the first area are output; the first prediction parameter and the second prediction parameter are input into the infection prediction network, and the second predicted number of infected people and the second predicted number of recovered people in the first area are output.

[0026] In one embodiment, the fusing calculation of the first result and the second result to obtain the infectious disease prediction result of each first region includes:

[0027] The first predicted number of infected people and the second predicted number of infected people are combined to obtain a predicted result of the number of infected people in the first area;

[0028] The first predicted number of recovered people and the second predicted number of recovered people are combined and calculated to obtain a predicted result of the number of recovered people in the first area.

[0029] In a second aspect, the present application also provides a device for predicting infectious diseases by integrating macro and micro images. The device comprises:

[0030] A graph calculation processing module, configured to obtain macroscopic infection data of multiple first regions within a set time period to generate a macroscopic graph, and to obtain microscopic infection data of multiple second regions within the same set time period to generate a microscopic graph corresponding to each of the first regions; the first region includes multiple second regions;

[0031] a spatial fusion calculation module, configured to input the macroscopic image into a first graph convolutional neural network to obtain a first hidden layer vector; input the microscopic image into a second graph convolutional neural network to obtain a second hidden layer vector; and perform a fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions;

[0032] a time fusion calculation module, configured to perform time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and input each of the time series hidden layer information into the first prediction network and the second prediction network to obtain corresponding first results and second results;

[0033] The fusion result module is used to perform fusion calculation on the first result and the second result to obtain the infectious disease prediction results of each of the first areas.

[0034] In a third aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the content of the first aspect when executed by a processor.

[0035] The above-mentioned infectious disease infection prediction method, device and storage medium using macro-micro graph fusion generate a macrograph by obtaining macro infection data of multiple first areas within a set time period, and generate micrographs corresponding to each first area by obtaining micro infection data of multiple second areas within the same set time period; the first area includes multiple second areas; the macrograph is input into the first graph convolutional neural network to obtain a first hidden layer vector; the micrograph is input into the second graph convolutional neural network to obtain a second hidden layer vector; the first hidden layer vector and the second hidden layer vector are fused and calculated to obtain fused hidden layer information of each first area; the fused hidden layer information is time-series calculated to obtain time-series hidden layer information of each first area, and each time-series hidden layer information is respectively input into the first prediction network and the second prediction network to obtain the corresponding first result and second result; the first result and the second result are fused and calculated to obtain the infectious disease prediction result of each first area, realizing infectious disease prediction using macro-micro graph fusion mining, improving the interpretability of prediction results, and the accuracy and efficiency of infectious disease prediction.

[0036] The details of one or more embodiments of the present application are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the present application will become apparent from the description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to better describe and illustrate the embodiments and / or examples of the inventions disclosed herein, reference may be made to one or more of the accompanying drawings. The additional details or examples used to describe the accompanying drawings should not be considered to limit the scope of the disclosed inventions, the presently described embodiments and / or examples, and any of the best modes currently understood for these inventions.

[0038] FIG1 is a diagram illustrating an application environment of a method for predicting infectious diseases using macro-micro image fusion according to an embodiment;

[0039] FIG2 is a schematic flow chart of a method for predicting infectious disease infection by integrating macro and micro graphs in one embodiment;

[0040] FIG3 is a schematic diagram of constructing a macrograph in one embodiment;

[0041] FIG4 is a schematic diagram of constructing a micrograph in one embodiment;

[0042] FIG5 is a structural block diagram of an infectious disease infection prediction device integrating macro-micro graphs in one embodiment;

[0043] FIG6 is a diagram showing the internal structure of an FPGA device in one embodiment. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0045] Unless otherwise defined, the technical or scientific terms used in this application should have the ordinary meaning understood by a person of ordinary skill in the technical field to which this application belongs. The words "one", "a", "the" and the like used in this application do not indicate a limit on quantity and may indicate the singular or plural. The terms "include", "comprise", "have" and any variations thereof used in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units that are inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The word "multiple" used in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0046] The infectious disease infection prediction method using macro-micro image fusion provided in the embodiments of this application can be applied in the application environment shown in Figure 1. Terminal 102 communicates with server 104 via a network. Data storage system 106 can store data that server 104 needs to process. Data storage system 106 can be integrated with server 104 or placed in the cloud or on other network servers.

[0047] On server 104, macro infection data of multiple first areas within a set time period is obtained to generate a macro graph, and micro infection data of multiple second areas within the same set time period is obtained to generate a micro graph corresponding to each first area, where the first area includes multiple second areas. The macro graph is input into a first graph convolutional neural network to obtain a first hidden layer vector, and the micro graph is input into a second graph convolutional neural network to obtain a second hidden layer vector. The first hidden layer vector and the second hidden layer vector are fused and calculated to obtain fused hidden layer information of each first area. A time series calculation is performed on the fused hidden layer information to obtain time series hidden layer information of each first area, and each time series hidden layer information is respectively input into the first prediction network and the second prediction network to obtain the corresponding first result and second result. The first result and the second result are fused and calculated to obtain the infectious disease prediction result of each first area.

[0048] The terminal 102 may be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0049] In one embodiment, as shown in FIG2 , a method for predicting infectious disease infection by integrating macro and micro graphs is provided. The method is described using the server side in FIG1 as an example, including the following steps:

[0050] Step 201: Obtain macro infection data of multiple first areas within a set time period to generate a macro map, and obtain micro infection data of multiple second areas within the same set time period to generate micro maps corresponding to each of the first areas; the first area includes multiple second areas.

[0051] The macro infection data includes macro personnel data and macro geographic data of each of the first areas, and the micro infection data includes micro personnel data and micro geographic data of each of the second areas.

[0052] Specifically, by acquiring macro-level demographic and macro-geographic data on the spread of infectious diseases in multiple first regions, and micro-level demographic and micro-geographic data on multiple second regions within each first region, macro- and micro-maps of each first region are constructed. This integration of multiple infectious disease factors increases the size of the processed dataset for infectious disease prediction, resulting in more accurate predictions.

[0053] In step 202, the macroscopic image is input into a first graph convolutional neural network to obtain a first hidden layer vector; the microscopic image is input into a second graph convolutional neural network to obtain a second hidden layer vector; and the first hidden layer vector and the second hidden layer vector are fused to obtain fused hidden layer information of each of the first regions.

[0054] Specifically, an infectious disease prediction model is constructed, comprising a first graph convolutional neural network and a second graph convolutional neural network. A macrograph is input into the first graph convolutional neural network to obtain a first hidden layer vector for the macrograph, while a micrograph is input into the second graph convolutional neural network to obtain a second hidden layer vector. The first and second hidden layer vectors are then fused to obtain fused hidden layer information for each first region. Using the fused macro-micro graph hidden layer vectors, the model comprehensively mines the associations between various data types and infectious disease prediction results.

[0055] Step 203 , performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and inputting each of the time series hidden layer information into the first prediction network and the second prediction network to obtain corresponding first results and second results.

[0056] Specifically, the infectious disease prediction model also includes a cyclic control network, a first prediction network, and a second prediction network. The second prediction network includes an infectious disease mechanism model. The fused hidden layer information is input into the cyclic control network for time series calculation to obtain time series hidden layer information for each of the first regions. This time series hidden layer information is then input into the first prediction network and the second prediction network, respectively, to obtain prediction results for the two infectious diseases.

[0057] Step 204: perform a fusion calculation on the first result and the second result to obtain an infectious disease prediction result for each of the first areas.

[0058] Specifically, the two infectious disease prediction results are fused and calculated to obtain the corresponding infectious disease prediction results of the first region. The spatiotemporal deep learning and infectious disease mechanism model are integrated to improve the interpretability of the final output infectious disease prediction results.

[0059] In the above-mentioned infectious disease infection prediction method of macro-micro graph fusion, a macro graph is generated by obtaining macro infection data of multiple first areas within a set time period, and micro infection data of multiple second areas within the same set time period are obtained to generate micro graphs corresponding to each first area; the first area includes multiple second areas; the macro graph is input into the first graph convolutional neural network to obtain a first hidden layer vector; the micro graph is input into the second graph convolutional neural network to obtain a second hidden layer vector; the first hidden layer vector and the second hidden layer vector are fused and calculated to obtain the fused hidden layer information of each first area; the fused hidden layer information is time-series calculated to obtain the time-series hidden layer information of each first area, and each time-series hidden layer information is respectively input into the first prediction network and the second prediction network to obtain the corresponding first result and second result; the first result and the second result are fused and calculated to obtain the infectious disease prediction result of each first area, realizing the infectious disease prediction of macro-micro graph fusion mining, improving the interpretability of the prediction result and the accuracy and efficiency of infectious disease prediction.

[0060] In one embodiment, obtaining macro infection data of multiple first areas within a set time period to generate a macro map in step 201 specifically includes the following:

[0061] Each of the first regions is taken as the first node of the macro graph, the macro personnel data of each of the first regions within a set time period is obtained as the node feature of each of the first nodes, and the first connection edge of each of the first nodes is determined based on the macro geographical data of each of the first regions to generate the macro graph.

[0062] The macro-personnel data includes: the total population, population density, number of infections, and number of recoveries in each of the first areas within a set time period.

[0063] Specifically, within the set time period t, as shown in FIG3 , the macro infection data of each of the first regions is abstracted into a macro graph Among them, V s t represents the first node corresponding to each of the first regions, Represents the combination of the first edges of each of the first nodes. The total population, population density, number of infected people, and number of recovered people in each of the first regions within the set time period t are taken as the corresponding first node V s t The node features s.

[0064] The determining of the first edge of each first node based on the macro-geographic data of each first area specifically includes the following:

[0065] Based on the macro-geographic data of each of the first regions, a first edge connection probability between each of the first nodes is determined, and two of the first nodes whose first edge connection probability is greater than a set threshold are connected to obtain the first edge connection.

[0066] The macro-geographic data includes: the number of infected people flowing from the first area to another first area within a set time period, the total number of infected people flowing out of the first area, the historical number of infected people in the first area, the historical number of infected people in another first area, and the geographical index between the first area and another first area.

[0067] Specifically, by setting the number of infected people from the first area i to another first area j within the time period t The total number of infected people leaving the first area i The number of historical infections in the first region i The historical number of infections in another first region j And the geographical index ∈ between the first region i and another first region j, the first edge probability between the first nodes is calculated using the following formula

[0068] Where a and b are manually set weight parameters, simil represents the cosine similarity calculation function, N represents the total number of adjacent regions to the first region i, and k represents the other first regions to which the number of infected people in the first region i flows. It is used to calculate the similarity of the infection situation between two first regions. The calculation formula for the geographic index ∈ between the first region i and another first region j is:

[0069] In this embodiment, the first connection probability of the macro map is calculated by the macro geographic data of the first area, and the two first areas whose first connection probability is greater than the set threshold are connected to realize the construction of the macro map, and the population flow, total population, geographical location, population density, and number of infections are combined to improve the comprehensiveness of the macro map, thereby enhancing the reliability of the infectious disease prediction results.

[0070] In one embodiment, step 201 obtains microscopic infection data of multiple second areas within the same set time period to generate microscopic images corresponding to each of the first areas, specifically including the following:

[0071] Each second area is used as a second node of the micro graph, micro personnel data of each second area is obtained as the node feature of each second node, and the second connection edge of each second node is determined based on the micro geographic data of each second area to generate the micro graph.

[0072] The micro-personnel data includes the total population, population density and number of hospitals in each of the second regions within a set time period.

[0073] Specifically, within the same set time period t, as shown in FIG4 , the microscopic infection data of each second region is abstracted into the microscopic map of the corresponding first region. Among them, V c t represents the second node corresponding to each of the second regions, The total population, population density and number of hospitals in each second region within the set time period t are taken as the corresponding second node V c t Node feature c.

[0074] The determining of the second edge of each second node based on the micro-geographic data of each second region specifically includes the following:

[0075] Based on the micro-geographic data of each second region, a second edge connection probability of each second node is determined, and two second regions whose second edge connection probability is greater than a set threshold are connected to obtain the second edge connection.

[0076] The micro-geographic data includes: the number of infected people flowing from the second area to another second area within a set time period, the total number of infected people flowing out of the second area, and the geographical index between the second area and another second area.

[0077] Specifically, the number of infected people who flow from the second area m to another second area n within the same set time period t The total number of infected people leaving the second area m The geographic index ∈' between the second region m and another second region n is calculated using the following formula to obtain the second edge probability between the second nodes.

[0078] Wherein, a' is a weight parameter set by the user, and the calculation formula of the geographic index ∈' between the second region m and another second region n is:

[0079] In this embodiment, the second connection probability of the micrograph is calculated by the micro-geographic data of the second area, and the two second areas whose second connection probability is greater than the set threshold are connected to realize the construction of the micrograph, and the population flow, total population, geographical location, population density, number of infected people and number of hospitals are combined to improve the comprehensiveness of the micrograph, thereby enhancing the reliability of the infectious disease prediction results.

[0080] In one embodiment, in step 202, the first hidden layer vector and the second hidden layer vector are fused and calculated to obtain fused hidden layer information of each first region, which specifically includes the following:

[0081] The first hidden layer vector and the second hidden layer vector Perform fusion calculation to obtain the fusion hidden layer information of the first region i Among them, w i is the macro weight, w a is the micro-weight, and the other two are the parameters automatically updated during the iterative training of the infectious disease prediction model.

[0082] In one embodiment, in step 203, performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each first region specifically includes the following:

[0083] Obtain the temporal hidden layer information of each of the first regions in the previous time period, input each of the fused hidden layer information and the corresponding temporal hidden layer information of the previous time period into a recurrent neural network, and obtain the temporal hidden layer information of each of the first regions in the current time period.

[0084] Optionally, if the current time period is an initial time period, the fused hidden layer information of the current time period and the randomly generated random number are input into the recurrent neural network to obtain the temporal hidden layer information of each of the first regions in the current time period.

[0085] Specifically, according to the temporal hidden layer information of the first region i in the previous time period The fused hidden layer information of the first region i Use the following formula to get the time series hidden layer information of the first region i in the current time period:

[0086] Among them, GRU represents a recurrent neural network.

[0087] In this embodiment, the fused hidden layer information is input into the recurrent neural network to implement spatiotemporal deep learning of macro-micro graph information and obtain the temporal hidden layer information of the first region.

[0088] In one embodiment, the first result includes a first predicted number of infected people and a first predicted number of recovered people, the second result includes a second predicted number of infected people and a second predicted number of recovered people, and the second prediction network includes a parameter prediction network and an infection prediction network. In step 203, each of the time series hidden layer information is input into the first prediction network and the second prediction network to obtain the corresponding first result and second result, which specifically includes the following:

[0089] The time series hidden layer information is input into the first prediction network to output the first predicted number of infected people and the first predicted number of recovered people in the first area. The time series hidden layer information is input into the parameter prediction network to output the first predicted parameter and the second predicted parameter for the first area. The first and second predicted parameters are input into the infection prediction network to output the second predicted number of infected people and the second predicted number of recovered people in the first area.

[0090] Specifically, the temporal hidden layer information Input the first prediction network MLP_1, and output the first predicted number of infected people I' and the first predicted number of recovered people R' in the first region i:

[0091] The temporal hidden layer information Input parameter prediction network MLP_2, output the first prediction parameter β and the second prediction parameter γ of the first region i: β, γ = sigmoid(MLP_2(h i t )). The first prediction parameter β and the second prediction parameter γ are input into the infection prediction network SIR, and the second predicted number of infected people I" and the second predicted number of recovered people R" in the first area i are output: I", R"=SIR(β,γ).

[0092] In one embodiment, in step 204, the first result and the second result are fused and calculated to obtain infectious disease prediction results for each of the first regions, which specifically include the following:

[0093] The first predicted number of infected people and the second predicted number of infected people are combined and calculated to obtain the predicted result of the number of infected people in the first area; the first predicted number of recovered people and the second predicted number of recovered people are combined and calculated to obtain the predicted result of the number of recovered people in the first area.

[0094] Specifically, the first predicted number of infected people I' and the second predicted number of infected people I" are fused and calculated to obtain the predicted result I of the number of infected people in the first area i: I = p*I'+q*I". The first predicted number of recovered people R' and the second predicted number of recovered people R" are fused and calculated to obtain the predicted result R of the recovered people in the first area i: R = p*R'+q*R". Among them, p and q are weight parameters, and p+q=1.

[0095] The infectious disease prediction model is continuously trained during the execution of steps 201-204, and its loss value loss is calculated using the MSE method: loss = p*(MSE(I, I′)+MSE(R, R′))+q*(MSE(I, I″)+MSE(R, R″)).

[0096] In one example embodiment, a method for predicting infectious disease infection using a fusion of macro and micro graphs is provided, wherein a first region represents a province and a second region represents a county or city within the province. The method specifically includes the following steps:

[0097] S1, within a set time period t, multiple provinces are used as the first node, and the number of infected people flowing from province i to province j is used. Total number of infected people leaving province i Historical number of infections in province i Historical number of infections in province j And the geographic index ∈ between province i and province j, calculate the first edge probability between each first node

[0098] The first edge probability The two first nodes with a value greater than the set threshold are connected to obtain the first connection and generate a macro map of multiple provinces. Among them, V s t Indicates the first node corresponding to each province, Represents the combination of the first connecting edges between provinces. Set the total population of each province Ps in time period t i , population density Ds i , number of infections and the number of people who recovered As the corresponding first node V s t Node features.

[0099] S2, in the same time period t, takes the counties and cities under each province as the second node, and the number of infected people flowing from county and city m to county and city n The total number of infected people flowing out of counties and cities m The geographic index ∈' between county m and county n is used to calculate the second edge probability between the second nodes.

[0100] The second edge probability The two second nodes with values ​​greater than the set threshold are connected to obtain the second connection, and a micro-map of the counties and cities corresponding to each province is generated. Among them, V c t Indicates the second node corresponding to each county and city, Represents the combination of the second edge between the second nodes. The total population of the county Pc i , population density Dc i and the number of hospitals Hc iAs the corresponding second node V c t Node features.

[0101] S3, the macro map Input the first graph convolutional neural network to get the first hidden layer vector The microscopic image Input the second graph convolutional neural network to get the second hidden layer vector

[0102] For the first hidden layer vector and the second hidden layer vector Perform fusion calculation to obtain the fusion hidden layer information of province i

[0103] S4, based on the time series hidden layer information of province i in the previous time period Fusion hidden information of province i Get the time series hidden layer information of province i in the current time period t

[0104] S5, the time series hidden layer information of province i Input the first prediction network MLP_1, and output the first predicted number of infected people I' and the first predicted number of recovered people R' in province i:

[0105] The time series hidden layer information of province i Input parameter prediction network MLP_2, output the first prediction parameter β and the second prediction parameter γ of province i: β, γ = sigmoid(MLP_2(h i t )). The first prediction parameter β and the second prediction parameter γ are input into the infection prediction network SIR, and the second predicted number of infected people I” and the second predicted number of recovered people R” in province i are output: I”, R”=SIR(β,γ).

[0106] S6. Fusion calculation is performed on the first predicted number of infected people I' and the second predicted number of infected people I", to obtain the predicted result I of the number of infected people in the first region i: I = p*I'+q*I". Fusion calculation is performed on the first predicted number of recovered people R' and the second predicted number of recovered people R", to obtain the predicted result R of the number of recovered people in the first region i: R = p*R'+q*R". Wherein, p and q are weight parameters, and p+q=1.

[0107] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0108] Based on the same inventive concept, the embodiments of the present application also provide a macro-micro image fusion infectious disease infection prediction device for implementing the macro-micro image fusion infectious disease infection prediction method involved above. The implementation solution provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations of one or more macro-micro image fusion infectious disease infection prediction device embodiments provided below can be referred to the limitations of the macro-micro image fusion infectious disease infection prediction method above, and will not be repeated here.

[0109] In one embodiment, as shown in FIG5 , a macro-micro graph fusion infectious disease infection prediction device is provided, comprising: a graph calculation processing module 51 , a spatial fusion calculation module 52 , a temporal fusion calculation module 53 , and a fusion result module 54 , wherein:

[0110] The graph calculation processing module 51 is used to obtain the macro infection data of multiple first areas within a set time period to generate a macro graph, and to obtain the micro infection data of multiple second areas within the same set time period to generate a micro graph corresponding to each of the first areas; the first area includes multiple second areas.

[0111] The spatial fusion calculation module 52 is used to input the macroscopic image into the first graph convolutional neural network to obtain a first hidden layer vector; input the microscopic image into the second graph convolutional neural network to obtain a second hidden layer vector; and perform a fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions.

[0112] The time fusion calculation module 53 is used to perform time series calculation on the fused hidden layer information to obtain time series hidden layer information of each first region, and input each time series hidden layer information into the first prediction network and the second prediction network to obtain the corresponding first result and second result.

[0113] The fusion result module 54 is used to perform fusion calculation on the first result and the second result to obtain the infectious disease prediction result of each of the first areas.

[0114] In one embodiment, the macro infection data includes macro personnel data and macro geographic data of each of the first regions, and the graph calculation processing module 51 is further configured to:

[0115] Each of the first regions is taken as the first node of the macro graph, the macro personnel data of each of the first regions within a set time period is obtained as the node feature of each of the first nodes, and the first connection edge of each of the first nodes is determined based on the macro geographical data of each of the first regions to generate the macro graph; the macro personnel data includes: the total population, population density, number of infected people and number of recovered people of each of the first regions within a set time period.

[0116] In one embodiment, the graph calculation processing module 51 is further configured to:

[0117] Based on the macro-geographic data of each first region, the first connection probability between each first node is determined, and two first nodes whose first connection probability is greater than a set threshold are connected to obtain the first connection; the macro-geographic data includes: within a certain set time period, the number of infected people flowing from the first region to another first region, the total number of infected people flowing out of the first region, the historical number of infected people in the first region, the historical number of infected people in another first region, and the geographical index between the first region and another first region.

[0118] In one embodiment, the microscopic infection data includes microscopic personnel data and microscopic geographic data of each second area, and the graph calculation processing module 51 is further configured to:

[0119] Each second region is used as a second node of the micrograph, micro-personnel data of each second region is obtained as the node feature of each second node, and the second connection edge of each second node is determined based on the micro-geographic data of each second region to generate the micrograph; the micro-personnel data includes the total population, population density and number of hospitals of each second region within a set time period.

[0120] In one embodiment, the graph calculation processing module 51 is further configured to:

[0121] Based on the micro-geographical data of each second region, the second connection probability of each second node is determined, and two second regions whose second connection probability is greater than a set threshold are connected to obtain the second connection; the micro-geographical data includes: within a certain set time period, the number of infected people flowing from the second region to another second region, the total number of infected people flowing out of the second region, and the geographical index between the second region and another second region.

[0122] In one embodiment, the time fusion calculation module 53 is further configured to:

[0123] Obtain the temporal hidden layer information of each of the first regions in the previous time period, input each of the fused hidden layer information and the corresponding temporal hidden layer information of the previous time period into a recurrent neural network, and obtain the temporal hidden layer information of each of the first regions in the current time period.

[0124] In one embodiment, the first result includes a first predicted number of infected people and a first predicted number of recovered people, the second result includes a second predicted number of infected people and a second predicted number of recovered people, the second prediction network includes a parameter prediction network and an infection prediction network, and the time fusion calculation module 53 is further configured to:

[0125] The time series hidden layer information is input into the first prediction network, and the first predicted number of infected people and the first predicted number of recovered people in the first area are output; the time series hidden layer information is input into the parameter prediction network, and the first predicted parameter and the second predicted parameter of the first area are output; the first predicted parameter and the second predicted parameter are input into the infection prediction network, and the second predicted number of infected people and the second predicted number of recovered people in the first area are output.

[0126] In one embodiment, the fusion result module 54 is further configured to:

[0127] The first predicted number of infected people and the second predicted number of infected people are combined and calculated to obtain the predicted result of the number of infected people in the first area; the first predicted number of recovered people and the second predicted number of recovered people are combined and calculated to obtain the predicted result of the number of recovered people in the first area.

[0128] Each module in the aforementioned macro-micro image fusion infectious disease prediction device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a computer device's memory in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0129] In one embodiment, an FPGA device is provided, and its internal structure diagram can be shown in Figure 6. The FPGA device includes a control unit, a general graph computing processing module, a spatiotemporal fusion computing module, a fusion result module, an on-chip cache, and an input / output interface (Input / Output, referred to as I / O). The control unit is connected to the general graph computing processing module, the spatiotemporal fusion computing module, the fusion result module and the input / output interface via a system bus. The control unit of the FPGA device is used to provide computing and control capabilities. The on-chip cache of the FPGA device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The input / output interface of the computer device is used to exchange information between the general graph computing processing module and an external device. When the computer program is executed by the control unit, a method for predicting infectious disease infection based on macro-micro graph fusion is implemented.

[0130] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0131] Obtaining macroscopic infection data of multiple first areas within a set time period to generate a macroscopic map, and obtaining microscopic infection data of multiple second areas within the same set time period to generate microscopic maps corresponding to each of the first areas; the first area includes multiple second areas;

[0132] Inputting the macroscopic image into a first graph convolutional neural network to obtain a first hidden layer vector; inputting the microscopic image into a second graph convolutional neural network to obtain a second hidden layer vector; performing a fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions;

[0133] Performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and inputting each of the time series hidden layer information into the first prediction network and the second prediction network to obtain corresponding first results and second results;

[0134] The first result and the second result are fused and calculated to obtain infectious disease prediction results for each of the first areas.

[0135] In one embodiment, the macro infection data includes macro personnel data and macro geographical data of each of the first areas. In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: each of the first areas is used as the first node of the macro graph, the macro personnel data of each of the first areas within a set time period is obtained as the node feature of each of the first nodes, and the first connection edge of each of the first nodes is determined based on the macro geographical data of each of the first areas to generate the macro graph; the macro personnel data includes: the total population, population density, number of infected people and number of recovered people of each of the first areas within a set time period.

[0136] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: based on the macro-geographic data of each first area, the first connection probability between each first node is determined, and two first nodes whose first connection probability is greater than a set threshold are connected to obtain the first connection; the macro-geographic data includes: within a certain set time period, the number of infected people flowing from the first area to another first area, the total number of infected people flowing out of the first area, the historical number of infected people in the first area, the historical number of infected people in another first area, and the geographical index between the first area and another first area.

[0137] In one embodiment, the microscopic infection data includes microscopic personnel data and microscopic geographic data of each second area, and when the computer program is executed by the processor, the computer program further implements the following steps:

[0138] Each second region is used as a second node of the micrograph, micro-personnel data of each second region is obtained as the node feature of each second node, and the second connection edge of each second node is determined based on the micro-geographic data of each second region to generate the micrograph; the micro-personnel data includes the total population, population density and number of hospitals of each second region within a set time period.

[0139] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: based on the micro-geographical data of each second area, the second connection probability of each second node is determined, and two second areas whose second connection probability is greater than a set threshold are connected to obtain the second connection; the micro-geographical data includes: within a certain set time period, the number of infected people flowing from the second area to another second area, the total number of infected people flowing out of the second area, and the geographical index between the second area and another second area.

[0140] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining the temporal hidden layer information of each of the first regions in the previous time period, inputting each of the fused hidden layer information and the corresponding temporal hidden layer information of the previous time period into a recurrent neural network, and obtaining the temporal hidden layer information of each of the first regions in the current time period.

[0141] In one embodiment, the first result includes a first predicted number of infected people and a first predicted number of recovered people, the second result includes a second predicted number of infected people and a second predicted number of recovered people, the second prediction network includes a parameter prediction network and an infection prediction network, and when the computer program is executed by the processor, it further implements the following steps:

[0142] The time series hidden layer information is input into the first prediction network, and the first predicted number of infected people and the first predicted number of recovered people in the first area are output; the time series hidden layer information is input into the parameter prediction network, and the first predicted parameter and the second predicted parameter of the first area are output; the first predicted parameter and the second predicted parameter are input into the infection prediction network, and the second predicted number of infected people and the second predicted number of recovered people in the first area are output.

[0143] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: the first predicted number of infected people and the second predicted number of infected people are merged and calculated to obtain the predicted number of infected people in the first area; the first predicted number of recovered people and the second predicted number of recovered people are merged and calculated to obtain the predicted number of recovered people in the first area.

[0144] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0145] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for predicting infectious diseases based on macro-micro graph fusion, characterized in that: The method comprises: Obtaining macroscopic infection data of a plurality of first areas within a certain set time period to generate a macroscopic graph, and obtaining microscopic infection data of a plurality of second areas within the same set time period to generate a microscopic graph corresponding to each of the first areas, wherein the first area includes a plurality of the second areas; Inputting the macroscopic image into a first graph convolutional neural network to obtain a first hidden layer vector, inputting the microscopic image into a second graph convolutional neural network to obtain a second hidden layer vector, and performing fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions; Performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and inputting each of the time series hidden layer information into the first prediction network and the second prediction network to obtain the corresponding first result and second result; and The first result and the second result are fused and calculated to obtain infectious disease prediction results for each of the first areas.

2. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 1, wherein: The macro infection data includes macro personnel data and macro geographic data of each of the first areas, and obtaining the macro infection data of multiple first areas within a certain set time period to generate a macro map includes: Each of the first regions is taken as the first node of the macro graph, the macro personnel data of each of the first regions within a set time period is obtained as the node feature of each of the first nodes, and the first connection edge of each of the first nodes is determined based on the macro geographical data of each of the first regions to generate the macro graph; the macro personnel data includes: the total population, population density, number of infected people and number of recovered people of each of the first regions within a set time period.

3. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 2, wherein: The determining of the first edge of each of the first nodes based on the macro-geographic data of each of the first regions comprises: Based on the macro-geographic data of each of the first regions, the first connection probability between each of the first nodes is determined, and two of the first nodes whose first connection probability is greater than a set threshold are connected to obtain the first connection; the macro-geographic data include: within a certain set time period, the number of infected people flowing from the first region to another first region, the total number of infected people flowing out of the first region, the historical number of infected people in the first region, the historical number of infected people in another first region, and the geographical index between the first region and another first region.

4. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 1, wherein: The microscopic infection data includes microscopic personnel data and microscopic geographic data of each second area, and the step of acquiring microscopic infection data of multiple second areas within the same set time period to generate microscopic maps corresponding to each first area includes: Each second area is taken as a second node of the micro graph, micro personnel data of each second area is obtained as the node feature of each second node, and the second edge of each second node is determined based on the micro geographical data of each second area to generate the micro graph; the micro personnel data includes the total population, population density and number of hospitals of each second area within a set time period.

5. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 4, wherein: Determining the second edge of each second node based on the micro-geographic data of each second area includes: Based on the micro-geographic data of each second area, the second connection probability of each second node is determined, and two second areas whose second connection probability is greater than a set threshold are connected to obtain the second connection; the micro-geographic data include: within a set time period, the number of infected people flowing from the second area to another second area, the total number of infected people flowing out of the second area, and the geographical index between the second area and another second area.

6. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 1, wherein: The performing time series calculation on the fused hidden layer information to obtain the time series hidden layer information of each first region includes: Obtain the time series hidden layer information of each of the first regions in the previous time period, and combine the fused hidden layer information with the corresponding The time series hidden layer information of the previous time period is input into the recurrent neural network to obtain the time series hidden layer information of each of the first regions in the current time period.

7. The infectious disease infection prediction method based on macro-micro image fusion according to claim 1, wherein: The first result includes a first predicted number of infected people and a first predicted number of recovered people, the second result includes a second predicted number of infected people and a second predicted number of recovered people, the second prediction network includes a parameter prediction network and an infection prediction network, and the inputting of each of the time series hidden layer information into the first prediction network and the second prediction network to obtain the corresponding first result and second result includes: Inputting the time series hidden layer information into the first prediction network, and outputting a first predicted number of infected people and a first predicted number of recovered people in the first area; The time series hidden layer information is input into the parameter prediction network, and the first prediction parameter and the second prediction parameter of the first area are output; the first prediction parameter and the second prediction parameter are input into the infection prediction network, and the second predicted number of infected people and the second predicted number of recovered people in the first area are output.

8. The infectious disease infection prediction method based on macro-micro graph fusion according to claim 7, wherein: The performing fusion calculation on the first result and the second result to obtain the infectious disease prediction result of each of the first regions includes: The first predicted number of infected persons and the second predicted number of infected persons are combined to obtain a predicted result of the number of infected persons in the first area; The first predicted number of recovered people and the second predicted number of recovered people are combined and calculated to obtain a predicted result of the number of recovered people in the first area.

9. A macro-micro image fusion infectious disease infection prediction device, characterized in that: The device comprises: A graph calculation processing module, used to obtain macro infection data of multiple first areas within a certain set time period to generate a macro graph, and to obtain micro infection data of multiple second areas within the same set time period to generate micro graphs corresponding to each of the first areas; the first area includes multiple second areas; A spatial fusion calculation module is used to input the macroscopic image into a first graph convolutional neural network to obtain a first hidden layer vector; input the microscopic image into a second graph convolutional neural network to obtain a second hidden layer vector; perform fusion calculation on the first hidden layer vector and the second hidden layer vector to obtain fused hidden layer information of each of the first regions; A time fusion calculation module, used for performing time series calculation on the fused hidden layer information to obtain time series hidden layer information of each of the first regions, and inputting each of the time series hidden layer information into the first prediction network and the second prediction network to obtain the corresponding first result and second result; The fusion result module is used to perform fusion calculation on the first result and the second result to obtain the infectious disease prediction results of each of the first areas.

10. 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 8 are implemented.

Images