A complex network disease propagation tracing method based on a graph neural network

Abstract

The application relates to a complex network disease propagation tracing method based on a graph neural network. The method comprises the following steps: firstly, an individual infection sample on a contact network is obtained by using a propagation dynamics model; then, node topological features on the contact network are obtained based on a centrality method, and the individual infection sample is added; then, data preprocessing is performed on the individual infection sample, and the individual infection sample is divided into a training sample, a verification sample and a test sample; then, the training sample and the verification sample are put into a neural network model based on a message passing mechanism for training, wherein the neural network model based on the message passing mechanism comprises a feature encoding module, an MPNN network, a propagation controller module and a priori knowledge module; finally, the test sample is input into the trained neural network model based on the message passing mechanism, and a prediction result of a propagation source of the test sample is obtained. The propagation source on the complex network can be efficiently and accurately inferred.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of infectious disease monitoring, in particular to a complex network disease propagation tracing method based on a graph neural network. BACKGROUND

[0002] In response to emerging infectious diseases and public health emergencies, accurately tracing the source of disease propagation is crucial to controlling the spread of the epidemic and developing precise prevention and control measures. Currently, transmission source tracking and tracing focuses on locating the source of sudden transmission information in the real world, aiming to implement effective control and risk assessment through this process. Given that transmission phenomena caused by transmission sources, such as major epidemics threatening human health, are common in real life, rapid and accurate identification and location of transmission sources are of great significance in reducing social losses.

[0003] In the prior art, traditional disease propagation models are usually used for tracing, mainly relying on centrality algorithms of infectious sources to identify potential transmission sources by calculating the node centrality of the infection subgraph in the complex network. This method assumes that the source node is usually located in the key position of the infection subgraph during the propagation process and has a high centrality value, so it can be used as a potential source candidate. However, when dealing with transmission tracing problems in complex network environments, it often faces problems such as low accuracy and low computational efficiency.

[0004] Therefore, in related technologies, there is an urgent need for a way to improve the accuracy and efficiency of complex network disease propagation tracing. SUMMARY

[0005] Therefore, it is necessary to provide a complex network disease propagation tracing method based on a graph neural network to improve the accuracy and efficiency of complex network disease propagation tracing.

[0006] In a first aspect, the present application provides a complex network disease propagation tracing method based on a graph neural network. The method comprises:

[0007] obtaining individual infection samples on the contact network using a propagation dynamics model;

[0008] obtaining node topology features on the contact network based on a centrality method and adding the individual infection samples;

[0009] performing data preprocessing on the individual infection samples and dividing them into training samples, validation samples and test samples;

[0010] putting the training sample and the validation sample into a neural network model based on a message passing mechanism for training, wherein the neural network model based on the message passing mechanism comprises a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module;

[0011] putting the test sample into the trained neural network model based on the message passing mechanism to obtain a prediction result of a propagation source of the test sample.

[0012] Optionally, in an embodiment of the present application, the propagation dynamics model adopts an individual SIR-based propagation dynamics model.

[0013] Optionally, in an embodiment of the present application, the contact network comprises a random graph network, a scale-free network, a citation network, a dolphin social network, a karate club social network, and a Les Miserables character co-occurrence network.

[0014] Optionally, in an embodiment of the present application, the node topology feature comprises betweenness centrality, closeness centrality, and degree centrality, and a calculation formula of the node topology feature is:

[0015]

[0016] Z i =[B i ∥C i ∥D i ]

[0017] wherein B i denotes the end centrality, σ st denotes the number of shortest paths from a node s to a node t, σ st (i) denotes the number of shortest paths from s to t passing through a node i, C i denotes the closeness centrality, d(i,j) denotes the shortest path distance between the node i and the node j, D i denotes the degree centrality, k i denotes the degree of the node i, i.e., the number of nodes connected to the node, A ij is an element in the adjacency matrix of the contact network, and denotes whether the node i and the node j are connected, if connected, A ij = 1, otherwise 0, Z i denotes the topology feature of the node, and ∥ denotes a splicing operation.

[0018] Optionally, in an embodiment of the present application, the data preprocessing comprises:

[0019] eliminating unreasonable individual infection samples, and performing one-hot encoding on the node information and the edge information of the contact network.

[0020] Optionally, in one embodiment of the present application, placing the training samples and the verification samples into a neural network model based on a message passing mechanism for training includes:

[0021] The cross entropy loss function is used as the optimization objective, and the formula of the optimization objective function is:

[0022]

[0023] in, Represents the loss to be optimized, y i Represents the true label of node i. If node i is the source node, then y i =1, otherwise y i =0, represents the prediction score of node i as the source, represents the prediction score of node j as the source.

[0024] Optionally, in one embodiment of the present application, Top-1 accuracy, Top-5 accuracy and Hop-1 accuracy are used as evaluation indicators to evaluate the inference ability of the neural network model based on the message passing mechanism.

[0025] In a second aspect, the present application also provides a complex network disease transmission tracing device based on a graph neural network. The device includes:

[0026] A data acquisition module is used to obtain individual infection samples on the contact network using a transmission dynamics model;

[0027] A node topology feature acquisition module, used to acquire node topology features on the contact network based on a centrality method and add them to the individual infection sample;

[0028] A data preprocessing module is used to perform data preprocessing on the individual infection samples and divide them into training samples, verification samples and test samples;

[0029] A model training module, configured to place the training samples and the verification samples into a neural network model based on a message passing mechanism for training, wherein the neural network model based on a message passing mechanism includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module;

[0030] The propagation tracing module is used to input the test sample into the trained neural network model based on the message passing mechanism to obtain the prediction result of the propagation source of the test sample.

[0031] In a third aspect, the present application also provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor executes the steps of the method in each of the above embodiments.

[0032] In a fourth aspect, the present application also provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the method in each of the above embodiments.

[0033] The above-mentioned complex network disease propagation tracing method based on a graph neural network first obtains individual infection samples on a contact network by using a propagation dynamics model. Then, the node topology features on the contact network are obtained based on a centrality method and added to the individual infection samples. Then, the individual infection samples are preprocessed and divided into training samples, verification samples and test samples. Then, the training samples and the verification samples are put into a neural network model based on a message passing mechanism for training, wherein the neural network model based on the message passing mechanism comprises a feature encoding module, an MPNN network, a propagation controller module and a priori knowledge module. Finally, the test samples are input into the trained neural network model based on the message passing mechanism to obtain the prediction result of the propagation source of the test samples. That is, by using the propagation dynamics model to simulate the propagation process, recording the infection snapshot of individuals on the contact network, analyzing the network node characteristics based on the centrality-based topology features, considering the features of the nodes and edges on the contact network, and using the neural network model based on the message passing mechanism for deep learning, the propagation source on the complex network can be efficiently and accurately inferred, thereby providing technical support for disease prevention. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 An application environment diagram of the complex network disease propagation tracing method based on a graph neural network in one embodiment;

[0035] Figure 2 A flowchart of the complex network disease propagation tracing method based on a graph neural network in one embodiment;

[0036] Figure 3 A structure diagram of the neural network model based on a message passing mechanism in one embodiment;

[0037] Figure 4 A structure diagram of the feature encoding module in one embodiment;

[0038] Figure 5 A structure block diagram of the complex network disease propagation tracing device based on a graph neural network in one embodiment;

[0039] Figure 6 Fig. 1 is a schematic diagram of an internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0040] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0041] The complex network disease propagation tracing method based on a graph neural network provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 . The terminal 102 communicates with the server 104 through a network. The data storage system can store data required to be processed by the server 104. The data storage system can be integrated on the server 104, or placed on a cloud or other network server. The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0042] In one embodiment, as shown in Figure 2 , a complex network disease propagation tracing method based on a graph neural network is provided. Taking the server in Figure 1 as an example, the method includes the following steps:

[0043] S201: Obtain individual infection samples on the contact network by using a propagation dynamics model.

[0044] In the embodiments of the present application, first, the propagation dynamics model is used to simulate propagation on the contact network, record the infection of individuals on the contact network at different times, and form individual infection samples. This process is called snapshot. Optionally, the propagation dynamics model includes SIR model, SEIR model, SIS model and other variants.

[0045] In one embodiment of the present application, the propagation dynamics model uses an individual SIR-based propagation dynamics model.

[0046] In an embodiment of the present application, the propagation dynamics model adopts an individual SIR-based propagation dynamics model. In the individual SIR-based propagation dynamics model, S represents a susceptible individual, I represents an infected individual, R represents a recovered individual, contact between individuals is represented as an edge on a contact network, a susceptible individual is infected with a probability of β when contacting an infected individual and thus turns into an infected individual, and an infected individual recovers with a probability of γ at each time step and thus turns into a recovered individual.

[0047] The process of simulating an infectious disease based on the SIR-based propagation dynamics model is as follows: randomly initializing a node on the network as a source node, and giving the number of propagation time steps t of this simulation, the basic reproduction number (R0) of the infectious disease, and the recovery rate (γ), and calculating the infectious disease parameter infection rate (β) through the following formula:

[0048]

[0049] where β represents the infection rate, R0 represents the basic reproduction number, γ represents the recovery rate, and λ represents the spectral radius of the contact network, i.e., the real part of the largest eigenvalue of the adjacency matrix of the contact network.

[0050] Simulating an infectious disease based on the SIR-based propagation dynamics model at each time step and recording the infection state of the network at each time, until one of the following conditions is met: the number of propagation time steps reaches t or there is no individual in the network in the infected state.

[0051] In an embodiment of the present application, the contact network includes a random graph network, a scale-free network, a citation network, a dolphin social network, a karate club social network, and a Les Mis characters co-occurrence network.

[0052] In an embodiment of the present application, the contact network includes an ER network, a BA network, a Cora network, a Dolphin network, a Karate network, and a Lesmis network. Details are shown in the following Table 1.

[0053]

[0054] S203: Obtain the node topology features on the contact network based on the centrality method, and add the individual infection sample.

[0055] In an embodiment of the present application, the node topology features include betweenness centrality, closeness centrality, and degree centrality, and the calculation formula of the node topology features is as follows:

[0056]

[0057] Z i =[B i ∥Ci ∥D i ]

[0058] wherein B i represents the closeness centrality, σ st represents the number of shortest paths from node s to node t, σ st (i) represents the number of shortest paths from s to t passing through node i, C u represents the closeness centrality, d(i,j) represents the shortest path distance between node i and node j, D i represents the degree centrality, k i represents the degree of node i, i.e., the number of nodes connected to the node, A ij is an element in the adjacency matrix of the contact network, indicating whether node i and node j are connected, if connected, A ij = 1, otherwise 0, Z i represents the topological feature of the node, and ∥ represents the concatenation operation.

[0059] In an embodiment of the present application, the betweenness centrality B i measures the intermediary role of a node in the propagation path, and explains the key role of the node in the information flow and infection spread; the closeness centrality C i reflects the average distance of the node to other nodes, indicating the efficiency of the node in the propagation process; the degree centrality D i represents the number of direct connections of the node, reflecting the influence of the node in the infection network. These three centrality indicators provide a quantitative basis for understanding the propagation mechanism and node importance in the infection state graph, and Z i as the topological feature of the node, obtains the topological feature matrix of the node

[0060] S205: data preprocessing is performed on the individual infection sample, and the individual infection sample is divided into a training sample, a validation sample and a test sample.

[0061] In an embodiment of the present application, the individual infection sample contains an infection state feature matrix, a topological feature matrix and an edge feature matrix of nodes on the contact network, and the infection state includes a susceptible state S, an infected state I and a recovered state R. After data preprocessing is performed on the individual infection sample, the individual infection sample is divided into a training sample, a validation sample and a test sample according to a ratio of 8:1:1.

[0062] In an embodiment of the present application, the data preprocessing includes:

[0063] unreasonable individual infection samples are removed, and the node information and edge information of the contact network are one-hot encoded.

[0064] In an embodiment of the present application, the records of over-infection and less infection status in the individual infection sample are removed to improve the quality of data and the effectiveness of analysis. Meanwhile, the node information of the contact network in the individual infection sample is encoded, and the three different infection states (susceptible state, infected state, and recovered state) of the nodes on the network are one-hot encoded to obtain the node infection state feature matrix where N is the number of nodes on the network; the edge information of the contact network in the individual infection sample is encoded, and the difference between the source node and the target node in the infection state is one-hot encoded, considering that each node has multiple possible states, and the edge information also corresponds to multiple potential state combinations, to obtain the edge state feature where E is the number of edges on the network.

[0065] S207: Put the training sample and the verification sample into the neural network model based on the message passing mechanism for training, wherein the neural network model based on the message passing mechanism includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module.

[0066] In an embodiment of the present application, as shown in Figure 3 , it is a structural diagram of the neural network model based on the message passing mechanism, which includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module, wherein, as shown in Figure 4 , it is a structural diagram of the feature encoding module, which is used to map the input feature information to the hidden layer, and the mapping formula is as follows:

[0067] H N =ReLU(W N ·x N )

[0068] H E =ReLU(W E ·x E )

[0069] H Z =ReLU(W Z ·Z N )

[0070] where H N represents the hidden layer representation of the node infection feature, W N represents the weight matrix of the node infection feature mapping, x N represents the infection state feature of the node, H E represents the hidden layer representation of the edge state feature, W E represents the weight matrix of the edge state feature mapping, x E represents the edge state feature, and H Zthe hidden layer representation of the node topology features, W Z the weight matrix of the node topology feature mapping, Z N representing the node topology features.

[0071] The MPNN network is a kind of graph neural network (GNN) that transmits information between nodes of a graph through a message passing mechanism to capture the relationship between nodes and their neighbors, and the MPNN network comprises two main modules, an attention module and a message passing module, wherein the formula of the attention module is as follows:

[0072]

[0073] Z i = H z [i,:]

[0074]

[0075] wherein l represents the number of message aggregation layers of the MPNN, and the maximum value is a preset hyperparameter L, and represents the node feature representation at the lth layer of the MPNN network, and the initial value is the corresponding feature of node i in the hidden layer representation H N of the infection feature, represents the edge feature representation at the lth layer of the MPNN network, and the initial value is the corresponding feature of edge ij in the hidden layer representation H E of the edge state feature, index ij represents the index value of edge ij, z i represents the corresponding feature of node i in the hidden layer representation H Z of the node topology feature, represents the attention coefficient of edge ij at the lth layer of the MPNN network, and a represents an attention weight vector, represents the attention calculation weight matrix of the lth layer of the MPNN network.

[0076] The message passing module comprises aggregation and update operations, and the formula is as follows:

[0077]

[0078] wherein, represents the information aggregated by edges and nodes at the lth layer of the MPNN network, W l represents the message passing weight matrix of the lth layer of the MPNN network, represents the weight matrix of the node itself information update, represents the first-order neighbor of node i on the contact network, It represents the prediction score of node i as the source, with a value between 0 and 1. FC(·) is a fully connected network.

[0079] The propagation controller module is used to adaptively control the number of message aggregation layers l of different nodes. The formula is as follows:

[0080]

[0081] in, represents the binary aggregate control signal of the control node i at the lth layer of the MPNN network, Indicates that node i participates in message passing at layer l, Indicates that node i does not participate in message passing at layer l, F controller Represents a two-layer multilayer perceptron (MLP) with output values ​​between 0 and 1, θ represents the weight parameter in the controller module, and u represents the hyperparameter that controls the signal.

[0082] The prior knowledge module introduces knowledge in the field of infectious diseases during network forward propagation, thereby improving the generalization ability and reasoning efficiency of the deep learning network. Considering that nodes in a susceptible state cannot be the source of transmission, the prior knowledge is introduced to adjust the output results. The formula is as follows:

[0083]

[0084] Among them, ε represents a positive number that is infinitely close to 0 but not equal to 0, which is used to correct the prediction results based on prior knowledge. i Represents the infection status of node i.

[0085] In one embodiment of the present application, placing the training samples and the verification samples into a neural network model based on a message passing mechanism for training includes:

[0086] The cross entropy loss function is used as the optimization objective, and the formula of the optimization objective function is:

[0087]

[0088] in, Represents the loss to be optimized, y i Represents the true label of node i. If node i is the source node, then y S =1, otherwise y i =0, represents the prediction score of node i as the source, represents the prediction score of node j as the source.

[0089] In one embodiment of the present application, a neural network model is trained using training samples and employs a cross-entropy loss function as an optimization objective to minimize the error between the model's predictions and the true labels. This training aims to ensure that the model can effectively distinguish between source nodes and non-source nodes. By minimizing this loss function, the neural network's ability to identify and infer source nodes is enhanced.

[0090] S209: Input the test sample into the trained neural network model based on the message passing mechanism to obtain the prediction result of the propagation source of the test sample.

[0091] In an embodiment of the present application, finally, the test sample is input into the trained neural network model based on the message passing mechanism, and the model outputs the prediction result of the propagation source of the test sample.

[0092] In one embodiment of the present application, Top-1 accuracy, Top-5 accuracy and Hop-1 accuracy are used as evaluation indicators to evaluate the inference ability of the neural network model based on the message passing mechanism.

[0093] In one embodiment of the present application, three types of indicators are selected to evaluate the model inference ability: Top-1 accuracy, Top-5 accuracy and Hop-1 accuracy. Top-1 accuracy is used to evaluate whether the model can predict the score in each inference. The highest node is correctly identified as the true source node (i.e., y i =1); Top-5 accuracy measures whether the model can predict the score in each inference The top five nodes include the actual source node. Even if the model's inferred preferred node is incorrect, as long as the source node is among the top five predicted scores, the Top-5 prediction is still considered accurate. The Hop-1 accuracy is used to evaluate whether the model can accurately infer the source node in each inference through the neighborhood information within one hop (including the node itself).

[0094] In the above-mentioned complex network disease propagation tracing method based on a graph neural network, first, an individual infection sample on a contact network is obtained by using a propagation dynamics model; then, node topology features on the contact network are obtained based on a centrality method and added to the individual infection sample; then, the individual infection sample is preprocessed and divided into training samples, verification samples, and test samples; then, the training samples and the verification samples are put into a neural network model based on a message passing mechanism for training, wherein the neural network model based on the message passing mechanism includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module; finally, the test samples are input into the trained neural network model based on the message passing mechanism to obtain the prediction result of the propagation source of the test samples. That is, by simulating the propagation process by using the propagation dynamics model, recording the infection snapshot of the individual on the contact network, analyzing the network node characteristics based on the centrality-based topology features, considering the features of the nodes and edges on the contact network, and performing deep learning by using the neural network model based on the message passing mechanism, the propagation source on the complex network can be efficiently and accurately inferred, thereby providing technical support for disease prevention.

[0095] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least part of other steps or steps or stages in other steps.

[0096] Based on the same inventive concept, the embodiments of the present application also provide a complex network disease propagation tracing device based on a graph neural network for implementing the above-mentioned complex network disease propagation tracing method based on a graph neural network. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more complex network disease propagation tracing device embodiments based on a graph neural network provided below can refer to the limitations of the complex network disease propagation tracing method based on a graph neural network in the above text, which will not be repeated here.

[0097] In one embodiment, as Figure 5As shown, a complex network disease propagation tracing device 500 based on a graph neural network is provided, comprising a data acquisition module 501, a node topology feature acquisition module 503, a data preprocessing module 505, a model training module 507, and a propagation tracing module 509, wherein:

[0098] The data acquisition module 501 is configured to acquire individual infection samples on a contact network by using a propagation dynamics model.

[0099] The node topology feature acquisition module 503 is configured to acquire node topology features on the contact network based on a centrality method and add the node topology features to the individual infection samples.

[0100] The data preprocessing module 505 is configured to perform data preprocessing on the individual infection samples and divide the individual infection samples into training samples, validation samples, and test samples.

[0101] The model training module 507 is configured to put the training samples and the validation samples into a neural network model based on a message passing mechanism for training, wherein the neural network model based on the message passing mechanism comprises a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module.

[0102] The propagation tracing module 509 is configured to input the test samples into the trained neural network model based on the message passing mechanism to obtain a prediction result of a propagation source of the test samples.

[0103] In an embodiment of the present application, the propagation dynamics model adopts an individual SIR-based propagation dynamics model.

[0104] In an embodiment of the present application, the contact network comprises a random graph network, a scale-free network, a citation network, a dolphin social network, a karate club social network, and a Les Miserables character co-occurrence network.

[0105] In an embodiment of the present application, the node topology features comprise betweenness centrality, closeness centrality, and degree centrality, and a calculation formula of the node topology features is:

[0106]

[0107] Z S =[B i ∥C i ∥D i ]

[0108] wherein B i represents the closeness centrality, and sigma st represents the number of shortest paths from a node s to a node t, and sigma st(i) represents the number of shortest paths from s to t passing through node i, C i represents closeness centrality, d(i,j) represents the shortest path distance between node i and node j, D i represents degree centrality, k i represents the degree of node i, i.e., the number of nodes connected to the node, A ij is an element in the adjacency matrix of the contact network, indicating whether node i and node j are connected, if connected, A ij = 1, otherwise 0, Z i represents the topological characteristics of the node, and || represents the concatenation operation.

[0109] In an embodiment of the present application, the data preprocessing module is further configured to:

[0110] reject unreasonable individual infection samples, and one-hot encode the node information and edge information of the contact network.

[0111] In an embodiment of the present application, the model training module is further configured to:

[0112] adopt a cross-entropy loss function as an optimization objective, and the formula of the optimization objective function is:

[0113]

[0114] wherein, represents a loss to be optimized, y i represents the true label of node i, if node i is a source node, then y i = 1, otherwise y i = 0, represents the predicted score of node i being a source, represents the predicted score of node j being a source.

[0115] In an embodiment of the present application, Top-1 accuracy, Top-5 accuracy and Hop-1 accuracy are adopted as evaluation indexes to evaluate the inference ability of the neural network model based on the message passing mechanism.

[0116] The above various modules in the complex network disease propagation tracing device based on the graph neural network can be all or part realized by software, hardware and combinations thereof. The above various modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to call and execute the operations corresponding to the above various modules by the processor.

[0117] In an embodiment, a computer device is provided, which can be a terminal, and the internal structure diagram thereof can be as shown in Figure 6As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a complex network disease propagation tracing method based on a graph neural network. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad provided on the shell of the computer device. It can also be an external keyboard, touchpad or mouse, etc.

[0118] Those skilled in the art can understand that, Figure 6 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0119] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0120] In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0121] In one embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0122] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0123] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0124] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0125] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A complex network disease transmission tracing method based on graph neural network, characterized by: The method comprises: A transmission dynamics model is used to obtain individual infection samples on the contact network; Obtaining node topology characteristics on the contact network based on a centrality method and adding them to the individual infection sample; Data preprocessing is performed on the individual infection samples, and they are divided into training samples, verification samples, and test samples. The individual infection samples contain the infection state feature matrix, topology feature matrix, and edge feature matrix of the nodes on the contact network. The infection states include susceptible state S, infected state I, and recovered state R. The training samples and verification samples are placed in a neural network model based on a message passing mechanism for training, wherein the neural network model based on a message passing mechanism includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module. The feature encoding module is used to map the input feature information to the hidden layer. The mapping formula is as follows: in, Hidden layer representation representing node infection characteristics, represents the weight matrix of the node infection feature map, Indicates the infection status characteristics of the node, The hidden layer representation of the edge state features, represents the weight matrix of the edge state feature map, Indicates the edge status characteristics, Hidden layer representation representing node topological features, represents the weight matrix of the node topology feature map, Represents node topology characteristics; The propagation controller module is used to adaptively control the number of message aggregation layers of different nodes , the formula is as follows: in, Represents the MPNN network Layer Control Node The binary aggregation control signal, Representation node In the Layers participate in message passing, Representation node In the Layers do not participate in message passing, Represents a two-layer multilayer perceptron (MLP), with output values ​​between 0 and 1. Represents the weight parameter in the controller module, represents the hyperparameters that control this signal; The prior knowledge module introduces knowledge in the field of infectious diseases during network forward propagation, thereby improving the generalization ability and reasoning efficiency of the deep learning network. Considering that nodes in a susceptible state cannot be the source of transmission, the prior knowledge is introduced to adjust the output results. The formula is as follows: in, Represents a positive number that is infinitely close to 0 but not equal to 0, and is used to correct the prediction results based on prior knowledge. Representation node infection status; The step of placing the training samples and the verification samples into a neural network model based on a message passing mechanism for training comprises: The cross entropy loss function is used as the optimization objective, and the formula for the optimization objective function is: in, represents the loss to be optimized, Representation node The true label of the node is the source node, then ,otherwise , Representation node is the prediction score of the source, Representation node is the predicted score of the source; The test sample is input into a trained neural network model based on a message passing mechanism to obtain a prediction result of the propagation source of the test sample.

2. A complex network disease transmission tracing method based on graph neural network according to claim 1, characterized in that: The propagation dynamics model adopts a propagation dynamics model based on individual SIR.

3. The method for tracing the source of disease transmission in complex networks based on graph neural networks according to claim 1 is characterized in that: The contact networks include random graph networks, scale-free networks, citation networks, dolphin social networks, karate club social networks, and Les Miserables character co-occurrence networks.

4. The method for tracing the spread of a disease in a complex network based on a graph neural network according to claim 1, characterized in that: The node topology features include betweenness centrality, closeness centrality, and degree centrality. The calculation formula of the node topology features is: in, represents betweenness centrality, represents the number of shortest paths from node 𝑠 to node 𝑡, Indicates passing through node From arrive The number of shortest paths, represents close centrality, Representation node To Node The shortest path distance between represents degree centrality, Representation node The degree of the node is the number of nodes connected to it. is an element in the contact network adjacency matrix, representing the node and nodes Are they connected? If so, , otherwise 0, Represents the topological characteristics of the node, Represents a splicing operation.

5. The method for tracing the spread of complex network diseases based on graph neural network according to claim 1 is characterized in that: The data preprocessing includes: Unreasonable individual infection samples are eliminated, and the node information and edge information of the contact network are one-hot encoded.

6. The method for tracing the source of disease transmission in complex networks based on graph neural networks according to claim 1 is characterized in that: The Top-1 accuracy, Top-5 accuracy and Hop-1 accuracy are used as evaluation indicators to evaluate the inference ability of the neural network model based on the message passing mechanism.

7. A complex network disease transmission tracing device based on graph neural network, characterized by: The device comprises: A data acquisition module is used to obtain individual infection samples on the contact network using a transmission dynamics model; A node topology feature acquisition module, used to acquire node topology features on the contact network based on a centrality method and add them to the individual infection sample; A data preprocessing module is used to preprocess the individual infection samples and divide them into training samples, verification samples and test samples. The individual infection samples include the infection state feature matrix, topology feature matrix and edge feature matrix of the nodes on the contact network. The infection state includes susceptible state S, infected state I and recovered state R; The model training module is used to put the training samples and verification samples into a neural network model based on a message passing mechanism for training. The neural network model based on a message passing mechanism includes a feature encoding module, an MPNN network, a propagation controller module, and a priori knowledge module. The feature encoding module is used to map the input feature information to the hidden layer. The mapping formula is as follows: in, Hidden layer representation representing node infection characteristics, represents the weight matrix of the node infection feature map, Indicates the infection status characteristics of the node, The hidden layer representation of the edge state features, represents the weight matrix of the edge state feature map, Indicates the edge status characteristics, Hidden layer representation representing node topological features, represents the weight matrix of the node topology feature map, Represents node topology characteristics; The propagation controller module is used to adaptively control the number of message aggregation layers of different nodes , the formula is as follows: in, Represents the MPNN network Layer Control Node The binary aggregation control signal, Representation node In the Layers participate in message passing, Representation node In the Layers do not participate in message passing, Represents a two-layer multilayer perceptron (MLP), with output values ​​between 0 and 1. Represents the weight parameter in the controller module, represents the hyperparameters that control this signal; The prior knowledge module introduces knowledge in the field of infectious diseases during network forward propagation, thereby improving the generalization ability and reasoning efficiency of the deep learning network. Considering that nodes in a susceptible state cannot be the source of transmission, the prior knowledge is introduced to adjust the output results. The formula is as follows: in, Represents a positive number that is infinitely close to 0 but not equal to 0, and is used to correct the prediction results based on prior knowledge. Representation node infection status; The step of placing the training samples and the verification samples into a neural network model based on a message passing mechanism for training comprises: The cross entropy loss function is used as the optimization objective, and the formula for the optimization objective function is: in, represents the loss to be optimized, Representation node The true label of the node is the source node, then ,otherwise , Representation node is the prediction score of the source, Representation node is the predicted score of the source; The propagation tracing module is used to input the test sample into the trained neural network model based on the message passing mechanism to obtain the prediction result of the propagation source of the test sample.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. 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 6 are implemented.

Images