A method, medium and device for predicting the transmission path of infectious diseases

Abstract

The present invention relates to the field of public health and safety technology, and specifically to a method, device and medium for predicting the transmission path of an infectious disease, wherein the method comprises the following steps: extracting an initial quadruple of an event from infectious disease outbreak data, mapping the initial quadruple to a numerical value, and obtaining a numerical quadruple; dividing the numerical quadruple by time to obtain a temporal knowledge graph; initializing the embedding representation of entities and relationships in the temporal knowledge graph; constructing a relational graph attention network, learning the structural information in the temporal knowledge graph, and obtaining entity embedding and relationship embedding; calling a gated recurrent unit to extract the temporal information of the temporal knowledge graph; using a decoder to combine the structural information with the temporal information, and generating a prediction result, and realizing the prediction of the transmission path of the infectious disease based on the prediction result. The method of the present invention uses a temporal knowledge graph to predict the transmission path of the infectious disease, and has stronger dynamics and multi-dimensional association analysis capabilities than traditional public health prevention methods.

Description

Technical Field

[0001] The present invention relates to the field of public health and safety technology, and in particular to a method, device and medium for predicting the transmission path of an infectious disease. Background Art

[0002] Traditional infectious disease transmission path prediction methods rely on external sources such as patient trajectory data, base station traffic data, or Internet big data for simple statistical analysis and visualization, but there are problems of lag and singleness, which are not enough to support the provision of refined predictions of the future trajectory of infectious diseases. Most of these methods are based on infection events that have already occurred, and cannot timely identify potential future transmission paths and infection risks, especially when responding to rapidly spreading epidemics. At the same time, these methods lack the ability to reason about complex transmission chains, and it is difficult to capture the risks of multi-level relationships and non-direct contact transmission, resulting in insufficient prediction accuracy and dynamic adaptability. In the case of changing epidemic transmission patterns, this limitation will significantly reduce the efficiency and effectiveness of public health emergency response.

[0003] In summary, there is an urgent need for a method, device and medium for predicting the transmission path of an infectious disease to solve the problems in the existing technology. Summary of the invention

[0004] The present invention aims to provide a method, device and medium for predicting the transmission path of an infectious disease. The specific technical scheme is as follows:

[0005] A method for predicting the transmission path of an infectious disease comprises the following steps:

[0006] S1: Extract the initial quadruple of the event from the infectious disease outbreak data, map the initial quadruple to a numerical value, and obtain a numerical quadruple;

[0007] S2: Divide the numerical quadruple by time to obtain a temporal knowledge graph;

[0008] S3: Initialize the embedding representation of entities and relations in the temporal knowledge graph;

[0009] S4: Build a relational graph attention network to learn the structural information in the temporal knowledge graph and obtain entity embedding and relation embedding;

[0010] S5: Call the gated recurrent unit to extract the timing information of the temporal knowledge graph;

[0011] S6: A decoder is used to combine the structural information with the time series information, and the probability value of the destination to which the disease may spread is obtained, thereby realizing the prediction of the disease transmission path. The calculation formula for calculating the probability value of the destination to which the disease may spread is as follows:

[0012] ;

[0013] ;

[0014] in, represents the predicted probability value of the place where the disease may arrive, represents the distance score, , , and Respectively represent the location, person, access relationship and timestamp, represents the L2 norm, Representing people In time Dynamic embedding, Representing relationships In time Dynamic embedding, Indicate location In time Dynamic embedding, represents the exponential function; Indicates one of the places where the disease may reach. Represents the set of all possible locations where the disease may arrive.

[0015] Optionally, in S1, the initial quadruple includes person, visit relationship, place and time.

[0016] Optionally, in S1, for the initial quadruple, the initial quadruple is mapped to a numerical value using a mapping relationship stored in a dictionary to obtain a numerical quadruple.

[0017] Optionally, in S2, all events under the same timestamp together form a subgraph, and all subgraphs arranged in chronological order constitute a temporal knowledge graph.

[0018] Optionally, in S4, a relational graph attention network is constructed to learn the temporal knowledge graph. The process is as follows:

[0019] Calculate the attention scores of the relationships between neighboring nodes of each subgraph;

[0020] Calculate the normalized attention weights of neighboring nodes;

[0021] Aggregate the features of neighboring nodes and combine them with normalized attention weights to update the features of the subgraph;

[0022] Compute entity embeddings and relation embeddings for each subgraph.

[0023] Optionally, in S5, the timing information includes entity timing information and relationship timing information, and the entity timing information and the relationship timing information are respectively extracted using separate gated recurrent units.

[0024] In addition, the present invention also includes a computer device, including a memory and a processor;

[0025] The memory is used to store a computer program executable on the processor;

[0026] The processor is used to implement the steps of the infectious disease transmission path prediction method as described above when executing the computer program.

[0027] In addition, the present invention also includes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the infectious disease transmission path prediction method as described above are implemented.

[0028] The application of the technical solution of the present invention has the following beneficial effects:

[0029] The method of the present invention provides a method, device and medium for predicting the transmission path of infectious diseases based on a time series knowledge graph. Compared with traditional public health prevention methods, the method of the present invention has stronger dynamics and multi-dimensional correlation analysis capabilities. Existing methods mainly rely on single-dimensional data (such as search index or infectious disease data) to predict infectious disease outbreaks, while the method of the present invention uses knowledge graphs combined with information in multiple dimensions (time, space, socioeconomic factors, medical resources) to form richer background information. In addition, the method of the present invention models the dynamic relationship between events through quadruplets (entities, relationships, timestamps, events), which can capture the complex causal relationship between events, adjust the prediction results in real time, and enhance the accuracy and foresight of the prediction. The method of the present invention also has good scalability, can continuously integrate new data, and improve the timeliness of early warning.

[0030] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 It is a flowchart of the steps of the infectious disease transmission path prediction method in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0034] like Figure 1 As shown, this embodiment provides a method for predicting the transmission path of an infectious disease, comprising the following steps:

[0035] S1: Extract the initial quadruple of the event from the infectious disease outbreak data, map the initial quadruple to a numerical value, and obtain a numerical quadruple.

[0036] It should be noted that this embodiment uses existing tools to extract entities and relations from the infectious disease outbreak data text, and combines the extracted person entities, location entities, access relations, and time into a four-tuple form (person, access relation, location, time) for output. Representative, access relationship by Representative, location by Representative, time by Represents, that is, the output quadruple form is .

[0037] S2: Divide the numerical quadruple by time to obtain a temporal knowledge graph.

[0038] Furthermore, the temporal knowledge graph can be represented as a multi-relation directed graph with timestamp edges, so the temporal knowledge graph can be viewed as a series of knowledge graph snapshots sorted by timestamp, i.e. . Snapshot Each fact in can be represented as a four-tuple The purpose of this embodiment is to: All historical facts that happened before, and events that happened in the future Predict the events at the moment, such as , thereby achieving the effect of public health prevention, It represents the predicted location. Through the location information in the predicted quadruple, the predicted infectious disease transmission path can be obtained, providing assistance for public health safety.

[0039] It should be noted that for the initial quadruple, the mapping relationship stored in the dictionary is used to map the initial quadruple to a numerical value to obtain a numerical quadruple. For the convenience of subsequent model training, the text needs to be vectorized, that is, the above quadruple is converted into a numerical form. Among them, different people, access relationships, and locations are mapped to unique numerical values. In this embodiment, considering the suddenness of the outbreak of infectious diseases, the time is divided into hours, and the continuous time is defined as a strictly increasing timestamp.

[0040] S3: Initialize the embedding representation of entities and relations in the temporal knowledge graph.

[0041] Specifically, the four-tuple , each node (including person, place and access relationship) has a corresponding feature vector at the beginning. This embodiment uses an embedding matrix to convert discrete data (person entity, place entity or access relationship category) into a low-dimensional continuous vector representation to capture its semantic relationship and features, thereby improving the model's expressiveness and learning effect. The specific formula is as follows:

[0042] ;

[0043] ;

[0044] ;

[0045] ;

[0046] in, , , , Representatives , access relationship ,Place and timestamp The initial embedding of , , , is the corresponding embedding matrix, which is used to convert the numerical value of the quadruple into a continuous low-dimensional vector representation.

[0047] S4: Build a relational graph attention network, learn the temporal knowledge graph, and obtain entity embedding and relation embedding.

[0048] Specifically, we construct a relational graph attention network and learn the temporal knowledge graph. The process is as follows:

[0049] Calculate the attention scores of the relationships between neighboring nodes in each subgraph. Specifically, calculate the node and neighboring nodes The relationship between The attention score measures the node For Node Importance:

[0050] ;

[0051] in, Representative Node and neighboring nodes The relationship between The attention score of The importance of the two nodes being connected. Further: is a relation-specific learnable weight matrix used to transform node feature representations; It's a relationship The attention vector is used to calculate the attention score between node pairs; Represents a vector concatenation operation; It is an activation function, usually used to increase the stability of the gradient.

[0052] Calculate the normalized attention weight of the neighboring nodes, the expression is as follows:

[0053] ;

[0054] in, Representation Node and neighboring nodes Through the relationship The normalized attention weight of node For Node influence; Representation Node The set of neighboring nodes; Represents an exponential function, which is used to enhance the discrimination of weights.

[0055] Aggregate the features of neighboring nodes and combine them with attention weights to update the features of the subgraph. The expression is as follows:

[0056] ;

[0057] in, Represents the updated node In the Feature representation of the layer; Refers to the activation function ReLU, which is used to introduce nonlinear transformation; Represents the set of all relations; Represents the sum of all element values ​​in the collection; Indicates the relationship through All neighboring nodes connected The sum of is a specific learnable weight matrix for the relation; For neighboring nodes In the Feature representation of the layer; Represents a learnable parameter that allows the model to retain the feature information of the previous layer and enhance the stability of the feature.

[0058] Entity embedding and relationship embedding of each subgraph. Specifically, the method in this embodiment calculates the relationship The embedded representation is used to deeply mine the local structural information of the relationship. In this embodiment, the entity pairs associated with the subgraph Information aggregation is performed and combined with temporal influence, using a weight-aware graph attention network to aggregate it.

[0059] Furthermore, the computational expression of relation embedding is as follows:

[0060] ;

[0061] in, Represents the updated relationship In the Feature representation of the layer; , They are and In the The embedded representation of the layer, which comes from the output of the previous layer of the RGAT model; For calculation and Relationship The attention weight reflects the influence of the entity on the relationship; Refers to the activation function ReLU, which is used to introduce nonlinear transformation; Indicates through relationship Connected and A collection of .

[0062] Furthermore, the attention weight The calculation expression is as follows:

[0063] ;

[0064] in, For relationship In the The transformation matrix of the layer, which is used to linearly transform the entity embedding; is the learned attention weight vector, which is responsible for calculating the contribution of each group of entities to the relationship; It is an activation function, usually used to increase the stability of the gradient; Represents all relationships Entity Group Perform the summation.

[0065] S5: Call the gated recurrent unit to extract the temporal information of the temporal knowledge graph. Specifically, in order to extract the temporal dependencies of entities and relationships, this embodiment uses gated recurrent units (GRU) designed specifically for entities and relationships to update entity and relationship representations along the time axis.

[0066] Entity temporal information learning:

[0067] The entity-centric GRU is calculated as follows:

[0068] ;

[0069] in, Indicates the last moment The temporal evolution result of the entity (i.e. the output of GRU at the previous moment); express Structural information of entities in the subgraph at a given moment.

[0070] Relational temporal information learning:

[0071] The relation-centric GRU is calculated as follows:

[0072] ;

[0073] in, Indicates the last moment The temporal evolution result of the time relationship (i.e. the output of GRU at the previous moment); express Structural information of relations in the moment subgraph.

[0074] S6: A decoder is used to combine the structural information with the time series information, and the probability value of the destination to which the disease may spread is obtained, thereby realizing the prediction of the disease transmission path. The calculation formula for calculating the probability value of the destination to which the disease may spread is as follows:

[0075] ;

[0076] ;

[0077] in, represents the predicted probability value of the place where the disease may arrive, represents the distance score, , , and Respectively represent the location, person, access relationship and timestamp, represents the L2 norm, Representing people In time Dynamic embedding, Representing relationships In time Dynamic embedding, Indicate location In time Dynamic embedding, represents the exponential function; Indicates one of the places where the disease may reach. Represents the set of locations where all diseases may arrive. It should be noted that the sum of the probabilities of all possible target entities calculated by the above formula is 1.

[0078] It should be noted that the prediction result is the possibility of an event occurring. This embodiment further maps the time series information through the aforementioned dictionary to obtain an initial four-tuple representation. Specifically, the entity value and the place entity are mapped to find the corresponding place entity, and the place name is the final event representation.

[0079] The process of using the method of this embodiment to predict the transmission path of infectious diseases is as follows:

[0080] First, we use the existing natural language processing library to extract the person, place, time entity and visit relationship from the public case movement trajectory text, and output it as a four-tuple of (person, visit relationship, place, time). For example, we convert the event "Liu, a confirmed infectious disease patient, took bus No. XXX to University A" into ("Liu", "arrival", "University A", "July 26 17:14").

[0081] For the convenience of subsequent model training, the entities in the above four-tuple need to be mapped to unique values. Use a dictionary to store the above mapping relationship, and a unique entity corresponds to a unique number. For example, ("Liu", "arrival", "A University", "July 26 17:14") is mapped to [76, 3, 215, 74].

[0082] All events in the text are converted into numerical quadruplets according to the above steps, and then each element of the numerical quadruplet is mapped to a low-dimensional vector representation in the continuous space. This step is called embedding. The numerical quadruplet is represented by embedding .

[0083] All events at the same timestamp together form a multi-relational directed graph , a series of graphs along the time sequence constitute the temporal knowledge graph , a graph at a certain time t is called a subgraph .

[0084] For each subgraph, first calculate the neighboring nodes and The attention score of the relationship between nodes is used to measure the importance of neighboring nodes to the node. Secondly, the normalized attention weights of the surrounding nodes are calculated to represent the influence of neighboring nodes on the node. Finally, the node features are updated by aggregating the features of neighboring nodes and combining the attention weights. For each relationship in the subgraph, the influence weight of each group of entities involved in the relationship on the relationship is first calculated. Then, the embedding representation of the relationship is updated by weighted summing of the embeddings of the entities.

[0085] The temporal knowledge graph is composed of a series of knowledge graph subgraphs. The method of this embodiment uses a gated recurrent unit GRU to extract temporal information. Specifically, the GRU output result of the previous timestamp and the embedding of entities and relationships obtained in the subgraph are respectively input into two gated recurrent units to extract temporal information.

[0086] Finally, the previously obtained entity embedding and relationship embedding are converted into prediction results, that is, the target entity, through the decoder. The target entity value can be mapped to the corresponding location entity through the previously constructed dictionary. For example ("Liu", "arrival", "A community", "July 26 21:00"), July 26 21:00 is the future time, that is, through Liu's historical trajectory and Internet big data, the method of this embodiment predicts that Liu may arrive at A community at 21:00 on July 26, and the health agency can deploy control in A community in advance to curb the spread of the epidemic in time.

[0087] The method of this embodiment uses a knowledge graph to combine information of multiple dimensions (time, space, socioeconomic factors, and medical resources). The method of this embodiment obtains the movement trajectory of existing public cases, combines socioeconomic factors and medical resources, and forms richer background information.

[0088] This example demonstrates the effectiveness of verification on an infectious disease time series knowledge graph dataset. First, the experimental setup is explained in detail, including the baseline algorithm and dataset, then the results are obtained, and finally an ablation study is performed.

[0089] Baseline algorithm: When evaluating the performance of the algorithm proposed in this patent, two types of models are selected as comparative baseline algorithms, namely, static knowledge graph reasoning models and temporal knowledge graph reasoning models. Static knowledge graph reasoning models include ConvE models, and temporal knowledge graph reasoning models include CyGNet models, TeMP models, and RE-NET models. The ConvE model is the simplest multi-layer convolutional structure for link prediction; the CyGNet model combines the replication mode and the generation mode; the TeMP model adopts a structure based on encoders and decoders; the RE-NET model is an autoregressive structure that predicts future interactions.

[0090] Dataset: The infectious disease event dataset was used during the experiment, which consists of event quadruplets. The relevant statistical data of the dataset are shown in Table 1:

[0091] Table 1 Dataset statistics

[0092]

[0093] Among them, the number of entities = the number of confirmed cases + the number of visited locations, the time interval represents the time granularity of adjacent timestamps, and the time interval of the data set used in this patent is divided into 24 hours.

[0094] Experimental evaluation indicators: When performing performance evaluation in the method of this embodiment, the performance evaluation indicators include Mean Reciprocal Ranking (MRR) and Hits@K (K=1, 3, 10). The following are descriptions of the above performance indicators:

[0095] MRR is the average of the inverse of the individual rankings, MRR∈[0,1]. The larger the MRR value, the better the performance of the algorithm.

[0096] Hits@K indicates the proportion of triples ranked in the top K by score. The larger the Hits@K value, the better the performance of the algorithm.

[0097] The experimental results are shown in Table 2.

[0098] Table 2 Experimental results

[0099]

[0100] It can be seen from the above table that compared with the ConvE model, CyGNet model, TeMP model and RE-NET model, the method in this embodiment has better performance in MRR and Hits@K, and has relative improvements in these two indicators.

[0101] In addition, this embodiment also discloses a computer device, including a memory and a processor;

[0102] The memory is used to store a computer program executable on the processor;

[0103] The processor is used to implement the steps of the above-mentioned infectious disease transmission path prediction method when executing the computer program.

[0104] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the computer device.

[0105] The computer device may be a computing device such as a mobile phone, a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may include, but is not limited to, a processor and a memory. For example, the computer device may also include an input / output device, a network access device, a bus, etc.

[0106] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the computer device, and uses various interfaces and lines to connect various parts of the entire computer device.

[0107] The memory can be used to store the computer program and / or module, and the processor implements the computer program by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0108] Wherein, if the module / unit integrated in the computer device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0109] In addition, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned infectious disease transmission path prediction method are implemented.

[0110] The method of the present invention provides a method, device and medium for predicting the transmission path of infectious diseases based on a time series knowledge graph. Compared with traditional public health prevention methods, the method of the present invention has stronger dynamics and multi-dimensional correlation analysis capabilities. Existing methods mainly rely on single-dimensional data (such as search index or infectious disease data) to predict infectious disease outbreaks, while the method of the present invention uses knowledge graphs combined with information in multiple dimensions (time, space, socioeconomic factors, medical resources) to form richer background information. In addition, the method of the present invention models the dynamic relationship between events through quadruplets (entities, relationships, timestamps, events), which can capture the complex causal relationship between events, adjust the prediction results in real time, and enhance the accuracy and foresight of the prediction. The method of the present invention also has good scalability, can continuously integrate new data, and improve the timeliness of early warning.

[0111] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the transmission path of an infectious disease, characterized in that: The following steps are involved: S1: Extract the initial quadruple of the event from the infectious disease outbreak data, map the initial quadruple to a numerical value, and obtain a numerical quadruple; S2: Divide the numerical quadruple by time to obtain a temporal knowledge graph; S3: Initialize the embedding representation of entities and relations in the temporal knowledge graph; S4: Build a relational graph attention network to learn the structural information in the temporal knowledge graph based on entity embedding and relation embedding; S5: Call the gated recurrent unit to extract the timing information of the temporal knowledge graph; S6: A decoder is used to combine the structural information with the time series information, and the probability value of the destination to which the disease may spread is obtained, thereby realizing the prediction of the disease transmission path. The calculation formula for calculating the probability value of the destination to which the disease may spread is as follows: f(s,r,o)=-||H s,t +R r,t -H o,t || 2 ; Among them, P(o|s,r,t) represents the predicted probability value of the place where the disease may arrive, f(s,r,o) represents the distance score, o, s, r and t represent the place, person, visit relationship and timestamp respectively, ||·|| represents the L2 norm, H s,t represents the dynamic embedding of character s at time t, R r,t represents the dynamic embedding of relation r at time t, H o,t represents the dynamic embedding of location o at time t, exp represents the exponential function; o′ represents one of the locations that the disease may reach, and O represents the set of all locations that the disease may reach.

2. The infectious disease transmission path prediction method according to claim 1, characterized in that: In S1, the initial quadruple includes person, visit relation, place and time.

3. The infectious disease transmission path prediction method according to claim 2, characterized in that: In S1, for the initial quadruple, the mapping relationship stored in the dictionary is used to map the initial quadruple to a numerical value, thereby obtaining a numerical quadruple.

4. The infectious disease transmission path prediction method according to claim 3, characterized in that: In S2, all events with the same timestamp together form a subgraph, and all subgraphs arranged in chronological order constitute a temporal knowledge graph.

5. The infectious disease transmission path prediction method according to claim 4, characterized in that: In S4, a relational graph attention network is constructed to learn the temporal knowledge graph. The process is as follows: Calculate the attention scores of the relationships between neighboring nodes of each subgraph; Calculate the normalized attention weights of neighboring nodes; Aggregate the features of neighboring nodes and combine them with normalized attention weights to update the features of the subgraph; Compute entity embeddings and relation embeddings for each subgraph.

6. The infectious disease transmission path prediction method according to claim 5, characterized in that: In S5, the timing information includes entity timing information and relationship timing information, and the entity timing information and the relationship timing information are extracted using separate gated recurrent units respectively.

7. A computer device, characterized in that: including memory and processor; The memory is used to store a computer program executable on the processor; The processor is used to implement the steps of the infectious disease transmission path prediction method as described in any one of claims 1 to 6 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the infectious disease transmission path prediction method as described in any one of claims 1 to 6.

Images