WO2021218207A1

WO2021218207A1 - Intra-urban dengue fever spatio-temporal forecasting method and system, and electronic device

Info

Publication number: WO2021218207A1
Application number: PCT/CN2020/139657
Authority: WO
Inventors: 刘康; 尹凌; 奚桂锴
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2020-04-27
Filing date: 2020-12-25
Publication date: 2021-11-04
Also published as: CN111554408B; CN111554408A

Abstract

An intra-urban dengue fever spatio-temporal forecasting method, comprising: acquiring intra-urban dengue fever related data and performing preprocessing (S1); constructing a graph structure reflecting spatial relationships among intra-urban regions (S2); selecting input features for dengue fever spatio-temporal forecasting (S3); and constructing and training a GCN model according to the preprocessed intra-urban dengue fever related data, the constructed graph structure, and the selected input features (S4), so as to perform intra-urban dengue fever spatio-temporal forecasting by using the GCN model. According to the method, the spatial relationships among the intra-urban regions are sufficiently considered, forecasting on a finer spatial scale is achieved, the forecasting performance is improved, and the precise prevention and control level of dengue fever is improved.

Description

Temporal and spatial prediction method, system and electronic equipment of dengue fever in city

Technical field

The invention relates to a method, a system and electronic equipment for predicting the time and space of dengue fever in a city.

Background technique

In recent decades, as a mosquito-borne infectious disease, dengue fever has been endemic in tropical and subtropical regions, especially in Southeast Asian countries and regions such as Singapore and Malaysia. In China, Guangdong Province in the subtropical region, especially Guangzhou City, as an area with developed economy, active trade and frequent movement of people, is affected by dengue fever virus every summer and autumn. There were more than 37,000 cases of dengue fever in Guangzhou in 2014, posing a serious threat to the lives and health of residents.

In the absence of effective vaccines, vector control (such as spraying mosquitoes to eliminate adult mosquitoes, removing breeding grounds for Aedes mosquitoes, etc.) is still the main method of dengue fever prevention and control. In this context, accurate prediction and early warning of the number and location of dengue fever in the future has become the key to prevention and control.

At present, there are many related studies on dengue fever prediction and early warning. Researchers mainly predict the number of dengue fever cases in the study area in the future based on traditional statistical models and machine learning models. However, the current studies are all based on the overall time series prediction of the number of cases in the country, state (province) or city in the future (such as 1 week, 2 weeks, 1 month, etc.), but the fine spatial scale (such as The forecast of township/sub-district administration) is relatively rare. The prediction of dengue fever at the fine spatial scale inside the city is quite challenging. The main reason is that the city is densely populated and the internal population flows frequently, and the disease spreads between the inner areas of the city more quickly. It is easy to ignore the area when each area is modeled separately. The spatial relationship between them cannot achieve a better prediction effect.

Summary of the invention

In view of this, it is necessary to provide a space-time prediction method, system and electronic equipment for dengue fever in a city.

The present invention provides a method for temporal and spatial prediction of dengue fever in a city. The method includes the following steps: a. Collect and preprocess related data on dengue fever in the city. The data on dengue fever in the city includes: dengue fever case data and meteorological data in the studied city , Population distribution data, township vector files; b. Construct a map structure that reflects the spatial relationship of the inner city; c. Select the input features for dengue spatiotemporal prediction; d. Construct a map based on the preprocessed dengue-related data within the city The structure and selected input features are used to construct and train the GCN model.

Wherein, the method further includes step e: evaluating the prediction performance of the GCN model.

The step a specifically includes:

Preprocess the collected data of dengue fever cases: convert the home address of the case to latitude and longitude coordinates; determine the township where each case is located; count the number of cases in each township in each week according to the onset date of each case, constituting W*N The number matrix of cases, W is the number of weeks, and N is the number of towns;

Preprocessing the collected meteorological data: Obtain the daily average temperature and rainfall recorded by all meteorological observation stations in the city, and use the kriging method to interpolate them separately; aggregate the interpolated data to the township level on a weekly basis, Calculate the average temperature and cumulative rainfall of each town in each week to form a W*N average temperature matrix and cumulative rainfall matrix;

The preprocessing of the collected population distribution data includes: aggregating the population distribution data to the township level to obtain the total population of each township.

The step b specifically includes the following steps:

Acquire the neighboring relationship between township and township;

Regarding the township as a point and the neighboring relationship between townships as an edge, a graph structure is constructed.

Said step c specifically includes:

The features commonly used in the literature and closely related to the spread and outbreak of dengue fever are selected as input features.

The GCN model is composed of one input layer, at least two hidden layers and one output layer; after the at least two hidden layers, the rectified linear function ReLU and the hyperbolic tangent function tanh are respectively used as activation functions.

The training of the GCN model in step d includes:

According to the input and output requirements of the GCN model and different prediction windows, organize K sets of data sets, and each set of the data sets is divided into a training set and a validation set;

Use the training set under each prediction window to train the constructed GCN model separately.

The step e specifically includes:

Input the verification set under each prediction window into the corresponding trained GCN model to obtain the prediction result of the t-th week in the future;

Use hit rate to evaluate prediction performance: The hit rate of the prediction result in week t is defined as follows:

Among them, N _m,t means that the number of cases in all towns within the city predicted in week t is ranked from highest to lowest, and the sum of the actual number of cases in the top m% of high-risk streets and towns; N _t means the number of cases in week t The total number of actual cases in the city.

The present invention provides a space-time prediction system for dengue fever in a city. The system includes a preprocessing unit, a graph structure building unit, a selection unit, and a model building unit. The preprocessing unit is used to collect data related to dengue fever in the city and perform preprocessing. The data related to dengue fever in the city includes: dengue fever case data, meteorological data, population distribution data, and township vector files of the studied city; the graph structure construction unit is used to construct a graph structure reflecting the spatial relationship of the inner city; The selection unit is used to select input features for the spatiotemporal prediction of dengue fever; the model construction unit is used to construct and train the GCN model based on the preprocessed intra-city dengue related data, the constructed graph structure, and the selected input features.

Wherein, the system further includes: an evaluation unit for evaluating the prediction performance of the GCN model.

The present invention also provides an electronic device, including:

At least one processor; and

A memory communicatively connected with the at least one processor; wherein,

The memory stores instructions that can be executed by the one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the urban interior described in any one of 1 to 8 above. The following operations of the infectious disease spread simulation method:

Step a: Collect and preprocess the dengue-related data in the city. The dengue-related data in the city includes: dengue case data, meteorological data, population distribution data, and township vector files in the studied city;

Step b: Construct a map structure that reflects the spatial relationship of the city's internal regions;

Step c: Select the input features for the spatiotemporal prediction of dengue fever;

Step d: Constructing and training a GCN model according to the preprocessed data related to dengue fever in the city, the constructed graph structure, and the selected input features, so as to use the GCN model to perform dengue fever spatiotemporal prediction.

Compared with the overall time series prediction of the country, province (state), and city in the prior art, the present invention is oriented to each area within the city, and realizes prediction on a finer spatial scale. When predicting the future number of dengue fever cases in each area within the city , Fully consider the spatial relationship between various regions, help to capture the characteristics of dengue fever in the city, effectively improve the prediction performance, and enhance the level of precision prevention and control of dengue fever.

Description of the drawings

Figure 1 is a flow chart of the method for predicting dengue fever within a city according to the present invention;

FIG. 2 is a schematic diagram of a process of constructing a spatial relationship within a city according to an embodiment of the present invention;

3 is a schematic structural diagram of a graph convolutional neural network model provided by an embodiment of the present invention;

FIG. 4 is a hardware architecture diagram of the dengue fever spatiotemporal prediction system in the city of the present invention;

5 is a schematic diagram of the hardware device structure of a method for predicting dengue fever in a city according to an embodiment of the present invention;

Fig. 6 is a schematic diagram of comparison of dengue fever prediction effects on a township scale in Guangzhou City in Example 1 of the present invention.

Detailed ways

The present invention will be further described in detail below with reference to the drawings and specific embodiments.

This embodiment is explained with the prediction of a township scale. The present invention is also applicable to urban internal spatial units divided in other ways, such as administrative districts, traffic analysis districts, grids, and the like.

Referring to FIG. 1, it is a flowchart of a preferred embodiment of the method for predicting dengue fever within a city according to the present invention.

Step S1, collecting data related to dengue fever in the city and preprocessing it. in particular:

The data related to dengue fever in the city includes: dengue fever case data, meteorological data, population distribution data, and township vector files (shapefile) of the studied city. The meteorological data includes daily average temperature and rainfall collected by meteorological monitoring stations in the city.

Wherein, the dengue fever case data is obtained from the disease prevention and control center of the country/province/city, and the dengue fever case data includes: the onset date and home address of each case; the meteorological data is obtained from the national/provincial/city meteorological The said population distribution data is obtained from the open source global population data project WorldPop website (https://www.worldpop.org/).

The preprocessing of the collected dengue case data includes: first use geocoding to convert the home address of the case into latitude and longitude coordinates, and import all case points into ArcGIS according to their latitude and longitude coordinates to obtain a vector file of point type; then use Spatial in ArcGIS software The Join tool associates cases (point-type vector files) with townships (face-type vector files, that is, township vector files) to determine the township where each case is located; finally, count each week according to the date of onset of each case The number of cases in each township constitutes a W*N matrix of the number of cases, where W is the number of weeks and N is the number of townships.

The preprocessing of the collected meteorological data includes: obtaining the daily average temperature and rainfall recorded by all meteorological observatories in the city, firstly using the kriging method to interpolate them separately; then, the interpolated data is aggregated into weekly At the township level, the average temperature and cumulative rainfall of each township in each week are counted to form the W*N average temperature matrix and cumulative rainfall matrix. In this embodiment, the spatial interpolation and data aggregation are processed in batches using the ArcPy toolkit of the Python language.

The preprocessing of the collected population distribution data includes: this embodiment uses ArcGIS software to download the population distribution data with a resolution of 100 meters from the WorldPop website in 2015 and aggregate it to the township level to obtain the total population of each township.

Step S2, constructing a graph structure reflecting the spatial relationship of the inner regions of the city according to the neighboring relationship between the regions.

Specifically, the step S2 includes:

Step 201: Use the Spatial Join function of ArcGIS software to obtain the neighboring relationship between the township and the township from the township vector file.

Step 202: Regard the township as a point, and the adjacency relationship between the townships as an edge, and construct a graph structure. Please refer to FIG. 2 for a schematic diagram of the construction process of structures A and B in this embodiment.

Step S3, selecting input features for dengue fever prediction. in particular:

This embodiment selects four types of features commonly used in the literature that are closely related to the spread and outbreak of dengue fever, including the number of cases in the current week and the past week, the average temperature, the cumulative rainfall, and the number of populations. As shown in Table 1, there are 13 features in total. Among them, the average moderate and the accumulated rainfall are related to the survival suitability of mosquito vectors; because dengue fever is an infectious disease, the number of future cases is also closely related to the number of past cases and the number of population.

It is worth noting that the input features selected in this embodiment do not compulsorily limit the 13 types used in the present invention, and selection of other reasonable input features and combinations thereof are also within the protection scope of the present invention.

Table 1. Input features used for dengue fever prediction

In step S4, the GCN model is constructed and trained according to the preprocessed data related to dengue fever in the city, the constructed graph structure, and the selected input features.

Specifically, the step S4 includes:

Step 401: Model construction. The graph convolutional neural network model used in this embodiment was proposed by Kipf Thomas N and Max Welling in 2016, and its basic structure is shown in FIG. 3. The model consists of an input layer, two hidden layers (more hidden layers can also be set), and an output layer; after the two hidden layers, the rectified linear function ReLU and the hyperbolic tangent function tanh are used as the activation functions.

The input data of the input layer is: 1) the graph structure A constructed in step S2; 2) the feature matrix X=N*D of N*D, where N is the number of nodes (ie, towns), and D is the number of features. The output layer outputs the number of dengue fever cases at N nodes (ie, townships) in the next T+k week, where k is the prediction window, and k=1, 2,...,K.

Step 402: Model training. According to the input and output requirements of the GCN model, and the different prediction windows, organize K sets of data sets; each set of data sets are divided into training sets and validation sets in a certain proportion: in this embodiment, the top 75% of all weeks in the data set The weekly data is used for training, and the last 25% of the weekly data is used for verification; the training set under each prediction window is used to train the constructed GCN model.

To achieve the construction and training of the GCN model, please refer to the following open source code based on the mainstream deep learning framework:

https://github.com/tkipf/gcn

https://github.com/tkipf/pygcn

https://github.com/tkipf/keras-gcn.

Step S5: Evaluate the prediction performance of the GCN model. in particular:

Input the validation set under each prediction window into the corresponding trained GCN model, and obtain the predicted value of the kth week in the future (that is, the number of cases in each town). Since the main purpose of prediction is to identify high-risk streets and towns among multiple streets and towns within the city, and to deploy prevention and control measures in a targeted manner, this embodiment uses hit rates to evaluate the prediction performance. The hit rate of the forecast results in week t is defined as follows:

Refer to FIG. 4, which is a hardware architecture diagram of the dengue fever spatiotemporal prediction system 10 in a city of the present invention. The system includes: a preprocessing unit 101, a graph structure construction unit 102, a selection unit 103, a model construction unit 104, and an evaluation unit 105.

The preprocessing unit 101 is used to collect and preprocess the data related to dengue fever in the city. in particular:

The preprocessing of the collected dengue case data by the preprocessing unit 101 includes: firstly using a geocoding method to convert the home address of the case into latitude and longitude coordinates, and import all case points into ArcGIS according to their latitude and longitude coordinates to obtain a vector file of point type; Use the SpatialJoin tool in ArcGIS software to associate the case (point-type vector file) with the township (face-type vector file, that is, the township vector file) to determine the township where each case is located; finally, according to the date of onset of each case , Count the number of cases in each town in each week to form a matrix of the number of cases in W*N, where W is the number of weeks and N is the number of towns.

The pre-processing of the collected meteorological data by the pre-processing unit 101 includes: obtaining the daily average temperature and rainfall recorded by all meteorological observatories in the city, firstly using the kriging method to perform spatial interpolation on them; The data is aggregated to the township level by week, and the average temperature and cumulative rainfall of each township in each week are counted to form the average temperature matrix and cumulative rainfall matrix of W*N. In this embodiment, the spatial interpolation and data aggregation are processed in batches using the ArcPy toolkit of the Python language.

The preprocessing unit 101 preprocessing the collected population distribution data includes: using ArcGIS software in this embodiment to download the population distribution data with a resolution of 100 meters in 2015 from the WorldPop website to the township level to obtain the total population of each township.

The graph structure construction unit 102 is used for constructing a graph structure reflecting the spatial relationship of the inner region of the city according to the neighboring relationship between the regions. in particular:

The graph structure construction unit 102 uses the Spatial Join function of ArcGIS software to obtain the neighboring relationship between the township and the township from the township vector file.

Regarding the township as a point and the neighboring relationship between townships as an edge, a graph structure is constructed. Please refer to FIG. 2 for a schematic diagram of the construction process of structures A and B in this embodiment.

The selection unit 103 is used to select input features for dengue fever prediction. in particular:

In this embodiment, the selection unit 103 selects four types of features commonly used in the literature that are closely related to the spread and outbreak of dengue fever, including the number of cases in the current week and the past week, the average temperature, the cumulative rainfall, and the number of population. As shown in Table 1, there are 13 features in total. Among them, the average moderate and the accumulated rainfall are related to the survival suitability of mosquito vectors; because dengue fever is an infectious disease, the number of future cases is also closely related to the number of past cases and the number of population.

Table 1. Input features used for dengue fever prediction

The model construction unit 104 is used for constructing and training the GCN model according to the preprocessed intra-city dengue fever related data, the constructed graph structure, and the selected input features. in particular:

The model construction unit 104 performs model construction. The graph convolutional neural network model used in this embodiment was proposed by Kipf Thomas N and Max Welling in 2016, and its basic structure is shown in FIG. 3. The model consists of an input layer, two hidden layers (more hidden layers can also be set), and an output layer; after the two hidden layers, the rectified linear function ReLU and the hyperbolic tangent function tanh are used as the activation functions.

The input data of the input layer is: 1) the graph structure A constructed in step S2; 2) the feature matrix X=N*D of N*D, where N is the number of nodes (ie, towns), and D is the number of features. The output layer outputs the number of dengue fever cases of N nodes (ie, townships) in the next T+k week, where k is the prediction window, and k=1, 2,...,K.

The model construction unit 104 performs model training. According to the input and output requirements of the GCN model, and the different prediction windows, organize K sets of data sets; each set of data sets are divided into training sets and validation sets in a certain proportion: in this embodiment, the top 75% of all weeks in the data set The weekly data is used for training, and the last 25% of the weekly data is used for verification; the training set under each prediction window is used to train the constructed GCN model.

https://github.com/tkipf/gcn

https://github.com/tkipf/pygcn

https://github.com/tkipf/keras-gcn.

The evaluation unit 105 is used to evaluate the prediction performance of the GCN model. in particular:

The evaluation unit 105 inputs the verification set under each prediction window into the corresponding trained GCN model, and accordingly obtains the prediction value of the kth week in the future (that is, the number of cases in each town). Since the main purpose of prediction is to identify high-risk streets and towns among multiple streets and towns in the city, and to deploy prevention and control measures in a targeted manner, the hit rate is used in this embodiment to evaluate the prediction performance. The hit rate of the forecast results in week t is defined as follows:

FIG. 5 is a schematic diagram of the hardware device structure of the method for simulating the spread of infectious diseases in a city provided by an embodiment of the present application. As shown in Figure 5, the device includes one or more processors and memory. Taking a processor as an example, the device may also include: an input system and an output system.

The processor, the memory, the input system, and the output system may be connected by a bus or other methods. In FIG. 5, the connection by a bus is taken as an example.

As a non-transitory computer-readable electronic device, the memory can be used to store non-transitory software programs, non-transitory computer executable programs, and modules. The processor executes various functional applications and data processing of the electronic device by running non-transitory software programs, instructions, and modules stored in the memory, that is, realizing the processing methods of the foregoing method embodiments.

The memory may include a program storage area and a data storage area, where the program storage area can store an operating system and an application program required by at least one function; the data storage area can store data and the like. In addition, the memory may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely provided with respect to the processor, and these remote memories may be connected to the processing system through a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

The input system can receive input digital or character information, and generate signal input. The output system may include display devices such as a display screen.

The one or more modules are stored in the memory, and when executed by the one or more processors, the following operations of any of the foregoing method embodiments are performed:

Step a: Collect and preprocess the dengue fever-related data in the city. The dengue fever-related data in the city includes: dengue fever case data, meteorological data, population distribution data, and township vector files of the studied city;

The above-mentioned products can execute the methods provided in the embodiments of the present application, and have functional modules and beneficial effects corresponding to the execution methods. For technical details not described in detail in this embodiment, please refer to the method provided in the embodiment of this application.

The embodiments of the present application provide a non-transitory (non-volatile) computer electronic device, the computer electronic device stores computer-executable instructions, and the computer-executable instructions can perform the following operations:

The embodiment of the present application provides a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable electronic device, the computer program includes program instructions, when the program instructions are executed by a computer To make the computer do the following:

Step a: Collect and preprocess the dengue fever-related data in the city. The dengue fever-related data in the city include: dengue fever case data, meteorological data, population distribution data, and township vector files of the studied city;

Experimental results of Example 1 of this application:

Example 1 of this application took 167 villages and towns in Guangdong Province as an example to conduct experiments. The study period is from January 1, 2015 to September 22, 2019, with a total of 247 weeks. Among them, the data from week 5 to week 195 is used for model training, and the data from week 196 to week 247 is used for model verification. The prediction window k is 1, 2, ..., 8.

The comparison methods are LASSO (least absolute shrinkage and selection operator) and SVM (support vector machine) regression models that are commonly used in current dengue fever prediction research and have been proven to be relatively effective. Use the above two models to make individual predictions for each township.

Figure 6 is a comparison diagram of model effects with hit rate as an evaluation index. It can be seen that, compared with the dengue fever prediction method based on the LASSO and SVM regression model, the dengue fever prediction method using GCN provided in the present invention has better overall prediction performance, which fully demonstrates the effectiveness of the present invention.

The present invention introduces the deep learning model Graph Convolutional Network (GCN) for the first time, which fully considers the spatial relationship between the inner regions of the city to capture the spread of diseases in space, and conducts joint prediction of each region, and achieves better results. Accurate prediction effect. In order to provide decision support for relevant prevention and control departments, avoid wasting manpower and material resources, and reduce the loss of life, health and property.

Although the present invention has been described with reference to the current preferred embodiments, those skilled in the art should understand that the above preferred embodiments are only used to illustrate the present invention and are not used to limit the scope of protection of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle scope of the invention shall be included in the protection scope of the present invention.

Claims

A spatiotemporal prediction method of dengue fever in a city is characterized in that the method includes the following steps:

a. Collect and preprocess the dengue-related data within the city. The dengue-related data within the city include: dengue case data, meteorological data, population distribution data, and township vector files in the studied city;

b. Construct a map structure that reflects the spatial relationship of the city's internal regions;

c. Select the input features used for the spatiotemporal prediction of dengue fever;

d. Construct and train the GCN model according to the preprocessed internal dengue fever-related data in the city, the constructed graph structure, and the selected input features, so as to use the GCN model to perform dengue fever spatiotemporal prediction.
The method according to claim 1, wherein the method further comprises step e:

The prediction performance of the GCN model is evaluated.
The method according to claim 1 or 2, wherein the step a specifically includes:

Preprocess the collected data of dengue fever cases: convert the home address of the case into latitude and longitude coordinates; determine the township where each case is located; count the number of cases in each township in each week according to the date of onset of each case, constituting W*N The number matrix of cases, W is the number of weeks, and N is the number of towns;

Preprocess the collected meteorological data: Obtain the daily average temperature and rainfall recorded by all meteorological observation stations in the city, and use the kriging method to interpolate them separately; aggregate the interpolated data to the township level on a weekly basis, Calculate the average temperature and cumulative rainfall of each township in each week to form a W*N average temperature matrix and cumulative rainfall matrix;

The preprocessing of the collected population distribution data includes: aggregating the population distribution data to the township level to obtain the total population of each township.
The method according to claim 3, wherein said step b specifically includes the following steps:

Acquire the neighboring relationship between township and township;

Regarding the township as a point and the neighboring relationship between townships as an edge, a graph structure is constructed.
The method according to claim 4, wherein said step c specifically comprises:

The features commonly used in the literature and closely related to the spread and outbreak of dengue fever are selected as input features.
The method according to claim 5, wherein the GCN model is composed of one input layer, at least two hidden layers, and one output layer; after the at least two hidden layers, the rectified linear function ReLU is used respectively. And the hyperbolic tangent function tanh as the activation function.
The method according to claim 6, wherein the training of the GCN model in step d comprises:

According to the input and output requirements of the GCN model and different prediction windows, organize K sets of data sets, and each set of the data sets is divided into a training set and a validation set;

Use the training set under each prediction window to train the constructed GCN model separately.
8. The method according to claim 7, wherein said step e specifically comprises:

Input the verification set under each prediction window into the corresponding trained GCN model to obtain the prediction result of the t-th week in the future;

Use hit rate (hit rate) to evaluate prediction performance: The hit rate of the prediction result in week t is defined as follows:

Among them, N m,t means that the number of cases in all towns within the city predicted in week t is ranked from highest to lowest, and the sum of the actual number of cases in the top m% of high-risk streets and towns; N t means the number of cases in week t The total number of actual cases in the city.
A spatiotemporal prediction system for dengue fever in a city, which is characterized in that the system includes a preprocessing unit, a graph structure construction unit, a selection unit, and a model construction unit, wherein:

The preprocessing unit is used to collect and preprocess the dengue fever related data in the city. The dengue fever related data in the city includes: dengue fever case data, meteorological data, population distribution data, and township vector files of the studied city;

The graph structure construction unit is used to construct a graph structure that reflects the spatial relationship of the inner regions of the city;

The selection unit is used to select input features for spatiotemporal prediction of dengue fever;

The model construction unit is used to construct and train the GCN model according to the preprocessed internal dengue fever related data, the constructed graph structure, and the selected input features, so as to use the GCN model to perform dengue fever spatiotemporal prediction.
The system of claim 9, wherein the system further comprises:

The evaluation unit is used to evaluate the prediction performance of the GCN model.
An electronic device including:

At least one processor; and

A memory communicatively connected with the at least one processor; wherein,

The memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute any one of the above claims 1 to 8. The following operations of the method of simulating the spread of infectious diseases within cities:

Step a: Collect and preprocess the dengue fever-related data in the city. The dengue fever-related data in the city includes: dengue fever case data, meteorological data, population distribution data, and township vector files of the studied city;

Step b: Construct a map structure that reflects the spatial relationship of the city's internal regions;

Step c: Select the input features for the spatiotemporal prediction of dengue fever;

Step d: Constructing and training a GCN model according to the preprocessed data related to dengue fever in the city, the constructed graph structure, and the selected input features, so as to use the GCN model to perform dengue fever spatiotemporal prediction.