Infectious disease prediction method and device, equipment and storage medium

Abstract

The invention discloses an infectious disease prediction method, device and equipment and a storage medium, and relates to the technical field of infectious disease prediction, and the method considers historical individual infection number, population distribution, age structure, medical facility distribution and climate condition multi-dimensional information; in the feature extraction stage, single time window information extraction is abandoned, multi-time window information extraction is adopted, and then long-term and short-term dimension features are extracted; in the model building stage, multi-scale information calculation is realized by adopting a mode of integrating an LSTM model, and an attention distribution mechanism is introduced to realize effective fusion of model data, so that the accuracy and robustness of model prediction are improved; the idea of a residual network is introduced, rapid convergence in a model training stage is accelerated, and resources are saved; and by utilizing a hyper-parameter optimization strategy, the model building efficiency is improved, and the prediction accuracy of the model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of infectious disease prediction, and in particular to an infectious disease prediction method, device, equipment and storage medium based on an integrated long short-term memory neural network model. Background Art

[0002] The occurrence of infectious diseases not only brings a heavy economic burden to society, but also seriously affects social order and human health. Accurately predicting the incidence trends of infectious diseases can effectively help people formulate prevention and control measures.

[0003] The prediction of the spread of infectious diseases not only needs to consider the historical data of infected individuals, but other factors can also often have a great impact. For infected individuals, different ages have different resistance to infectious viruses. Young people are more resistant to infectious diseases, while the elderly and children are relatively susceptible to infection. The age structure data of the population in each region can be obtained through government departments, and then a regional age structure is formed. Other external environmental factors also affect the spread of infectious diseases, such as urban climate, whose data can be obtained from the local meteorological bureau, including temperature, air humidity, wind speed, etc. Since government agencies have accumulated a large amount of data in the process of informatization construction of social development, machine learning has a great advantage in processing big data and can be applied to the prediction of infectious diseases.

[0004] In common prediction schemes, data feature extraction is often relatively simple, which leads to insufficient information extraction. At the same time, the models built by common prediction schemes are often single models, which may lead to poor accuracy of calculation results. Multi-model fusion has the advantages of high accuracy and good robustness. Summary of the invention

[0005] In view of the above problems, the present invention provides an infectious disease prediction method, apparatus, device and storage medium for overcoming the above problems or at least partially solving the above problems.

[0006] The present invention provides the following scheme:

[0007] A method for predicting infectious diseases, comprising:

[0008] Obtain relevant information of the target area, the relevant information at least including the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions and distribution of medical facilities;

[0009] Numerically representing the data other than the historical individual infection number of the infectious disease; and serializing and expanding the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed;

[0010] Extracting feature data sets at different time scales from the numerical data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure;

[0011] Extracting key information from the feature data set and fusing it to obtain input data by allocating attention;

[0012] The input data is input into a fully connected network so that the fully connected network outputs a prediction result.

[0013] Preferably: the characteristics of the population distribution are numerically represented by calculating the population density of the target area;

[0014] The characteristics of the age structure are numerically represented using three quantitative statistical values, including minors, the elderly, and others;

[0015] The characteristics of the distribution of medical facilities include statistics on the acceptance capacity and distribution density of medical units capable of handling infectious diseases in the target area.

[0016] Preferably: the long short-term memory neural network included in each group of the basic network group is constructed according to the principle that the longer the time window, the more complex the model structure.

[0017] Preferably: after the long short-term memory neural network included in each group of the basic network group is built, each network is trained separately, and the model training stage includes:

[0018] The acquired time window representation data minus the data of the last time node is used as the model input data, and the label value of the model is the data of the last time node; the model will be calculated based on the historical data to obtain the prediction of the last time point of the existing historical data, thereby realizing the training of the basic network.

[0019] Preferably: converting the feature data set into two-dimensional data to obtain two-dimensional data;

[0020] A convolutional neural network is built for the two-dimensional data to perform attention allocation calculation to obtain the input data.

[0021] Preferably: during the training process of the convolutional neural network, the residual network model is used to sum the original target B and the predicted value B_p at the output position of the convolutional neural network as the input data.

[0022] Preferably: a hyperparameter automatic optimization mechanism is used to implement model tuning of the convolutional neural network and the fully connected network.

[0023] An infectious disease prediction device, used to execute the above infectious disease prediction method, the device comprising:

[0024] An information acquisition unit, used to acquire relevant information of the target area, wherein the relevant information includes at least the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions, and distribution of medical facilities;

[0025] A numerical processing unit, used to numerically represent the data other than the historical individual infection number of the infectious disease; and to serialize and expand the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed;

[0026] A feature data extraction unit is used to extract feature data sets at different time scales from the digitized data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure;

[0027] An attention allocation unit, used for extracting key information from the feature data set and fusing it to obtain input data by means of attention allocation;

[0028] The prediction result output unit is used to input the input data into the fully connected network so that the fully connected network outputs the prediction result.

[0029] An infectious disease prediction device, the device comprising a processor and a memory:

[0030] The memory is used to store program code and transmit the program code to the processor;

[0031] The processor is used to execute the above-mentioned infectious disease prediction method according to the instructions in the program code.

[0032] A computer-readable storage medium is used to store program code, and the program code is used to execute the above-mentioned infectious disease prediction method.

[0033] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0034] The embodiments of the present application provide an infectious disease prediction method, apparatus, equipment and storage medium, which takes into account multi-dimensional information such as historical individual infection numbers, population distribution, age structure, distribution of medical facilities, and climate conditions; in the feature extraction stage, a single time window is abandoned to extract information, and multiple time windows are adopted to extract information, thereby extracting long-term and short-term dimensional features; in the model building stage, an integrated LSTM model is adopted to realize multi-scale information calculation, and an attention allocation mechanism is introduced to realize effective fusion of model data, thereby improving the accuracy and robustness of model prediction; the idea of ​​residual network is introduced to accelerate the rapid convergence of the model training stage and save resources; the hyperparameter optimization strategy is used to improve the efficiency of model building and improve the prediction accuracy of the model.

[0035] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0037] Figure 1 is a flow chart of an infectious disease prediction method provided by an embodiment of the present invention;

[0038] Figure 2 is a schematic diagram of a model framework provided by an embodiment of the present invention;

[0039] Figure 3 is a schematic diagram of dividing data into time windows provided by an embodiment of the present invention;

[0040] Figure 4 1 is a schematic diagram of the network model structure of LSTM provided by an embodiment of the present invention;

[0041] Figure 5 is a schematic diagram of an infectious disease prediction device provided by an embodiment of the present invention;

[0042] Figure 6 It is a schematic diagram of an infectious disease prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0044] See also Figure 1 , is a method for predicting infectious diseases provided by an embodiment of the present invention, such as Figure 1 As shown, the method may include:

[0045] S101: Obtain relevant information of the target area, wherein the relevant information includes at least the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions, and distribution of medical facilities;

[0046] S102: numerically characterize the data except the historical individual infection number of the infectious disease; and serialize and expand the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed; in specific implementation, the embodiment of the present application can provide the characteristics of the population distribution to be numerically characterized by calculating the population density of the target area;

[0047] The characteristics of the age structure are numerically represented using three quantitative statistical values, including minors, the elderly, and others;

[0048] The characteristics of the distribution of medical facilities include statistics on the acceptance capacity and distribution density of medical units capable of handling infectious diseases in the target area.

[0049] S103: extracting feature data sets at different time scales from the digitized data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, each of the basic network groups including a plurality of long short-term memory neural networks with the same structure; further, the long short-term memory neural network included in each of the basic network groups is constructed in accordance with the principle that the longer the time window, the more complex the model structure.

[0050] In order to speed up the training of the model, the embodiment of the present application can also provide that after the long short-term memory neural network included in each group of the basic network group is built, each network is trained separately, and the model training stage includes:

[0051] The acquired time window representation data minus the data of the last time node is used as the model input data, and the label value of the model is the data of the last time node; the model will be calculated based on the historical data to obtain the prediction of the last time point of the existing historical data, thereby realizing the training of the basic network.

[0052] S104: extracting key information from the feature data set and fusing it to obtain input data by means of attention allocation; in specific implementation, the embodiment of the present application may also provide for converting the feature data set into two-dimensional data to obtain two-dimensional data;

[0053] A convolutional neural network is built for the two-dimensional data to perform attention allocation calculation to obtain the input data.

[0054] Furthermore, during the training process of the convolutional neural network, the residual network model is used to sum the original target B and the predicted value B_p at the output position of the convolutional neural network as the input data.

[0055] S105: Input the input data into a fully connected network so that the fully connected network outputs a prediction result.

[0056] In specific implementation, the embodiments of the present application can provide a hyperparameter automatic optimization mechanism to achieve model tuning of the convolutional neural network and the fully connected network.

[0057] The infectious disease prediction method provided in the embodiment of the present application is unable to cope with comprehensive sudden situations when only a single data source of historical individual infection number is considered in the traditional infectious disease prediction process. Therefore, the present invention comprehensively considers the influence of multiple factors, such as population distribution, age structure, climate conditions, medical facilities, etc., so as to enhance the accuracy of model prediction.

[0058] The infectious disease prediction method provided in the embodiment of the present application is described in detail below. Figure 2 .

[0059] 1. How to process input data.

[0060] First, the analysis area is divided and relevant information of the area is obtained from relevant official agencies, including historical infection data, population data, medical data, climate data, etc. The data that needs to be processed in the present invention includes specific data of a region: the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions, distribution of medical facilities, etc., and multi-dimensional data processing is required.

[0061] The acquired data is sorted and the data of each dimension is quantified; first, other data except the historical number of individual infections are digitally represented. The population distribution, age structure, distribution of medical facilities, etc. are relatively stable in a region over a period of time. The update frequency of this data is usually not too fast, so it will be a stable value over a period of time. These data features need to select an area first during the extraction process.

[0062] (1) Population distribution characteristics are characterized by calculating the population density of a specified area;

[0063] (2) The age structure characteristics are not represented by the average age, because this representation method will make the areas with a relatively high concentration of elderly people and teenagers have similar values ​​to those of areas with more young and middle-aged people. The present invention uses three quantitative statistical values ​​for representation, namely: minors, elderly, and others;

[0064] (3) Distribution of medical facilities: The acceptance capacity and distribution density of medical units capable of handling infectious diseases in the main statistical area;

[0065] Individual infection data of infectious diseases are obtained through relevant local health departments, mainly infectious diseases that need to be analyzed and predicted.

[0066] The data on climate conditions are mainly obtained through the local meteorological bureau, and the data obtained include historical temperature, humidity, air quality, air pressure, wind speed, etc.

[0067] Arrange the numerical data into the format required by the model. Individual infection data of infectious diseases and climate conditions are both serial data, while population, age, medical and other data are non-serialized. The two types of data need to be adapted, and the non-serialized data should be serialized and expanded according to the sampling rate of the serial data.

[0068] The acquired data needs to be divided into time windows before being input into the model so that the model can process it correctly. The schematic diagram is as follows Figure 3 shown.

[0069] 2. Basic network set.

[0070] In the information extraction process, if a single time scale is used, the selection of too large or too small time scale may lead to redundancy and loss of feature information extraction. The present invention proposes a multi-time scale information extraction method, which can effectively avoid this problem and enhance the model information extraction capability. Multi-scale data feature extraction is mainly achieved by setting different time windows and building multiple processing models.

[0071] Further, the acquired data is set to M different time scales;

[0072] Furthermore, according to M time scales, M basic networks with different complexities are built, and the network structure is set to a simple structure. However, as the time scale increases, the network structure will also be set to a more complex structure.

[0073] Recurrent neural networks (RNNs) and their variants have natural advantages in predicting time series data. Taking resources and costs into consideration, the present invention will build some simple long short-term memory neural network (LSTM) models and perform integrated operations on these models to improve the robustness of the prediction model. LSTM is built according to different feature scales, and a group of basic networks with the same structure are built at one scale.

[0074] Furthermore, multiple basic networks of each scale will be built to ensure stable computing;

[0075] Furthermore, each network is trained according to the set target value to obtain a trained model.

[0076] The basic network of the integrated model of the present invention is a LSTM network model with a simple structure. Different data sets are obtained by selecting different time windows in the data processing stage. The LSTM model will be built according to different time windows to extract richer feature information. The schematic diagram of the LSTM network model structure is shown in Figure 4 shown.

[0077] Figure 4 The diagram shows a two-hidden-layer LSTM model. You can also build a basic network with another structure. The model building follows the principle that the longer the time window, the more complex the model structure. The structure of the basic network that processes the same time window is the same. After the basic network is built, each network needs to be trained separately. In the model training phase, the acquired time window representation data minus the data of the last time node is used as the model input data, and the label value of the model is the data of the last time node. At this time, the model will calculate based on the historical data to obtain the prediction of the last time point of the existing historical data, thereby realizing the training of the basic network.

[0078] 3. Attention allocation.

[0079] For the results predicted by the basic network set, the information is fused through the attention allocation mechanism. The data obtained by the basic network set is processed in a specific way to obtain a format that can be processed by the attention allocation mechanism. Then the attention mechanism model is built. After the data passes through the basic network set, some values ​​will be output. These values ​​represent the feature information extracted at different time scales. These feature information needs to be extracted and fused through the attention allocation method.

[0080] The prediction results are converted into two-dimensional data; a convolutional neural network (CNN) is built for two-dimensional data to perform attention allocation calculations; the idea of ​​residual network is introduced to enable the model to converge quickly during the training phase.

[0081] First, the acquired information is organized into a two-dimensional data list A, and the outputs of the same network structure are placed in the same row. Assuming that there are M groups of the same network structure, and each group has N networks, the dimension of the acquired list A is M×N. At this time, each row of list A is spliced ​​to obtain a one-dimensional list B with a length of M×N. For two-dimensional data, the commonly used processing method is CNN. At this time, CNN is built, the input of CNN is list A, and the label value of CNN is list B. Assuming that after the CNN model is trained, when an input A is given, the output of CNN is B_p. In order to ensure that the overall model can converge as soon as possible, the present invention also introduces the idea of ​​a residual network model. At the output position of CNN, the original target B and the predicted value B_p are summed as the input for calculating the prediction result.

[0082] 4. Calculate the prediction results.

[0083] After the calculation in the attention allocation phase, the prediction results are calculated and the model hyperparameters are optimized in the model training phase to improve the efficiency of model adjustment. A fully connected network is built to output the results of attention allocation through the fully connected network. The model hyperparameters of CNN and fully connected networks are considered, and the hyperparameter automatic optimization mechanism is used to achieve model tuning.

[0084] The calculation of prediction results is achieved by building a fully connected neural network. The input of the network layer is the result obtained in the attention allocation mechanism step, and the label value is the final prediction value. In the attention allocation stage and the calculation of prediction results stage, the built network model can be trained end-to-end during training. The hyperparameters of the network model structure can be automatically optimized through methods including but not limited to: network search, genetic algorithm, Bayesian optimization, etc., to further improve the efficiency of model training.

[0085] In summary, the infectious disease prediction method provided by this application takes into account multi-dimensional information such as historical individual infection numbers, population distribution, age structure, distribution of medical facilities, and climate conditions; in the feature extraction stage, the single time window method of extracting information is abandoned, and multi-time windows are adopted to extract information, thereby extracting long-term and short-term dimensional features; in the model building stage, an integrated LSTM model is adopted to realize multi-scale information calculation, and an attention allocation mechanism is introduced to realize effective fusion of model data, thereby improving the accuracy and robustness of model prediction; the idea of ​​residual network is introduced to accelerate the rapid convergence of the model training stage and save resources; the hyperparameter optimization strategy is used to improve the efficiency of model building and improve the prediction accuracy of the model.

[0086] See also Figure 5 , the present application embodiment can also provide an infectious disease prediction device, such as Figure 5 As shown, for executing the above infectious disease prediction method, the device may include:

[0087] An information acquisition unit 501 is used to acquire relevant information of a target area, wherein the relevant information includes at least the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions, and distribution of medical facilities;

[0088] The numerical processing unit 502 is used to numerically represent the data other than the historical individual infection number of the infectious disease; and to serialize and expand the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed;

[0089] A feature data extraction unit 503 is used to extract feature data sets at different time scales from the digitized data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure;

[0090] An attention allocation unit 504 is used to extract key information from the feature data set and fuse it to obtain input data by means of attention allocation;

[0091] The prediction result output unit 505 is used to input the input data into the fully connected network so that the fully connected network outputs the prediction result.

[0092] The present application embodiment may also provide an infectious disease prediction device, the device comprising a processor and a memory:

[0093] The memory is used to store program code and transmit the program code to the processor;

[0094] The processor is used to execute the steps of the above-mentioned infectious disease prediction method according to the instructions in the program code.

[0095] like Figure 6 As shown, an infectious disease prediction device provided by an embodiment of the present application may include: a processor 10, a memory 11, a communication interface 12 and a communication bus 13. The processor 10, the memory 11 and the communication interface 12 communicate with each other through the communication bus 13.

[0096] In the embodiment of the present application, the processor 10 may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array or other programmable logic devices, etc.

[0097] The processor 10 may call a program stored in the memory 11. Specifically, the processor 10 may execute operations in an embodiment of the infectious disease prediction method.

[0098] The memory 11 is used to store one or more programs, which may include program codes, and the program codes include computer operation instructions. In the embodiment of the present application, the memory 11 at least stores programs for implementing the following functions:

[0099] Obtain relevant information of the target area, the relevant information at least including the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions and distribution of medical facilities;

[0100] Numerically representing the data other than the historical individual infection number of the infectious disease; and serializing and expanding the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed;

[0101] Extracting feature data sets at different time scales from the numerical data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure;

[0102] Extracting key information from the feature data set and fusing it to obtain input data by allocating attention;

[0103] The input data is input into a fully connected network so that the fully connected network outputs a prediction result.

[0104] In one possible implementation, the memory 11 may include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required for at least one function (such as a file creation function, a data reading and writing function), etc.; the data storage area can store data created during use, such as initialization data, etc.

[0105] In addition, the memory 11 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0106] The communication interface 12 may be an interface of a communication module, and is used to connect to other devices or systems.

[0107] Of course, it should be noted that Figure 6 The structure shown does not constitute a limitation on the infectious disease prediction device in the embodiment of the present application. In actual applications, the infectious disease prediction device may include Figure 6 More or fewer components than shown, or combinations of certain components.

[0108] An embodiment of the present application may also provide a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the steps of the above-mentioned infectious disease prediction method.

[0109] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0110] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0111] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, 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 they 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 scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for predicting infectious diseases, characterized in that: include: Obtain relevant information of the target area, the relevant information at least including the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions and distribution of medical facilities; Numerically representing the data other than the historical individual infection number of the infectious disease; and serializing and expanding the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed; Extracting feature data sets at different time scales from the numerical data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure; Extracting key information from the feature data set and fusing it to obtain input data by allocating attention; The input data is input into a fully connected network so that the fully connected network outputs a prediction result.

2. The infectious disease prediction method according to claim 1, characterized in that: The characteristics of the population distribution are numerically represented by calculating the population density of the target area; The characteristics of the age structure are numerically represented using three quantitative statistical values, including minors, the elderly, and others; The characteristics of the distribution of medical facilities include statistics on the acceptance capacity and distribution density of medical units capable of handling infectious diseases in the target area.

3. The infectious disease prediction method according to claim 1, characterized in that: The long short-term memory neural network included in each group of the basic network group is constructed according to the principle that the longer the time window, the more complex the model structure.

4. The infectious disease prediction method according to claim 1, characterized in that: After the long short-term memory neural network included in each group of the basic network group is built, each network is trained separately. The model training stage includes: The acquired time window representation data minus the data of the last time node is used as the model input data, and the label value of the model is the data of the last time node; the model will be calculated based on the historical data to obtain the prediction of the last time point of the existing historical data, thereby realizing the training of the basic network.

5. The infectious disease prediction method according to claim 1, characterized in that: Converting the feature data set into two-dimensional data to obtain two-dimensional data; A convolutional neural network is built for the two-dimensional data to perform attention allocation calculation to obtain the input data.

6. The infectious disease prediction method according to claim 5, characterized in that: During the training process of the convolutional neural network, the residual network model is used to sum the original target B and the predicted value B_p at the output position of the convolutional neural network as the input data.

7. The infectious disease prediction method according to claim 5, characterized in that: The model tuning of the convolutional neural network and the fully connected network is achieved using a hyperparameter automatic optimization mechanism.

8. An infectious disease prediction device, characterized in that: For executing the infectious disease prediction method according to any one of claims 1 to 7, the device comprises: An information acquisition unit, used to acquire relevant information of the target area, wherein the relevant information includes at least the number of historical individual infections of infectious diseases, population distribution, age structure, climate conditions, and distribution of medical facilities; A numerical processing unit, used to numerically represent the data other than the historical individual infection number of the infectious disease; and to serialize and expand the non-serialized data according to the sampling rate of the serial data to obtain the numerical data to be processed; A feature data extraction unit is used to extract feature data sets at different time scales from the digitized data to be processed using a basic network set; the basic network set includes a plurality of basic network groups constructed according to time windows of different time scales, and each of the basic network groups includes a plurality of long short-term memory neural networks with the same structure; An attention allocation unit, used for extracting key information from the feature data set and fusing it to obtain input data by means of attention allocation; The prediction result output unit is used to input the input data into the fully connected network so that the fully connected network outputs the prediction result.

9. An infectious disease prediction device, characterized in that: The device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the infectious disease prediction method according to any one of claims 1-7 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and the program code is used to execute the infectious disease prediction method according to any one of claims 1-7.

Images