Infectious disease trend prediction method, device, equipment, storage medium and program product

Abstract

The application provides an infectious disease trend prediction method and device based on artificial intelligence, an electronic device, a computer readable storage medium and a computer program product, relates to artificial intelligence technology, and the method comprises the following steps: acquiring environmental characteristics of a prediction period, and acquiring search data for infectious diseases in the prediction period, wherein the search data comprises keywords for searching for infectious diseases; performing feature extraction processing based on the search data to obtain search characteristics of the infectious diseases in the prediction period; performing multi-source fusion processing on the environmental characteristics of the prediction period and the search characteristics to obtain multi-source fusion characteristics of the infectious diseases; and performing trend prediction processing on the multi-source fusion characteristics of the infectious diseases to obtain the infectious disease trend in the prediction period. Through the application, the infectious disease trend can be automatically and accurately predicted.

Description

Technical Field

[0001] This application relates to artificial intelligence technology, and more particularly to an artificial intelligence-based method, device, electronic device, computer-readable storage medium, and computer program product for predicting infectious disease trends. Background Technology

[0002] Artificial Intelligence (AI) is a comprehensive technology within computer science that studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities. AI technology is a multidisciplinary field encompassing a wide range of areas, including natural language processing and machine learning / deep learning. With technological advancements, AI will be applied in more fields and play an increasingly important role.

[0003] Infectious disease trend prediction is one of the important applications of artificial intelligence. It can predict the trend of infectious diseases in real life (such as the percentage of influenza-like cases and the number of influenza-like cases), and thus provide early warning of future outbreaks of infectious diseases based on infectious disease trends.

[0004] There is a lack of effective methods for predicting infectious disease trends in related technologies, with most relying on historical case counts to provide early warnings of future outbreaks. However, this approach is overly dependent on historical case counts, leading to inaccurate predictions of infectious disease trends. Summary of the Invention

[0005] This application provides an artificial intelligence-based method, apparatus, electronic device, computer-readable storage medium, and computer program product for predicting infectious disease trends, which can automatically and accurately predict infectious disease trends.

[0006] The technical solution of this application embodiment is implemented as follows:

[0007] This application provides an artificial intelligence-based method for predicting infectious disease trends, including:

[0008] The environmental characteristics of the prediction period are obtained, and search data for infectious diseases during the prediction period are obtained, wherein the search data includes keywords for searching for the infectious disease;

[0009] Based on the search data, feature extraction processing is performed to obtain the search features for the infectious disease during the prediction period;

[0010] The environmental characteristics of the prediction period and the search characteristics are subjected to multi-source fusion processing to obtain multi-source fusion characteristics for the infectious disease.

[0011] Trend prediction processing is performed on the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend for the prediction period.

[0012] This application provides an artificial intelligence-based infectious disease trend prediction device, comprising:

[0013] The acquisition module is used to acquire environmental characteristics during the prediction period and acquire search data for infectious diseases during the prediction period, wherein the search data includes keywords for searching for the infectious diseases;

[0014] The feature extraction module is used to perform feature extraction processing based on the search data to obtain the search features for the infectious disease during the prediction period;

[0015] The fusion module is used to perform multi-source fusion processing on the environmental features of the prediction period and the search features to obtain multi-source fusion features for the infectious disease.

[0016] The prediction module is used to perform trend prediction processing on the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend for the prediction period.

[0017] In the above technical solution, the feature extraction module is further used to determine the keyword lexicon for the infectious disease during the prediction period;

[0018] The keywords in the keyword database are matched with the keywords in the search data to obtain the search keywords for the prediction period;

[0019] Based on the search data, the search keywords are statistically processed to obtain the search metrics for the search keywords;

[0020] The search keywords are subjected to feature transformation processing to obtain the semantic features of the search keywords;

[0021] The set of semantic features of the search keywords and the search indicators of the search keywords are used as the search features for the infectious disease during the prediction period.

[0022] In the above technical solution, the feature extraction module is further used to determine the first correlation between multiple candidate keywords of the prediction period and the target words of the infectious disease;

[0023] Based on the first relevance, the multiple candidate keywords are filtered to obtain keywords for updating the thesaurus;

[0024] The virus feature terminology is updated based on the keywords used to update the terminology, resulting in the updated virus feature terminology.

[0025] Determine the second correlation between keywords in the updated virus feature word library and target words of the infectious disease;

[0026] Based on the second correlation, the keywords in the updated virus feature word library are filtered to obtain multiple target keywords;

[0027] Based on the multiple target keywords, a keyword lexicon for the infectious disease during the prediction period is constructed.

[0028] In the above technical solution, before constructing the keyword database for the predicted period of the infectious disease, the device further includes:

[0029] The update module is used to obtain the predicted infectious disease trend of samples during the prediction period and the infectious disease trend annotation of samples during the prediction period.

[0030] When the difference between the predicted infectious disease trend and the labeled infectious disease trend is greater than a difference threshold, it is determined that the operation of constructing a keyword thesaurus for the infectious disease during the prediction period will be performed.

[0031] In the above technical solution, the feature extraction module is further used to perform vector mapping processing on the search keywords to obtain the embedding features of the search keywords;

[0032] The semantic features of the search keywords are obtained by performing semantic transformation on the embedding features of the search keywords.

[0033] In the above technical solution, the semantic conversion is achieved through a fully connected neural network, which includes multiple cascaded fully connected layers;

[0034] The feature extraction module is also used to perform semantic mapping processing on the embedded features of the search keywords through the first fully connected layer in the plurality of cascaded fully connected layers;

[0035] The mapping result of the first fully connected layer is output to the subsequent cascaded fully connected layers. The semantic mapping process and mapping result output are continued through the subsequent cascaded fully connected layers until the last fully connected layer is output.

[0036] The mapping result output by the last fully connected layer is used as the semantic feature of the search keyword.

[0037] In the above technical solution, before performing multi-source fusion processing on the environmental features of the prediction period and the search features, the fusion module is also used to standardize the environmental features of the prediction period to obtain standardized environmental features.

[0038] performing normalization processing on the search index in the search features to obtain normalized search features;

[0039] performing multi-source fusion processing on the normalized environmental features and the normalized search features to obtain multi-source fusion features for the infectious disease.

[0040] In the above technical solution, the fusion module is further configured to perform concatenation processing on the environmental features and the search features of the prediction period to obtain the multi-source fusion features for the infectious disease; or,

[0041] performing weighted summation processing on the environmental features and the search features of the prediction period to obtain the multi-source fusion features for the infectious disease;

[0042] wherein the environmental features include at least one of the following: meteorological features and air features.

[0043] In the above technical solution, the prediction module is further configured to perform trend prediction processing on the multi-source fusion features for the infectious disease based on a plurality of prediction models, to obtain an infectious disease trend corresponding to each of the prediction models;

[0044] performing fusion processing on the infectious disease trend corresponding to each of the prediction models based on the weight of each of the prediction models, to obtain the infectious disease trend of the prediction period.

[0045] In the above technical solution, the prediction module is further configured to, for any j-th prediction model in the plurality of prediction models, perform the following processing:

[0046] performing trend prediction processing on the multi-source fusion features of the prediction period sample through the j-th prediction model, to obtain a predicted infectious disease trend corresponding to the j-th prediction model;

[0047] determining the influence coefficient of the j-th prediction model based on the annotated infectious disease trend of the prediction period sample and the predicted infectious disease trend corresponding to the j-th prediction model;

[0048] determining the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of the plurality of prediction models;

[0049] where 0<j≤N, N represents the total number of the prediction models, and j and N are natural numbers.

[0050] In the above technical solution, the prediction module is further configured to calculate the first variance between the annotated infectious disease trend of the prediction period sample and the predicted infectious disease trend corresponding to the j-th prediction model;

[0051] The first variance is normalized to obtain the influence coefficient of the j-th prediction model;

[0052] Determine the second variance of the constant and the influence coefficient of the j-th prediction model;

[0053] The second variance is normalized based on the influence coefficients of the multiple prediction models to obtain the weight of the j-th prediction model.

[0054] In the above technical solution, the infectious disease trend prediction method is implemented through a neural network model, and the device further includes:

[0055] The training module is used to perform the following processes using the initialized neural network model:

[0056] Based on the search data samples for infectious diseases within the prediction period and the environmental characteristics of the prediction period, infectious disease trend prediction processing is performed to obtain the predicted infectious disease trend of the prediction period sample.

[0057] Based on the predicted infectious disease trends of the samples during the prediction period and the infectious disease trend labels of the samples during the prediction period, the loss function of the neural network model is constructed.

[0058] The parameters of the neural network model are updated based on the loss function, and the updated parameters of the neural network model are used as the parameters of the trained neural network model.

[0059] This application provides an electronic device for predicting infectious disease trends, the electronic device comprising:

[0060] Memory, used to store executable instructions;

[0061] When the processor executes the executable instructions stored in the memory, it implements the artificial intelligence-based infectious disease trend prediction method provided in the embodiments of this application.

[0062] This application provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, implement the artificial intelligence-based infectious disease trend prediction method provided in this application.

[0063] This application provides a computer program product, including a computer program or instructions, characterized in that, when the computer program or instructions are executed by a processor, they implement the artificial intelligence-based infectious disease trend prediction method provided in this application.

[0064] The embodiments of this application have the following beneficial effects:

[0065] By combining environmental characteristics of the prediction period with search data on infectious diseases during that period, infectious disease trend prediction can be made full use of multi-source data on infectious diseases to achieve accurate trend prediction, which has important reference value for prevention and control work. Compared with the scheme that relies on the number of historical cases of infectious diseases for prediction, the prediction of infectious disease trend through multi-source data during the prediction period improves the real-time performance of infectious disease trend prediction. Attached Figure Description

[0066] Figure 1 This is a schematic diagram illustrating an application scenario of the infectious disease trend prediction system provided in this application embodiment;

[0067] Figure 2 This is a schematic diagram of the structure of an electronic device for predicting infectious disease trends provided in an embodiment of this application;

[0068] Figures 3-5 This is a flowchart illustrating the artificial intelligence-based infectious disease trend prediction method provided in an embodiment of this application.

[0069] Figure 6 This is a flowchart of the AI ​​word selection system provided in the embodiments of this application;

[0070] Figure 7 This is a schematic diagram of the architecture of an influenza prediction model provided in an embodiment of this application;

[0071] Figure 8 This is a schematic diagram of the model construction provided in the embodiments of this application;

[0072] Figure 9 This is a graph of the training set predictions provided in the embodiments of this application;

[0073] Figure 10 This is a graph of the test set predictions provided in the embodiments of this application. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0075] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0077] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0078] 1) Responding to: used to indicate the conditions or states on which the operation is performed depends. When the conditions or states on which it depends are met, one or more operations can be performed in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations are performed.

[0079] 2) Infectious disease trends: Data that can characterize the trend of infectious disease transmission, such as the percentage of infectious disease cases, the number of infectious disease cases, and the inflection point of new infectious disease cases.

[0080] This application provides an artificial intelligence-based method, apparatus, electronic device, computer-readable storage medium, and computer program product for predicting infectious disease trends, which can automatically and accurately predict infectious disease trends.

[0081] The AI-based infectious disease trend prediction method provided in this application can be implemented by a terminal alone; or it can be implemented collaboratively by a terminal and a server. For example, the terminal can independently implement the AI-based infectious disease trend prediction method described below, or the terminal can send a trend prediction request for infectious diseases to the server. The server can then perform trend prediction based on the received trend prediction for infectious diseases, combined with the environmental characteristics of the prediction period and the search data for infectious diseases during the prediction period, to obtain the infectious disease trend (e.g., the percentage of influenza-like cases) during the prediction period. This can then be used for applications such as early warning of future infectious disease outbreaks based on the infectious disease trend.

[0082] The electronic device for predicting infectious disease trends provided in this application can be various types of terminals or servers. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, smart TV, in-vehicle device, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0083] Taking servers as an example, such as server clusters deployed in the cloud, artificial intelligence cloud services (AI as a Service, AIaaS) are opened to users (such as medical personnel, government health department personnel or the general public). The AIaaS platform will break down several common AI services and provide independent or packaged services in the cloud. This service model is similar to an AI-themed mall. All users can access and use one or more artificial intelligence services provided by the AIaaS platform through application programming interfaces.

[0084] For example, one type of artificial intelligence cloud service could be an infectious disease trend prediction service, whereby a cloud server encapsulates the infectious disease trend prediction program provided in this application embodiment. Users invoke the infectious disease trend prediction service in the cloud service through a terminal (running a client, such as an infectious disease early warning client), causing the cloud-deployed server to invoke the encapsulated infectious disease trend prediction program. Based on search data, feature extraction processing is performed to obtain search features for the infectious disease during the prediction period. Multi-source fusion processing is then performed on the environmental features and search features during the prediction period to obtain multi-source fusion features for the infectious disease. Finally, trend prediction processing is performed on these multi-source fusion features to obtain the infectious disease trend for the prediction period.

[0085] As an application example, for infectious disease early warning applications, the artificial intelligence-based infectious disease trend prediction method of this application combines the environmental characteristics of the prediction period and the search data for infectious diseases during the prediction period to predict the trend of infectious diseases during the prediction period (e.g., the percentage of influenza-like cases). This allows the government or the public to take corresponding prevention and control measures based on the infectious disease trend (e.g., the percentage of influenza-like cases) during the prediction period, including related measures such as the organization / allocation of medical personnel, the production / allocation of medical supplies, and the organization and management of social production activities.

[0086] See Figure 1 , Figure 1 This is a schematic diagram of the application scenario of the infectious disease trend prediction system 10 provided in the embodiments of this application. The terminal 200 is connected to the server 100 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two.

[0087] Terminals (running clients, such as infectious disease early warning clients) can be used to obtain trend prediction requests for infectious diseases. For example, users (such as medical personnel, government health department personnel, or the general public) input the prediction period through the terminal, and the terminal automatically obtains the trend prediction request for infectious diseases (including the prediction period).

[0088] In some embodiments, the client running on terminal 200 may be embedded with an infectious disease trend prediction plugin to implement an AI-based infectious disease trend prediction method locally on the client. For example, terminal 200 calls the infectious disease trend prediction plugin to implement an AI-based infectious disease trend prediction method, acquires environmental characteristics of the prediction period and search data for infectious diseases during the prediction period, performs feature extraction processing based on the search data to obtain search features for the infectious disease during the prediction period, performs multi-source fusion processing on the environmental characteristics and search features of the prediction period to obtain multi-source fusion features for the infectious disease, and performs trend prediction processing on the multi-source fusion features for the infectious disease to obtain the infectious disease trend (e.g., percentage of influenza-like cases) during the prediction period, so that the government or the public can take corresponding prevention and control measures based on the infectious disease trend during the prediction period, including related measures such as the organization / allocation of medical personnel, the production / allocation of medical supplies, and the organization and management of social production activities.

[0089] In some embodiments, after receiving a trend prediction request for an infectious disease, the terminal 200 calls the infectious disease trend prediction interface of the server 100 (which can be provided as a cloud service, i.e., an infectious disease trend prediction service). Based on the trend prediction request, the server 100 obtains the environmental characteristics of the prediction period and the search data for the infectious disease during the prediction period. Based on the search data, it performs feature extraction processing to obtain the search features for the infectious disease during the prediction period. It then performs multi-source fusion processing on the environmental characteristics and search features during the prediction period to obtain multi-source fusion features for the infectious disease. Finally, it performs trend prediction processing on the multi-source fusion features for the infectious disease to obtain the infectious disease trend during the prediction period (e.g., the percentage of influenza-like cases). The server 100 then sends the infectious disease trend during the prediction period to the terminal 200 to present the infectious disease trend during the prediction period, so that the government or the public can take corresponding prevention and control measures based on the infectious disease trend during the prediction period, including related measures such as the organization / allocation of medical personnel, the production / allocation of medical supplies, and the organization and management of social production activities.

[0090] In some embodiments, the terminal or server can implement the artificial intelligence-based infectious disease trend prediction method provided in this application by running a computer program, wherein the computer program is as follows: Figure 1 The client running in the terminal 200 shown can be, for example, a computer program that is a native program or software module in an operating system; it can be a native application (APP), i.e., a program that needs to be installed in the operating system to run; it can also be a small program, i.e., a program that only needs to be downloaded to a browser environment to run; or it can be a small program that can be embedded in any APP. In short, the above-mentioned computer program can be any form of application, module or plugin.

[0091] In some embodiments, multiple servers can form a blockchain, and server 100 is a node on the blockchain. Information connections can exist between each node in the blockchain, and information can be transmitted between nodes through these connections. The data related to the AI-based infectious disease trend prediction method provided in this application embodiment (e.g., the logic of infectious disease trend prediction, the infectious disease trend during the prediction period) can be stored on the blockchain.

[0092] The structure of the electronic device for predicting infectious disease trends provided in the embodiments of this application is described below. See also... Figure 2 , Figure 2 This is a schematic diagram of the structure of an electronic device 500 for predicting infectious disease trends provided in an embodiment of this application. The explanation uses an example where the electronic device 500 is a terminal. Figure 2 The illustrated electronic device 500 for knowledge generation includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 540.

[0093] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0094] Memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory. Memory 550 may optionally include one or more storage devices physically located away from processor 510.

[0095] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0096] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0097] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0098] In some embodiments, the AI-based infectious disease trend prediction device provided in this application can be implemented in software. The AI-based infectious disease trend prediction device provided in this application can be provided in various software embodiments, including various forms such as applications, software, software modules, scripts, or code.

[0099] Figure 2 An AI-based infectious disease trend prediction device 555, stored in memory 550, is shown. It can be software in the form of programs and plug-ins, and includes a series of modules, including an acquisition module 5551, a feature extraction module 5552, a fusion module 5553, a prediction module 5554, an update module 5555, and a training module 5556. These modules are logically connected and can therefore be arbitrarily combined or further split according to the functions they implement. The functions of each module will be described below.

[0100] As previously stated, the artificial intelligence-based infectious disease trend prediction method provided in this application can be implemented by various types of electronic devices. See also Figure 3 , Figure 3 This is a flowchart illustrating the artificial intelligence-based infectious disease trend prediction method provided in this application embodiment, combined with...Figure 3 The steps shown are explained.

[0101] In the following steps, the forecast period can be in days, weeks, or months, that is, the infectious disease trend can be predicted for a certain day, a certain week, or a certain month.

[0102] In the following steps, environmental features characterize environmental information related to the infectious disease during the prediction period. These environmental features include at least one of the following: meteorological features and air quality features. For example, meteorological features include temperature, humidity, perceived temperature, maximum temperature, minimum temperature, wind speed V component, and wind speed U component; air quality features include PM2.5, PM10, sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and ozone (O3). Search data includes keywords related to infectious diseases searched on search engines, such as "fever," "fatigue," and "antiviral"—keywords related to infectious diseases entered into search engines.

[0103] In step 101, the environmental characteristics of the prediction period are obtained, and the search data for infectious diseases during the prediction period are obtained, including keywords for searching for infectious diseases.

[0104] As an example of acquiring environmental characteristics and searching data, medical personnel can input the predicted period of an infectious disease on the terminal's input interface. After the input is completed, the terminal automatically generates a trend prediction request for the infectious disease (including the predicted period) and forwards the trend prediction request to the server. After receiving the trend prediction request for the infectious disease, the server parses the request to obtain the predicted period and performs data crawling based on a search engine to obtain search data for the infectious disease during the predicted period, such as keywords for the infectious disease entered in the search engine during the predicted period, such as "fever," "fatigue," and "antiviral." Meteorological characteristics such as temperature, humidity, perceived temperature, maximum temperature, minimum temperature, wind speed V component, and wind speed U component are obtained through a meteorological platform, and air characteristics such as PM2.5, PM10, sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and ozone (O3) are obtained through an air pollutant platform.

[0105] In step 102, feature extraction is performed based on the search data to obtain search features for infectious diseases during the prediction period.

[0106] For example, after acquiring search data, since neural network models process data for digital information, it is necessary to transform the search data into digital search features.

[0107] See Figure 4 , Figure 4This is an optional flowchart illustrating an artificial intelligence-based infectious disease trend prediction method provided in an embodiment of this application. Figure 4 Show Figure 3 Step 102 in the process can be achieved through Figure 4 Steps 1021-1025 are implemented as follows: In step 1021, a keyword lexicon for the infectious disease during the prediction period is determined; in step 1022, the keywords in the keyword lexicon are matched with the keywords in the search data to obtain the search keywords for the prediction period; in step 1023, the search keywords are statistically processed based on the search data to obtain the search indicators of the search keywords; in step 1024, the search keywords are subjected to feature transformation processing to obtain the semantic features of the search keywords; in step 1025, the set of the semantic features of the search keywords and the search indicators of the search keywords is used as the search features for the infectious disease during the prediction period.

[0108] For example, after acquiring search data for infectious diseases during the prediction period, a keyword thesaurus for infectious diseases during the prediction period is determined. When a keyword in the search data exists in the keyword thesaurus, that keyword is used as the search keyword for the prediction period. That is, the keywords in the keyword thesaurus are matched with the keywords in the search data, and the successfully matched keywords are used as the search keywords for the prediction period. Statistical processing is performed on the search keywords based on the search data to obtain search indicators (such as search frequency) for the search keywords. Feature transformation processing is then performed on the search keywords to obtain semantic features for the search keywords. The set of semantic features and search indicators of the search keywords is used as the search features for infectious diseases during the prediction period. Subsequently, trend prediction is performed by combining search features and environmental features. Thus, based on multiple feature dimensions, it is possible to accurately predict the trend of infectious diseases and take timely preventive and protective measures to prevent the spread of infectious diseases.

[0109] In some embodiments, determining a keyword lexicon for an infectious disease during the prediction period includes: determining a first correlation between multiple candidate keywords for the prediction period and target words of the infectious disease; filtering the multiple candidate keywords based on the first correlation to obtain keywords for updating the lexicon; updating a virus feature lexicon based on the keywords for updating the lexicon to obtain an updated virus feature lexicon; determining a second correlation between keywords in the updated virus feature lexicon and target words of the infectious disease; filtering the keywords in the updated virus feature lexicon based on the second correlation to obtain multiple target keywords; and constructing a keyword lexicon for the infectious disease during the prediction period based on the multiple target keywords.

[0110] It should be noted that users' search information for infectious diseases may change over time. In order to address the issue of changes in human language behavior over time, this application embodiment updates the keyword lexicon regularly and then uses a keyword lexicon that closely resembles human language behavior to collect search data. When updating the keyword thesaurus, the first step is to calculate the primary relevance (e.g., vector relevance) between multiple candidate keywords for the prediction period and the target term for the infectious disease (e.g., "influenza"). Based on this primary relevance, the candidate keywords are sorted in descending order. The top-ranked keywords are used to update the thesaurus. These updated keywords are then added to the virus feature thesaurus (also known as a virus-feature-based keyword thesaurus, which includes keywords related to influenza, symptoms, treatment, drugs, and transmission from infectious disease literature). This results in an updated virus feature thesaurus. The second relevance (e.g., vector relevance) between the keywords in the updated virus feature thesaurus and the target term for the infectious disease is then calculated. Based on this second relevance, the keywords in the updated virus feature thesaurus are sorted in descending order. The top-ranked keywords are used as target keywords. This allows for the construction of a keyword thesaurus for the prediction period based on these target keywords. This continuous updating of the keyword thesaurus based on the target term for the infectious disease improves its real-time performance, addresses the issue of changes in human language behavior over time, and ultimately enhances the accuracy of search data.

[0111] In some embodiments, before constructing a keyword lexicon for infectious diseases during the prediction period, the predicted infectious disease trend and the infectious disease trend annotation of the sample during the prediction period are obtained; when the difference between the predicted infectious disease trend and the infectious disease trend annotation is greater than the difference threshold, it is determined that the operation of constructing a keyword lexicon for infectious diseases during the prediction period will be performed.

[0112] For example, when human language behavior doesn't change significantly, there's no need to update the keyword database. Therefore, updating the keyword database is only necessary when human language behavior changes substantially. For instance, if the difference between the predicted infectious disease trend and the labeled infectious disease trend exceeds a certain threshold, it indicates that the infectious disease trend prediction is inaccurate, possibly due to a significant change in human language behavior. The difference between the predicted infectious disease trend and the labeled infectious disease trend can be the difference between the two, or it can be the quadratic variance between them.

[0113] In some embodiments, feature transformation processing is performed on search keywords to obtain semantic features of search keywords, including: performing vector mapping processing on search keywords to obtain embedding features of search keywords; and performing semantic transformation processing on the embedding features of search keywords to obtain semantic features of search keywords.

[0114] For example, by using the word embedding method, search keywords are mapped to another numerical vector space to obtain the embedding features of the search keywords. Then, the semantic features of the search keywords are obtained by performing semantic transformation on the embedding features of the search keywords through a fully connected neural network.

[0115] In some embodiments, semantic transformation is achieved through a fully connected neural network, which includes multiple cascaded fully connected layers. Semantic transformation is performed on the embedded features of the search keywords to obtain semantic features of the search keywords. This includes: performing semantic mapping on the embedded features of the search keywords through the first fully connected layer in the multiple cascaded fully connected layers; outputting the mapping result of the first fully connected layer to subsequent cascaded fully connected layers, continuing semantic mapping processing and outputting mapping results through subsequent cascaded fully connected layers until the last fully connected layer is reached, and using the mapping result output by the last fully connected layer as the semantic features of the search keywords.

[0116] For example, a fully connected neural network consists of multiple cascaded fully connected layers. The first fully connected layer performs semantic mapping on the embedded features of the search keywords and outputs the mapping result to the second fully connected layer. The second fully connected layer continues semantic mapping and outputs the mapping, and so on, until the mapping result output by the last fully connected layer is used as the semantic feature of the search keywords. Through cascaded semantic mapping, progressively accurate semantic features can be obtained, ultimately yielding features that accurately contain all semantic information.

[0117] In step 103, the environmental characteristics and search characteristics of the prediction period are subjected to multi-source fusion processing to obtain multi-source fusion characteristics for infectious diseases.

[0118] For example, after obtaining the environmental and search features for the prediction period, the environmental and search features for the prediction period are fused to obtain multi-source fusion features for infectious diseases. Then, trend prediction is made by using multi-source fusion features that combine search and environmental features. Based on multiple feature dimensions, this makes it easier to accurately predict the trend of infectious diseases and take timely preventive and protective measures to prevent the spread of infectious diseases.

[0119] In some embodiments, before performing multi-source fusion processing on the environmental features and search features for the prediction period, the environmental features for the prediction period are standardized to obtain standardized environmental features; the search index in the search features is standardized to obtain standardized search features; and then the standardized environmental features and standardized search features are subjected to multi-source fusion processing to obtain multi-source fusion features for infectious diseases.

[0120] For example, since environmental features and search features (multi-source data) come from different domains and have different units of measurement, in order to enable the model to converge faster when learning parameter optimization (such as gradient descent), the environmental features and search features are standardized before the multi-source data is input into the model.

[0121] In some embodiments, multi-source fusion processing is performed on environmental features and search features during the prediction period to obtain multi-source fusion features for infectious diseases, including: splicing environmental features and search features during the prediction period to obtain multi-source fusion features for infectious diseases; or, weighted summation processing is performed on environmental features and search features during the prediction period to obtain multi-source fusion features for infectious diseases.

[0122] For example, environmental features include at least one of the following: meteorological features and air features. The meteorological features, air features, and search features of the prediction period are spliced ​​together to obtain multi-source fusion features for infectious diseases.

[0123] For example, by performing a weighted summation of meteorological characteristics, air characteristics, and search characteristics during the prediction period, multi-source fusion characteristics for infectious diseases can be obtained.

[0124] In step 104, trend prediction processing is performed on the multi-source fusion characteristics of infectious diseases to obtain the infectious disease trend for the prediction period.

[0125] For example, after acquiring multi-source fusion features for infectious diseases, trend prediction can be performed by integrating multi-source fusion features with search features and environmental features. This allows for accurate prediction of infectious disease trends based on multiple feature dimensions, enabling timely preventive and protective measures to prevent the spread of infectious diseases.

[0126] See Figure 5 , Figure 5 This is an optional flowchart illustrating an artificial intelligence-based infectious disease trend prediction method provided in an embodiment of this application. Figure 5 Show Figure 3 Step 104 in the process can be achieved through Figure 5The illustrated steps 1041 to 1042 are implemented as follows: In step 1041, trend prediction processing is performed on multi-source fusion features for infectious diseases based on a plurality of prediction models, to obtain the infectious disease trend corresponding to each prediction model; In step 1042, fusion processing is performed on the infectious disease trend corresponding to each prediction model based on the weight of each prediction model, to obtain the infectious disease trend for the prediction period.

[0127] For example, the multi-source fusion features for infectious diseases are input into the eXtreme Gradient Boosting (XGBoost) machine learning regression model to predict ILI%, so as to obtain the infectious disease trend (predicted value of ILI%) for the prediction period, where the plurality of prediction models may be models with different structures. Thus, by fusing the prediction results of multiple prediction models, comprehensive prediction information is obtained, and the prediction effect is improved.

[0128] In some embodiments, before performing fusion processing on the infectious disease trend corresponding to each prediction model based on the weight of each prediction model to obtain the infectious disease trend for the prediction period, for any j-th prediction model among the plurality of prediction models, the following processing is performed: performing trend prediction processing on the multi-source fusion features of the prediction period sample through the j-th prediction model, to obtain the predicted infectious disease trend corresponding to the j-th prediction model; determining the influence coefficient of the j-th prediction model based on the annotated infectious disease trend of the prediction period sample and the predicted infectious disease trend corresponding to the j-th prediction model; determining the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of the plurality of prediction models; where 0<j≤N, N represents the total number of prediction models, and j and N are natural numbers.

[0129] For example, the embodiments of the present application may determine the weight of each prediction model through the influence coefficient of the prediction model. For example, for the j-th prediction model, the influence coefficient of the j-th prediction model may be determined based on the annotated infectious disease trend of the prediction period sample and the predicted infectious disease trend corresponding to the j-th prediction model. The larger the influence coefficient of the j-th prediction model, the worse the prediction capability of the j-th prediction model, so the smaller the weight of the j-th prediction model, so as to avoid negative impact on infectious disease trend prediction.

[0130] In some embodiments, determining the influence coefficient of the j-th prediction model based on the infectious disease trend labeling of the sample during the prediction period and the predicted infectious disease trend corresponding to the j-th prediction model includes: labeling the infectious disease trend of the sample during the prediction period and the first variance of the predicted infectious disease trend corresponding to the j-th prediction model; normalizing the first variance to obtain the influence coefficient of the j-th prediction model; and determining the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of multiple prediction models, including: determining the second variance of the constant and the influence coefficient of the j-th prediction model; and normalizing the second variance based on the influence coefficients of multiple prediction models to obtain the weight of the j-th prediction model.

[0131] For example, the influence coefficient of the j-th prediction model is Where i represents a sample for any prediction period, x i Let represent the infectious disease trend label for the sample during the prediction period, and represent the predicted infectious disease trend corresponding to the j-th prediction model. The weight of the j-th prediction model is .

[0132] It should be noted that the influence coefficient of the embodiments in this application is not limited to... Other variations of the formula are also possible; the weights of the j-th prediction model in this application embodiment are not limited to... It can also be other variations of the formula.

[0133] In some embodiments, the infectious disease trend prediction method is implemented through a neural network model. The training process of the neural network model includes: performing the following processing on the initialized neural network model: performing infectious disease trend prediction processing based on the search data samples for infectious diseases within the prediction period sample and the environmental characteristics of the prediction period to obtain the predicted infectious disease trend of the prediction period sample; constructing the loss function of the neural network model based on the predicted infectious disease trend of the prediction period sample and the infectious disease trend label of the prediction period sample; updating the parameters of the neural network model based on the loss function, and using the updated parameters of the neural network model as the parameters of the trained neural network model.

[0134] For example, after determining the value of the loss function of the neural network model, it can be determined whether the value of the loss function has reached a preset threshold. When the value of the loss function has not reached the preset threshold, the error signal of the neural network model is determined based on the loss function, the error information is backpropagated in the neural network model, and the model parameters of each layer are updated during the propagation process.

[0135] Here, we explain backpropagation. Training sample data is input into the input layer of the neural network model, passes through the hidden layers, and finally reaches the output layer to output the result. This is the forward propagation process of the neural network model. Since there is an error between the output result of the neural network model and the actual result, the error between the output result and the actual value is calculated and backpropagated from the output layer to the hidden layers until it reaches the input layer. During the backpropagation process, the values ​​of the model parameters are adjusted according to the error. That is, a loss function is constructed based on the error between the output result and the actual value, and the partial derivatives of the loss function with respect to the model parameters are calculated layer by layer to generate the gradient of the loss function with respect to the model parameters of each layer. Since the direction of the gradient indicates the direction of error amplification, the gradient of the model parameters is inverted and summed with the original parameters of each layer. The summation result is used as the updated model parameters of each layer, thereby reducing the error caused by the model parameters. The above process is iterated until convergence.

[0136] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0137] The embodiments of this application can be applied to various scenarios of infectious disease trend prediction, such as predicting influenza trends by predicting the percentage of influenza-like illness (ILI).

[0138] The following explanation uses influenza as an example of an infectious disease:

[0139] Influenza surveillance and early warning systems play a crucial role in influenza prevention and control. However, with global economic integration, informatization, and climate change, these systems are no longer adequate to adapt to such complex changes. The current influenza surveillance and early warning system suffers from a series of problems, including high operating costs, low network coverage, low information reporting efficiency, over-reliance on historical influenza data without multi-dimensional and multi-source data support, overly simplistic data mining and prediction methods, difficulty in real-time data acquisition and updates, and challenges in extrapolating models. Furthermore, sentinel hospitals collect only localized, random samples, not the entire population, making data acquisition difficult and expensive, and resulting in delayed warnings, sometimes as late as one week. Therefore, developing a more efficient and accurate real-time early warning system and establishing an influenza early warning system based on multi-source data to further improve the ability to predict and warn of influenza outbreaks is imperative.

[0140] To address the aforementioned issues, this application introduces multi-source data, including AI-automated keyword indexing of internet data, weather data, and air pollutant data, to establish an influenza trend prediction method. The multi-source data includes: 1) search engine data based on AI-automated keyword selection, highly correlated with influenza symptoms and characteristics; 2) high spatiotemporal resolution meteorological data: perceived temperature, maximum temperature, minimum temperature, wind speed V component, wind speed U component, humidity, and air temperature; 3) high spatiotemporal resolution air pollutant data: fine particulate matter (PM2.5), sulfur dioxide (SO2), nitrogen dioxide (NO2), inhalable particulate matter (PM10), carbon monoxide (CO), and ozone (O3). This application uses ILI% data from 2011 to 2019. Based on this multi-source data, an influenza trend prediction model is established. Over time, the multi-source data used by the influenza trend prediction model will be updated, the ILI% training data will increase, and the key parameters of the influenza trend prediction model will be automatically updated to further improve prediction accuracy and enable real-time prediction of influenza trends.

[0141] The influenza trend prediction model (implemented by a neural network model) used in this application to predict real-time influenza outbreaks has the following significance:

[0142] 1. It helps alleviate public anxiety. The spread of infectious diseases affects almost everyone's life. Predicting the trend of the epidemic can help the public understand the development of the epidemic and help the audience take appropriate protective measures and make reasonable arrangements for their lives.

[0143] 2. It helps relevant agencies to introduce relevant policies in advance to prevent the influenza pandemic. By predicting the influenza trend through influenza trend prediction models, the severity of the future epidemic can be judged in advance, similar to weather forecasts, typhoon forecasts, and earthquake forecasts. In response to the potential future spread of influenza, it can provide a reference for local governments to introduce corresponding control measures and make full preparations.

[0144] 3. It can help public health and safety experts understand the factors influencing the virus. The correlation between the keyword index, meteorological characteristics, air pollutant characteristics and the degree of influenza prevalence included in the influenza trend prediction model can provide relevant experts with more information on the factors influencing the influenza virus.

[0145] It should be noted that the embodiments of this application include several important parts, namely, the AI ​​word selection system, data collection, data processing, multi-source data fusion and model building. The following is a detailed description of the AI ​​word selection system, data collection, data processing, multi-source data fusion and model building:

[0146] 1) AI word selection system

[0147] Search indexes were used as the characteristics of internet search indexes. Based on the increasing trend of residents using computers and smartphones for information searches, the search data source was the "Personal Computer (PC) + Mobile" index. On the one hand, starting from the characteristics of the influenza virus and referring to frequently cited literature related to influenza, a preliminary keyword vocabulary related to influenza disease, symptoms, treatment, drugs, and transmission was established, such as fever, fatigue, and antiviral. On the other hand, to address the issue of changes in human language behavior over time, an AI-based word selection system was designed to update and filter the keyword vocabulary, and then the final keyword vocabulary was used to collect search data. The specific structure and principle of the AI ​​word selection system are as follows:

[0148] Word vectors are a way to mathematically represent words in a language; a word vector represents a word as a vector. A suitable method is needed to convert text data into numerical data, which leads to the concept of word embedding. If we consider a word as the smallest unit of text, word embedding can be understood as a mapping; the process is to map or embed a word in the text space into another numerical vector space. In this embodiment, the keyword vocabulary includes multiple words, and word embedding technology based on word-to-vector conversion methods (such as word2vec) is used. Figure 6 As shown, the vector similarity between candidate words and "influenza" is first calculated, and then sorted in descending order based on the vector similarity. The top 100 keywords are selected, and then the keyword vocabulary based on virus characteristics is updated based on the 100 keywords to obtain the updated vocabulary. The vector similarity between the words in the updated vocabulary and "influenza" is calculated again, and then sorted in descending order based on the vector similarity. The top 50 words are selected as the final keyword vocabulary. Data is collected based on the final keyword vocabulary, and an influenza prediction model is built based on the collected data.

[0149] 2) Data collection

[0150] Weather and meteorological data: Based on evidence from relevant literature that there is a certain correlation between weather and virus transmission, meteorological data such as temperature, humidity, perceived temperature, maximum temperature, minimum temperature, wind speed V component, and wind speed U component are obtained through meteorological platforms.

[0151] Air pollutant data: Based on evidence from relevant literature regarding the relationship between air pollutants and virus transmission, air data such as PM2.5, PM10, sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and ozone (O3) are obtained through a platform in the field of air pollutants.

[0152] 3) Data processing

[0153] 1. The multi-source data collected in this application embodiment are all daily data. Based on the characteristic that the incubation period of influenza virus is generally 1-7 days, the time interval is set to a week. The weekly average is calculated using the daily data and then input into the influenza prediction model.

[0154] 2. Since the data from multiple sources are from different domains and have different units of measurement, in order to enable the model to converge faster when learning parameters (such as gradient descent), the features are standardized using the following formula (1) before inputting the data from multiple sources into the model:

[0155]

[0156] Where X represents a random variable (in this embodiment, it represents an input feature), X_min represents the minimum value of the random variable X within its range, and X_max represents the maximum value of the random variable X within its range. std This represents a standardized random variable.

[0157] 3. Although a large amount of keyword index search data, meteorological data, and air pollutant data were collected, in order to prevent the introduction of more noise into the model, the Pearson correlation coefficient between each feature (including search features, meteorological features, and air features) and influenza cases was first calculated, and hypothesis testing of the correlation coefficient was performed. Finally, features with significant statistical correlation with influenza were input to build an influenza prediction model to improve the accuracy of the model prediction.

[0158] 4) Multi-source data fusion and model building

[0159] Multi-source data fusion architecture: such as Figure 7 As shown, for the search keywords selected by the AI ​​keyword selection system, the pre-trained Word Embedding is used as the input layer. Through a fully connected neural network, the semantic features of a hidden layer are output. In the hidden layer, the semantic features are fused with features from meteorological data, air quality data, and the standardized features of the search index. Finally, these features are input into a machine learning regression model using extreme gradient boosting (XGBoost) to predict the ILI%. This data fusion model can capture the semantic information of keywords and combine other relevant macroscopic features to obtain comprehensive information and improve prediction performance.

[0160] For the final prediction layer, this embodiment employs XGBoost, an ensemble machine learning method based on decision trees, which iteratively combines a set of weak prediction models into a strong prediction model. XGBoost uses a second-order Taylor series to approximate the value of the loss function and further reduces the possibility of overfitting by applying regularization. In setting the hyperparameters, the optimal parameter values ​​are found by performing a grid search on possible combinations of values. The main fine-tuning parameters in the prediction model of this embodiment include the number of trees (n_estimator = 100), the learning rate (eta = 0.1), the maximum tree depth (max_depth = 3), and the subsample ratio (subsample = 1). Figure 8 As shown, this application embodiment collected data from 2011 to 2020. Considering the impact and interference caused by the pneumonia epidemic, data before 2020 was selected. Data from 2011 to 2019 was used to train the model and adjust parameters, and data from 2019 to 2020 was used as the test set to evaluate the model accuracy. In the model evaluation, the Mean Absolute Percent Error (MAPE) as shown in formula (2) was used:

[0161]

[0162] Where, n samples Let y represent the sample size and y represent the ILI% observation value. This represents the ILI% model prediction value, where i represents each sample. (∈,|y i |) represents y i The expected value.

[0163] like Figures 9-10 As shown, the predicted curves closely match the true curves on both the training and test sets, with an average deviation of 13.1% on the test set and 0.9% on the training set, indicating that the model's predictions are accurate.

[0164] To compare the influenza prediction model based on AI-selected multi-source data proposed in this application with other technical solutions, the following comparative experiments were designed: The influenza prediction model based on AI-selected multi-source data incorporates all features from multi-source data for model training; the meteorological data + air data model excludes search data from search engines and trains the model with the same parameters; the search data + air data model excludes weather and meteorological data and trains the model with the same parameters; and the search data + meteorological data model excludes air pollutant feature data and trains the model with the same parameters.

[0165] Finally, the MAPE (Mean Absolute Percent Error) of these four models was compared horizontally on the training and test sets to determine the evaluation results. The comparative experiments are shown in Table 1.

[0166] Table 1

[0167]

[0168]

[0169] As shown in Table 1, the experimental results indicate that the evaluation metrics of other models are significantly lower than those of the influenza prediction model based on multi-source data. The influenza prediction model based on multi-source data outperforms other models in both the training and test sets in terms of MAPE. Specifically, the average prediction bias on the test set is 13.1%, which is 4.1%, 1.9%, and 0.2% higher than other models, respectively. Therefore, the influenza prediction model based on multi-source data can effectively improve the performance of the model.

[0170] Utilizing multi-source data enables real-time, rapid, and accurate prediction of regional influenza, significantly improving the accuracy of influenza forecasting. Embodiments of this application can be used to construct a real-time prediction and early warning system for regional influenza or other infectious diseases, allowing disease control centers and the public to anticipate current infectious disease trends and take timely preventative and protective measures to prevent the spread of infectious diseases.

[0171] In summary, the influenza prediction model based on multi-source data provided in this application has the following beneficial effects:

[0172] 1) In terms of data dimensions, instead of point-based local data collection, we capture multi-source data across the entire network in an open and automated manner. We use search data from search engines, meteorological data that affects the spread of influenza, and air pollutant data to construct multiple feature dimensions in order to accurately predict the regular trend of influenza, non-seasonal influenza activity, and the outbreak of influenza epidemics.

[0173] 2) For multi-source data, feature engineering in machine learning is used to automatically select predictive variables. As time goes by, the model parameters will be continuously fine-tuned and updated with the latest ILI% release to more accurately predict influenza trends.

[0174] 3) Keyword fields highly related to influenza are automatically selected and updated based on AI, rather than being fixed by humans. This method can adapt to changes in human language habits, thereby more accurately predicting influenza trends.

[0175] This concludes the description of the artificial intelligence-based infectious disease trend prediction method provided in this application, in conjunction with exemplary applications and implementations of the electronic devices provided in the embodiments of this application. This application also provides an artificial intelligence-based infectious disease trend prediction device. In practical applications, the functional modules in the artificial intelligence-based infectious disease trend prediction device can be collaboratively implemented by the hardware resources of the electronic device (such as a terminal, server, or server cluster), such as computing resources like processors, communication resources (such as those supporting various communication methods like optical fiber and cellular), and memory. Figure 2 The device 555, which is an artificial intelligence-based infectious disease trend prediction device stored in memory 550, is shown. It can be software in the form of programs and plug-ins, such as software modules designed in programming languages ​​such as C / C++ and Java, application software designed in programming languages ​​such as C / C++ and Java, or dedicated software modules, application programming interfaces, plug-ins, cloud services, etc. in large software systems. Examples of different implementation methods are given below.

[0176] The artificial intelligence-based infectious disease trend prediction device 555 includes a series of modules, including an acquisition module 5551, a feature extraction module 5552, a fusion module 5553, a prediction module 5554, an update module 5555, and a training module 5556. The following description further illustrates how the various modules in the artificial intelligence-based infectious disease trend prediction device 555 provided in this application cooperate to implement an infectious disease trend prediction scheme.

[0177] The acquisition module 5551 is used to acquire environmental characteristics of the prediction period and search data for infectious diseases during the prediction period, wherein the search data includes keywords for searching for the infectious disease; the feature extraction module 5552 is used to perform feature extraction processing based on the search data to obtain search features for the infectious disease during the prediction period; the fusion module 5553 is used to perform multi-source fusion processing on the environmental characteristics of the prediction period and the search features to obtain multi-source fusion features for the infectious disease; and the prediction module 5554 is used to perform trend prediction processing on the multi-source fusion features for the infectious disease to obtain the infectious disease trend during the prediction period.

[0178] In some embodiments, the feature extraction module 5552 is further configured to: determine a keyword lexicon for the infectious disease during the prediction period; match the keywords in the keyword lexicon with the keywords in the search data to obtain search keywords for the prediction period; perform statistical processing on the search keywords based on the search data to obtain search indicators for the search keywords; perform feature transformation processing on the search keywords to obtain semantic features of the search keywords; and use the set of semantic features of the search keywords and search indicators of the search keywords as search features for the infectious disease during the prediction period.

[0179] In some embodiments, the feature extraction module 5552 is further configured to determine a first correlation between multiple candidate keywords for the prediction period and the target words of the infectious disease; to filter the multiple candidate keywords based on the first correlation to obtain keywords for updating the terminology; to update the virus feature terminology based on the keywords for updating the terminology to obtain an updated virus feature terminology; to determine a second correlation between the keywords in the updated virus feature terminology and the target words of the infectious disease; to filter the keywords in the updated virus feature terminology based on the second correlation to obtain multiple target keywords; and to construct a keyword terminology for the infectious disease for the prediction period based on the multiple target keywords.

[0180] In some embodiments, before constructing the keyword lexicon for the infectious disease during the prediction period, the apparatus further includes: an update module 5555, configured to acquire the predicted infectious disease trend of the sample during the prediction period and the infectious disease trend label of the sample during the prediction period; when the difference between the predicted infectious disease trend and the infectious disease trend label is greater than a difference threshold, it is determined that the operation of constructing the keyword lexicon for the infectious disease during the prediction period will be performed.

[0181] In some embodiments, the feature extraction module 5552 is further configured to perform vector mapping processing on the search keywords to obtain the embedding features of the search keywords; and to perform semantic transformation processing on the embedding features of the search keywords to obtain the semantic features of the search keywords.

[0182] In some embodiments, the semantic conversion is implemented by a fully connected neural network, and the fully connected neural network includes a plurality of cascaded fully connected layers; the feature extraction module 5552 is further configured to perform semantic mapping processing on the embedding feature of the search keyword through the first fully connected layer among the plurality of cascaded fully connected layers; output the mapping result of the first fully connected layer to the subsequent cascaded fully connected layers, and continue performing semantic mapping processing and mapping result output through the subsequent cascaded fully connected layers until the result is output to the last fully connected layer; and take the mapping result output by the last fully connected layer as the semantic feature of the search keyword.

[0183] In some embodiments, before performing multi-source fusion processing on the environmental feature of the prediction period and the search feature, the fusion module 5553 is further configured to perform standardization processing on the environmental feature of the prediction period to obtain a standardized environmental feature; perform standardization processing on the search index in the search feature to obtain a standardized search feature; and perform multi-source fusion processing on the standardized environmental feature and the standardized search feature to obtain a multi-source fusion feature for the infectious disease.

[0184] In some embodiments, the fusion module 5553 is further configured to perform concatenation processing on the environmental feature of the prediction period and the search feature to obtain a multi-source fusion feature for the infectious disease; or perform weighted summation processing on the environmental feature of the prediction period and the search feature to obtain a multi-source fusion feature for the infectious disease; wherein the environmental feature includes at least one of the following: meteorological feature and air feature.

[0185] In some embodiments, the prediction module 5554 is further configured to perform trend prediction processing on the multi-source fusion feature for the infectious disease based on a plurality of prediction models, to obtain an infectious disease trend corresponding to each of the prediction models; and perform fusion processing on the infectious disease trend corresponding to each of the prediction models based on the weight of each of the prediction models, to obtain the infectious disease trend of the prediction period.

[0186] In some embodiments, the prediction module 5554 is further configured to, for any j-th prediction model among the plurality of prediction models, perform the following processing: perform trend prediction processing on the multi-source fusion feature of the prediction period sample through the j-th prediction model, to obtain a predicted infectious disease trend corresponding to the j-th prediction model; determine the influence coefficient of the j-th prediction model based on the annotated infectious disease trend of the prediction period sample and the predicted infectious disease trend corresponding to the j-th prediction model; determine the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of the plurality of prediction models; wherein 0<j≤N, N represents the total number of the prediction models, and j and N are natural numbers.

[0187] In some embodiments, the prediction module 5554 is further configured to: label the infectious disease trend of the sample during the prediction period and the first variance of the predicted infectious disease trend corresponding to the j-th prediction model; normalize the first variance to obtain the influence coefficient of the j-th prediction model; determine the second variance of the constant and the influence coefficient of the j-th prediction model; and normalize the second variance based on the influence coefficients of the plurality of prediction models to obtain the weight of the j-th prediction model.

[0188] In some embodiments, the infectious disease trend prediction method is implemented using a neural network model, and the apparatus further includes: a training module 5556, configured to perform the following processing using the initialized neural network model: perform infectious disease trend prediction processing based on search data samples for infectious diseases within the prediction period sample and environmental characteristics of the prediction period, to obtain the predicted infectious disease trend of the prediction period sample; construct a loss function for the neural network model based on the predicted infectious disease trend of the prediction period sample and the infectious disease trend label of the prediction period sample; update the parameters of the neural network model based on the loss function, and use the updated parameters of the neural network model as the parameters of the trained neural network model.

[0189] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the artificial intelligence-based infectious disease trend prediction method described above in this application.

[0190] This application provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to execute the artificial intelligence-based infectious disease trend prediction method provided in this application. For example... Figures 3-5 The example shown is an AI-based method for predicting infectious disease trends.

[0191] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0192] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0193] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0194] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0195] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for predicting infectious disease trends based on artificial intelligence, characterized in that, The method includes: The environmental characteristics of the prediction period are obtained, and search data for infectious diseases during the prediction period are obtained, wherein the search data includes keywords for searching for the infectious disease; Determine the first correlation between multiple candidate keywords for the prediction period and the target words of the infectious disease; Based on the first relevance, the multiple candidate keywords are filtered to obtain keywords for updating the thesaurus; The virus feature terminology is updated based on the keywords used to update the terminology, resulting in the updated virus feature terminology. Determine the second correlation between keywords in the updated virus feature word library and target words of the infectious disease; Based on the second correlation, the keywords in the updated virus feature word library are filtered to obtain multiple target keywords; Based on the multiple target keywords, a keyword lexicon for the infectious disease during the prediction period is constructed; Using the keyword database and performing feature extraction processing based on the search data, the search features for the infectious disease during the prediction period are obtained; Based on the Pearson correlation coefficient between the search features and environmental features and infectious disease cases, the search features and environmental features are screened to identify those that have a significant statistical correlation with the infectious disease. The environmental features of the predicted period and the search features after screening are subjected to multi-source fusion processing to obtain multi-source fusion features for the infectious disease. Trend prediction processing is performed on the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend for the prediction period.

2. The method according to claim 1, characterized in that, The step of utilizing the keyword database and performing feature extraction processing based on the search data to obtain search features for the infectious disease during the prediction period includes: The keywords in the keyword database are matched with the keywords in the search data to obtain the search keywords for the prediction period; Based on the search data, the search keywords are statistically processed to obtain the search metrics for the search keywords; The search keywords are subjected to feature transformation processing to obtain the semantic features of the search keywords; The set of semantic features of the search keywords and the search indicators of the search keywords are used as the search features for the infectious disease during the prediction period.

3. The method according to claim 1, characterized in that, Before constructing the keyword database for the infectious disease during the prediction period, the method further includes: Obtain the predicted infectious disease trend of the sample during the prediction period and the labeling of the infectious disease trend of the sample during the prediction period; When the difference between the predicted infectious disease trend and the labeled infectious disease trend is greater than a difference threshold, it is determined that the operation of constructing a keyword thesaurus for the infectious disease during the prediction period will be performed.

4. The method according to claim 2, characterized in that, The step of performing feature transformation processing on the search keywords to obtain the semantic features of the search keywords includes: The search keywords are subjected to vector mapping processing to obtain the embedding features of the search keywords; The semantic features of the search keywords are obtained by performing semantic transformation on the embedding features of the search keywords.

5. The method according to claim 4, characterized in that, The semantic transformation is achieved through a fully connected neural network, which includes multiple cascaded fully connected layers; The semantic transformation processing of the embedded features of the search keywords to obtain the semantic features of the search keywords includes: The embedding features of the search keywords are semantically mapped through the first fully connected layer in the plurality of cascaded fully connected layers. The mapping result of the first fully connected layer is output to the subsequent cascaded fully connected layers. The semantic mapping process and mapping result output are continued through the subsequent cascaded fully connected layers until the last fully connected layer is output. The mapping result output by the last fully connected layer is used as the semantic feature of the search keyword.

6. The method according to claim 1, characterized in that, Before performing multi-source fusion processing on the screened environmental features for the predicted period and the search features, the method further includes: The environmental characteristics of the predicted period after screening are standardized to obtain the standardized environmental characteristics. The search indices in the filtered search features are standardized to obtain standardized search features. The process of performing multi-source fusion processing on the screened environmental features for the predicted period and the search features to obtain multi-source fusion features for the infectious disease includes: The environmental features and search features after standardization are subjected to multi-source fusion processing to obtain multi-source fusion features for the infectious disease.

7. The method according to claim 1, characterized in that, The process of performing multi-source fusion processing on the screened environmental features for the predicted period and the search features to obtain multi-source fusion features for the infectious disease includes: The environmental features of the predicted period after screening and the search features are concatenated to obtain multi-source fusion features for the infectious disease; or, The environmental features of the predicted period after screening and the search features are weighted and summed to obtain multi-source fusion features for the infectious disease. The environmental characteristics include at least one of the following: meteorological characteristics and air characteristics.

8. The method according to claim 1, characterized in that, The trend prediction processing of the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend for the prediction period includes: Based on multiple prediction models, trend prediction processing is performed on the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend corresponding to each prediction model. Based on the weights of each prediction model, the infectious disease trends corresponding to each prediction model are fused to obtain the infectious disease trends for the prediction period.

9. The method according to claim 8, characterized in that, Before fusing the infectious disease trends corresponding to each prediction model based on the weights of each prediction model to obtain the infectious disease trend for the prediction period, the method further includes: For any j-th prediction model among the plurality of prediction models, the following processing is performed: By performing trend prediction processing on the multi-source fusion features of the sample during the prediction period using the j-th prediction model, the predicted infectious disease trend corresponding to the j-th prediction model is obtained. determining the influence coefficient of the j-th prediction model based on the infectious disease trend label of the sample in the prediction period and the predicted infectious disease trend corresponding to the j-th prediction model; determining the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of the plurality of prediction models; wherein 0<j≤N, N represents the total number of the prediction models, and j and N are natural numbers.

10. The method according to claim 9, characterized in that the step of determining the influence coefficient of the j-th prediction model based on the infectious disease trend label of the sample in the prediction period and the predicted infectious disease trend corresponding to the j-th prediction model comprises: determining the first variance of the predicted infectious disease trend corresponding to the j-th prediction model based on the infectious disease trend label of the sample in the prediction period; performing normalization processing on the first variance to obtain the influence coefficient of the j-th prediction model; the step of determining the weight of the j-th prediction model based on the influence coefficient of the j-th prediction model and the influence coefficients of the plurality of prediction models comprises: determining the second variance of a constant and the influence coefficient of the j-th prediction model; performing normalization processing on the second variance based on the influence coefficients of the plurality of prediction models to obtain the weight of the j-th prediction model.

11. The method according to claim 1, characterized in that, The infectious disease trend prediction method is implemented by a neural network model, and the training process of the neural network model comprises: executing the following processing by the initialized neural network model: performing infectious disease trend prediction processing based on the search data sample for infectious diseases in the prediction period sample and the environmental features of the prediction period to obtain the predicted infectious disease trend of the prediction period sample; constructing a loss function of the neural network model based on the predicted infectious disease trend of the prediction period sample and the infectious disease trend label of the prediction period sample; updating the parameters of the neural network model based on the loss function, and taking the updated parameters of the neural network model as the parameters of the trained neural network model.

12. An artificial intelligence-based infectious disease trend prediction device, characterized in that, The apparatus comprises: an acquisition module, configured to acquire the environmental features of a prediction period, and acquire search data for infectious diseases in the prediction period, wherein the search data comprises keywords searched for the infectious diseases; A feature extraction module is used to determine a first correlation between multiple candidate keywords for the prediction period and the target words of the infectious disease; to filter the multiple candidate keywords based on the first correlation to obtain keywords for updating the terminology; to update the virus feature terminology based on the keywords for updating the terminology to obtain an updated virus feature terminology; to determine a second correlation between the keywords in the updated virus feature terminology and the target words of the infectious disease; to filter the keywords in the updated virus feature terminology based on the second correlation to obtain multiple target keywords; to construct a keyword terminology for the infectious disease for the prediction period based on the multiple target keywords; to use the keyword terminology and perform feature extraction processing based on the search data to obtain search features for the infectious disease for the prediction period; and to filter the search features and environmental features based on the Pearson correlation coefficient between the search features, environmental features, and infectious disease cases to select the search features and environmental features that have a significant statistical correlation with the infectious disease. The fusion module is used to perform multi-source fusion processing on the screened environmental features of the prediction period and the search features to obtain multi-source fusion features for the infectious disease. The prediction module is used to perform trend prediction processing on the multi-source fusion characteristics of the infectious disease to obtain the infectious disease trend for the prediction period.

13. The apparatus according to claim 12, characterized in that, The feature extraction module is also used for: The keywords in the keyword database are matched with the keywords in the search data to obtain the search keywords for the prediction period; Based on the search data, the search keywords are statistically processed to obtain the search metrics for the search keywords; The search keywords are subjected to feature transformation processing to obtain the semantic features of the search keywords; The set of semantic features of the search keywords and the search indicators of the search keywords are used as the search features for the infectious disease during the prediction period.

14. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the artificial intelligence-based infectious disease trend prediction method according to any one of claims 1 to 11.

15. A computer-readable storage medium, characterized in that, It stores executable instructions for implementing the artificial intelligence-based infectious disease trend prediction method according to any one of claims 1 to 11 when executed by a processor.

16. A computer program product comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the artificial intelligence-based infectious disease trend prediction method according to any one of claims 1 to 11.

Images