Methods, devices, equipment, and products for infectious disease surveillance based on multi-source heterogeneous data

By collecting multi-source heterogeneous data through distributed web crawlers and combining hierarchical keyword libraries and autoregressive exogenous prediction models, the problems of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease surveillance have been solved, enabling efficient infectious disease surveillance and timely early warning.

CN123120364APending Publication Date: 2026-11-17SHENZHEN MSU-BIT UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611353559.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-02
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing infectious disease surveillance technologies suffer from immature multi-source fusion modeling and inaccurate feature extraction, making it difficult to achieve timely early warning.

Method used

Multi-source heterogeneous data is collected using distributed web crawler tools. Natural language processing features are extracted using a hierarchical keyword library and semantic filtering rules. Data fusion prediction is performed using a pre-built autoregressive exogenous prediction model, and an early warning notification is sent when the predicted value exceeds a safety threshold.

Benefits of technology

It has improved the efficiency of infectious disease surveillance, enabled timely early warning, and solved the problems of data lag and inaccurate feature extraction in traditional surveillance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120364A_ABST
    Figure CN123120364A_ABST
Patent Text Reader

Abstract

This application discloses a method, apparatus, device, and product for infectious disease monitoring based on multi-source heterogeneous data, relating to the field of health information management technology. The method includes: collecting data from an internet platform using a distributed web crawler to obtain a multi-source heterogeneous dataset; extracting features from the multi-source heterogeneous dataset using natural language processing based on a hierarchical keyword library and semantic filtering rules to obtain multi-source data features; fusing and predicting the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain an infectious disease activity trend prediction result; visualizing the infectious disease activity trend prediction result; and sending an early warning notification when the predicted value of the infectious disease activity trend prediction result exceeds a safety threshold. This improves the efficiency of infectious disease monitoring based on multi-source heterogeneous data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of health information management technology, and in particular to a method, device, equipment and product for infectious disease monitoring based on multi-source heterogeneous data. Background Technology

[0002] Surveillance and early warning of infectious diseases are the core of public health prevention and control, providing important support for disease intervention, resource allocation, and the formulation of prevention and control strategies. Current infectious disease surveillance technologies include two major systems: traditional passive surveillance and new digital surveillance. Traditional surveillance is centered on medical institutions and disease control institutions, relying on case reporting, laboratory testing, epidemiological investigation and analysis, etc., and completing the collection and release of epidemic data through the National Infectious Disease Surveillance Information System to achieve structured monitoring of routine epidemics. To make up for the shortcomings of traditional technologies, existing digital surveillance technologies have become a cutting-edge supplementary means. Relying on Internet big data information technology, based on multi-source data such as search engine data, public information on social media, drug sales, and population migration, combined with machine learning and data fusion algorithms, digital traces related to the epidemic are mined to achieve real-time judgment and tracking of infectious disease trends.

[0003] However, existing infectious disease surveillance technologies still have significant shortcomings. Traditional passive surveillance and reporting processes are cumbersome, with long data flow chains and an inherent lag of 1 to 3 weeks. Moreover, when facing emerging infectious diseases, the lack of clear diagnostic criteria and reporting categories in the early stages can easily lead to response delays. Existing digital surveillance technologies can only serve as supplementary means. Monitoring dimensions based on single data sources are one-sided, the maturity of multi-source fusion modeling solutions is insufficient, and the accuracy of data mining is limited. It is difficult to achieve comprehensive and high-precision forward-looking early warning of epidemics and cannot fully meet the core needs of rapid emergency response to emerging and re-emerging infectious diseases.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and product for infectious disease monitoring based on multi-source heterogeneous data, which aims to solve the technical problems of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which makes it difficult to achieve timely early warning.

[0006] To achieve the above objectives, this application proposes an infectious disease surveillance method based on multi-source heterogeneous data, the method comprising: Data was collected from internet platforms using distributed web crawler tools to obtain multi-source heterogeneous datasets; Based on the hierarchical keyword library and semantic filtering rules, feature extraction is performed on the multi-source heterogeneous dataset using natural language processing to obtain multi-source data features; The multi-source data features are fused and predicted using a pre-constructed autoregressive exogenous prediction model to obtain the prediction results of infectious disease activity trends. The infectious disease activity trend prediction results are visualized and output, and an early warning notification is sent when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold.

[0007] In one embodiment, before the step of fusing and predicting the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain the infectious disease activity trend prediction result, the method further includes: Obtain historical official monitoring data and perform logarithmic probability transformation on the historical official monitoring data to obtain the target variable; The optimal lag order of the historical official monitoring data and Internet data sources is determined by cross-validation, resulting in a set of lag orders. The historical multi-source data features are converted into exogenous input variables, and the historical autoregressive term of the target variable and the exogenous input variables are combined according to the set of lag orders to obtain the autoregressive exogenous prediction model formula. The parameters of the autoregressive exogenous prediction model formula are solved by the least squares method to obtain the autoregressive exogenous prediction model.

[0008] In one embodiment, the multi-source data features include self-media article proportion features, search index features, and social media proportion features. The step of extracting features from the multi-source heterogeneous dataset using natural language processing based on a hierarchical keyword library and semantic filtering rules to obtain multi-source data features includes: Collect core symptoms, disease names, and behavioral keywords related to the target infectious disease, and filter these core symptoms, disease names, and behavioral keywords using term frequency-inverse document frequency to obtain infectious disease keywords; The semantics of the infectious disease keywords were expanded using a word vector model to obtain expanded keywords; A hierarchical keyword library was constructed based on the expanded keywords and infectious disease keywords. The hierarchical keyword library is used to perform keyword matching on the self-media article data to obtain matching results. The matching results are then subjected to binary classification and filtering using a BERT pre-trained model to obtain the self-media article proportion feature. The hierarchical keyword library is used to combine keywords in the search engine query data to obtain combined keywords. The combined keywords are then weighted and summed according to the search frequency of the search engine query data to obtain a comprehensive search index feature. The social media data is classified into texts using the hierarchical keyword library to obtain classification results. The proportion of the classification results is then calculated to obtain the social media proportion feature.

[0009] In one embodiment, the step of fusing and predicting the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain the predicted trend of infectious disease activity includes: The initial predicted value is obtained by predicting the features of the multi-source data and the features of historical multi-source data using the autoregressive exogenous prediction model. The confidence interval of the initial predicted value is obtained by resampling the initial predicted value using a bootstrap sampling method. Based on the initial predicted value and the confidence interval, the predicted trend of infectious disease activity is determined.

[0010] In one embodiment, the step of visualizing the predicted infectious disease activity trend results includes: Based on the historical official monitoring data and the predicted trend of infectious disease activity, a comparison chart is generated using time series curves, and the comparison chart is differentiated by color to obtain a trend comparison chart. The confidence interval of the predicted trend of the infectious disease activity is displayed by shaded areas to obtain the error range; Based on the change curve of the predicted trend of infectious disease activity, a trend stage analysis is performed to obtain the trend analysis results. The trend comparison chart, error range, and trend analysis results are visualized and output.

[0011] In one embodiment, the step of sending an early warning notification when the predicted value of the infectious disease activity trend prediction result exceeds a safety threshold includes: Thresholds are divided based on historical safety averages to obtain a first safety threshold, a second safety threshold, and a third safety threshold. Duration conditions are then determined based on the first safety threshold, the second safety threshold, and the third safety threshold. The predicted value is used to make an alarm judgment based on the first security threshold, the second security threshold, and the third security threshold to obtain a judgment result. When the judgment result indicates that the predicted value exceeds the safety threshold and the duration condition is met, an early warning notification is generated and sent to the corresponding early warning processing terminal.

[0012] In one embodiment, after the steps of visualizing the infectious disease activity trend prediction results and sending an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold, the method further includes: New official monitoring data and internet data are collected to obtain an incremental dataset; The autoregressive exogenous prediction model is incrementally learned using the incremental dataset to obtain an updated autoregressive exogenous prediction model.

[0013] Furthermore, to achieve the above objectives, this application also proposes an infectious disease monitoring device based on multi-source heterogeneous data, the infectious disease monitoring device based on multi-source heterogeneous data comprising: The data collection module is used to collect data from Internet platforms using distributed web crawler tools to obtain multi-source heterogeneous datasets; The extraction module is used to extract features from the multi-source heterogeneous dataset through natural language processing based on the hierarchical keyword library and semantic filtering rules to obtain multi-source data features; The prediction module is used to fuse and predict the features of the multi-source data through a pre-built autoregressive exogenous prediction model to obtain the prediction results of infectious disease activity trends. The sending module is used to visualize the infectious disease activity trend prediction results and send an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold.

[0014] In addition, to achieve the above objectives, this application also proposes an infectious disease monitoring device based on multi-source heterogeneous data. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the infectious disease monitoring method based on multi-source heterogeneous data as described above.

[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the infectious disease monitoring method based on multi-source heterogeneous data as described above.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the infectious disease monitoring method based on multi-source heterogeneous data as described above.

[0017] One or more technical solutions proposed in this application have at least the following technical effects: This application proposes a method, apparatus, device, and product for infectious disease monitoring based on multi-source heterogeneous data. It collects data from an internet platform using a distributed web crawler to obtain a multi-source heterogeneous dataset. Based on a hierarchical keyword library and semantic filtering rules, it extracts features from the multi-source heterogeneous dataset using natural language processing to obtain multi-source data features. It then fuses and predicts the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain an infectious disease activity trend prediction result. The prediction result is then visualized and output, and an early warning notification is sent when the predicted value exceeds a safety threshold. Therefore, multi-source heterogeneous data is collected through distributed crawlers, and then NLP feature extraction is performed using a hierarchical keyword library and semantic filtering rules. Compared with traditional keyword matching, this significantly improves accuracy and solves the problem of inaccurate feature extraction. Then, a pre-built autoregressive exogenous prediction model is used to fuse multi-source features as exogenous variables with historical autoregressive terms to achieve mature multi-source data fusion. Finally, based on the visualization output of the prediction results and threshold warning, timely warning is achieved. This solves the problem that the multi-source fusion model is immature and the feature extraction is inaccurate in infectious disease monitoring technology, which makes it difficult to achieve timely warning. This improves the efficiency of infectious disease monitoring based on multi-source heterogeneous data. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating an embodiment of the infectious disease surveillance method based on multi-source heterogeneous data provided in this application. Figure 2 This is a flowchart illustrating Embodiment 2 of the infectious disease surveillance method based on multi-source heterogeneous data provided in this application; Figure 3 A simplified flowchart illustrating the infectious disease surveillance method based on multi-source heterogeneous data provided in Embodiment 2 of this application; Figure 4 This is a schematic diagram of the overall implementation architecture of the infectious disease surveillance method based on multi-source heterogeneous data in this application; Figure 5 This is a schematic diagram of the module structure of the infectious disease monitoring device based on multi-source heterogeneous data according to an embodiment of this application; Figure 6 This is a schematic diagram of the hardware operating environment of the infectious disease monitoring method based on multi-source heterogeneous data in the embodiments of this application.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0023] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0024] The main solution of this application embodiment is as follows: acquire historical official monitoring data, and perform log-probability transformation on the historical official monitoring data to obtain the target variable; determine the optimal lag order of the historical official monitoring data and the Internet data source through cross-validation to obtain a lag order set; convert the features of historical multi-source data into exogenous input variables, and combine the historical autoregressive term of the target variable and the exogenous input variables according to the lag order set to obtain the autoregressive exogenous prediction model formula; solve the parameters of the autoregressive exogenous prediction model formula using the least squares method to obtain the autoregressive exogenous prediction model. The process involves collecting core symptoms, disease names, and behavioral keywords related to a target infectious disease. These keywords are then filtered using word frequency-inverse document frequency (IF-IVF) to obtain infectious disease keywords. Semantic expansion of these keywords is performed using a word vector model to obtain expanded keywords. A hierarchical keyword library is constructed based on the expanded keywords and the infectious disease keywords. Keyword matching is performed on self-media article data using this hierarchical keyword library to obtain matching results. These matching results are then subjected to binary classification using a BERT pre-trained model to obtain the self-media article proportion feature. Keyword combination is performed on search engine query data using this hierarchical keyword library to obtain combined keywords. These combined keywords are then weighted and summed based on the search frequency of the search engine query data to obtain a comprehensive search index feature. Finally, text classification is performed on social media data using this hierarchical keyword library to obtain classification results. The proportion of these classification results is calculated to obtain the social media proportion feature. The autoregressive exogenous prediction model is used to predict the features of the multi-source data and historical multi-source data to obtain initial predicted values. These initial predicted values ​​are then resampled using a bootstrap sampling method to obtain confidence intervals. Based on the initial predicted values ​​and the confidence intervals, the predicted trend of infectious disease activity is determined. A comparison chart is generated using time series curves based on historical official monitoring data and the predicted trend of infectious disease activity, and the comparison chart is color-coded to obtain a trend comparison chart. The confidence intervals of the predicted trend of infectious disease activity are shaded to obtain the error range. Trend stage analysis is performed based on the change curves of the predicted trend of infectious disease activity to obtain trend analysis results. The trend comparison chart, error range, and trend analysis results are then visualized and output.Thresholds are defined based on historical safety averages to obtain a first safety threshold, a second safety threshold, and a third safety threshold. A duration condition is then determined based on these three thresholds. The predicted value is then used to make an alarm judgment based on these three thresholds, yielding a judgment result. When the judgment result indicates that the predicted value exceeds the safety threshold and meets the duration condition, an early warning notification is generated and sent to the corresponding early warning processing terminal. Newly added official monitoring data and internet data are collected to obtain an incremental dataset. This incremental dataset is then used to incrementally learn the autoregressive exogenous prediction model, resulting in an updated autoregressive exogenous prediction model. This addresses the problem of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which hinders timely early warning. It enables the monitoring of infectious diseases and improves monitoring efficiency. Based on the present invention, addressing the problems of cumbersome traditional passive monitoring and reporting processes, long data flow chains, inherent delays of 1 to 3 weeks, and low efficiency in the face of emerging infectious diseases due to the lack of clear diagnostic criteria and reporting categories in the early stages, a method for infectious disease monitoring based on multi-source heterogeneous data was designed. The effectiveness of the method was verified during infectious disease monitoring, and the efficiency of infectious disease monitoring based on multi-source heterogeneous data was significantly improved.

[0025] In this embodiment, for ease of description, the following description will focus on an infectious disease monitoring device based on multi-source heterogeneous data.

[0026] Due to limitations in existing data acquisition and fusion technologies, the efficiency of infectious disease surveillance based on multi-source heterogeneous data needs improvement. One issue is the single data source problem; relying solely on a single data source such as search engines or social media may lead to biased epidemic signals. The lack of effective integration of multiple available monitoring paths also reduces the accuracy of infectious disease surveillance. Another issue is data lag; traditional passive monitoring has a long data flow chain with an inherent lag of 1 to 3 weeks. Without considering the timeliness of real-time internet data, the timeliness of infectious disease surveillance will also decrease significantly. Furthermore, there is the issue of data noise. Accurate infectious disease surveillance requires extracting effective signals from massive amounts of internet information, but social media and other channels contain a large amount of irrelevant and false information, resulting in high data noise, which also affects the efficiency of infectious disease surveillance. Therefore, current public health surveillance systems also face challenges in domain adaptability. Since different infectious diseases (such as influenza and dengue fever) have different transmission characteristics and keywords, if the surveillance model does not undergo disease-specific adaptive learning and training, its efficiency will also decrease.

[0027] This application provides a solution that collects multi-source heterogeneous data through distributed crawlers, then uses a hierarchical keyword library and semantic filtering rules for NLP feature extraction, significantly improving accuracy compared to traditional keyword matching and solving the problem of inaccurate feature extraction. Next, a pre-built autoregressive exogenous prediction model is used to fuse multi-source features as exogenous variables with historical autoregressive terms, achieving mature multi-source data fusion. Finally, based on the visualization output of the prediction results and threshold-based early warning, timely alerts are achieved. This solves the problem of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which makes timely early warning difficult, thus improving the efficiency of infectious disease monitoring based on multi-source heterogeneous data.

[0028] Based on this, embodiments of this application provide a method for infectious disease surveillance based on multi-source heterogeneous data, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the infectious disease monitoring method based on multi-source heterogeneous data in this application.

[0029] In this embodiment, the infectious disease monitoring method based on multi-source heterogeneous data includes steps S01 to S04: Step S01: Collect data from the Internet platform using a distributed web crawler tool to obtain a multi-source heterogeneous dataset; Before the implementation of this embodiment, it should be clear that existing infectious disease surveillance technologies still have significant shortcomings. Traditional passive monitoring and reporting processes are cumbersome, data flow chains are long, and there is an inherent lag of 1 to 3 weeks. Moreover, when facing emerging infectious diseases, the lack of clear diagnostic criteria and reporting categories in the early stages can easily lead to response delays. Existing digital monitoring technologies can only serve as auxiliary and supplementary means. Single data source monitoring dimensions are one-sided, the maturity of multi-source fusion modeling schemes is insufficient, and the accuracy of data mining is limited. It is difficult to achieve comprehensive and high-precision forward-looking early warning of epidemics and cannot fully meet the core needs of rapid emergency response to emerging and re-emerging infectious diseases.

[0030] Therefore, in order to solve the above problems, this embodiment uses a distributed web crawler system deployed on a cloud server to collect three types of heterogeneous data sources from the Internet platform in real time, including self-media article data (such as WeChat official accounts, Toutiao, Sohu, Baijiahao, Zhihu columns, etc.), search engine query data (such as Baidu Index, Google Trends, etc.), and social media data (such as Weibo, Twitter, etc.). The collection frequency is set to once a day to ensure the timeliness of the data.

[0031] For self-media article data, the collected content includes article title, body content, publication time, publishing platform, author information, number of views, number of likes, etc. For search engine query data, the daily search frequency, search trends, and geographical distribution of keywords related to the target infectious disease are obtained through official APIs or simulated requests. For social media data, the body content, publication time, publishing user, geographical location, number of reposts, number of comments, number of likes, etc. of posts containing keywords related to the target infectious disease are collected through API interfaces or web crawlers.

[0032] After collection, the three types of data are aggregated into a multi-source heterogeneous dataset and stored in a distributed database (such as Elasticsearch). Indexes are created according to dimensions such as time, platform, and region to facilitate subsequent retrieval and analysis.

[0033] Step S02: Based on the hierarchical keyword library and semantic filtering rules, feature extraction is performed on the multi-source heterogeneous dataset using natural language processing to obtain multi-source data features; Subsequently, natural language processing techniques were employed, combined with a pre-defined hierarchical keyword database related to infectious diseases and semantic filtering rules, to filter information entries related to the target prevalent infectious disease from the preprocessed multi-source heterogeneous dataset. Differentiated feature extraction methods were designed for different data sources. Self-media articles are classified using a combination of keyword matching and BERT pre-trained model binary classification. Social media data is classified using a BERT-based deep learning text classification model. Search engine data is classified using keyword combination search frequency statistics and time-series feature extraction methods.

[0034] Finally, weekly or daily characteristic indicators for each data source are calculated, including the proportion of self-media articles, the comprehensive search index, and the proportion of social media-related posts, which together constitute the multi-source data characteristics.

[0035] Step S03: The multi-source data features are fused and predicted using a pre-constructed autoregressive exogenous prediction model to obtain the prediction result of infectious disease activity trend. Next, using historical official monitoring data (such as the proportion of influenza-like cases) as the prediction target, and taking the extracted multi-source data features as exogenous input variables, a hybrid prediction model integrating an autoregressive exogenous model and an attention mechanism is constructed.

[0036] By using an adaptive weight learning mechanism, the contribution of different data sources in the prediction is dynamically adjusted to achieve real-time prediction of the trend of infectious disease activity. The model in this embodiment is an autoregressive exogenous prediction model, which uses real-time data and historical data from each data source as exogenous input variables. The parameters are estimated by the least squares method, and the optimal lag order of each data source is determined by cross-validation. Finally, the prediction result of the infectious disease activity trend is output.

[0037] Step S04: Visualize the infectious disease activity trend prediction results and send an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds the safety threshold.

[0038] Finally, the model prediction results are displayed through a visualization interface, including time series curves (showing the comparison between historical official monitoring data and model predictions, with different colors to distinguish between measured and predicted values), prediction error range (the 95% confidence interval of the predicted value is calculated using the Bootstrap method and displayed as a shaded area), and trend analysis (based on the continuous changes in the prediction results, it is determined whether the epidemic is in an upward, peak, or downward phase).

[0039] Meanwhile, a three-tiered early warning mechanism is established. When the infectious disease activity level predicted by the model exceeds the preset threshold, the system automatically triggers a tiered early warning notification, which is sent to the designated contact person through various means such as email, SMS, and system pop-ups.

[0040] Specifically, before step S03 above, which involves fusing and predicting the multi-source data features using a pre-constructed autoregressive exogenous prediction model to obtain the prediction result of infectious disease activity trends, the method further includes: Step S0301: Obtain historical official monitoring data and perform logarithmic probability transformation on the historical official monitoring data to obtain the target variable; Step S0302: Determine the optimal lag order of the historical official monitoring data and the Internet data source through cross-validation to obtain the lag order set; Step S0303: Convert the historical multi-source data features into exogenous input variables, and combine the historical autoregressive term of the target variable and the exogenous input variables according to the set of lag orders to obtain the autoregressive exogenous prediction model formula. Step S0304: Solve the parameters of the autoregressive exogenous prediction model formula using the least squares method to obtain the autoregressive exogenous prediction model.

[0041] First, obtain historical weekly epidemic infectious disease surveillance data from official health institutions (such as the Chinese Center for Disease Control and Prevention), such as the percentage of influenza-like illness cases. ,right Perform a logit transformation to obtain the target variable. :

[0042] This transformation is used to map proportional data with values ​​ranging from 0 to 1 to the real number domain, making it more consistent with the assumptions of the linear model, stabilizing the variance, and improving the model fit.

[0043] Subsequently, cross-validation was used to determine the optimal lag order for each data source. For a single internet data source, simplified models (such as models containing only features of self-media articles) were constructed, and the lag orders were iterated through. We select the lag order that minimizes the root mean square error (RMSE) of the validation set, and use the same method to determine the optimal lag order for social media features. Optimal lag order for search engine features For official historical data, the lag order traversal By selecting the order that minimizes the prediction error of the base model, the set of lag orders is finally obtained. .

[0044] Multi-source data characteristics at historical moments (characteristics of the proportion of self-media articles) Comprehensive search index characteristics Social media proportion characteristics The historical autoregressive term of the target variable is converted into an exogenous input variable, and based on the lag order set determined in the above steps, the historical autoregressive term of the target variable is combined with the exogenous input variable to construct the autoregressive exogenous prediction model formula:

[0045] in, , , , Here are the parameters to be estimated in the model, and e is a constant term. It is a random error term, and .

[0046] Furthermore, the autoregressive exogenous prediction model in this embodiment also incorporates an attention mechanism, dynamically adjusting the proportion of self-media articles through adaptive weight learning. Comprehensive search index characteristics Social media proportion characteristics In terms of contribution to prediction, specifically, the three types of exogenous input variables are weighted by learnable attention weights to obtain a fused exogenous input vector, which is then substituted into the above autoregressive exogenous prediction model formula for prediction.

[0047] Attention weights and model parameters , , , Together with the constant term e, they are jointly estimated using the least squares method. This allows the attention weight of a data source to adaptively increase when it contains a stronger epidemic signal and to adaptively decrease when it has high noise or a weak signal. This suppresses noise interference and avoids bias caused by fixed weights.

[0048] Model parameters are estimated using the least squares method. , , , The parameters are calculated to minimize the sum of squared residuals. Once the parameters are solved, a usable autoregressive exogenous prediction model is obtained. Meanwhile, the variance inflation factor (VIF) is used to assess multicollinearity among the independent variables. A VIF value less than 4 indicates that the degree of collinearity is low and acceptable, while a VIF value greater than 10 indicates that there is severe collinearity and variable screening or regularization methods need to be considered.

[0049] The model performance is evaluated using repeated K-fold cross-validation. The data is divided into K folds (e.g., K=10), and the cross-validation process is repeated multiple times (e.g., 10 times). The average prediction error index is calculated, including root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), correlation between predicted and actual values, and incremental correlation.

[0050] More specifically, the aforementioned multi-source data features include self-media article proportion features, search index features, and social media proportion features. Step S02, which involves extracting features from the multi-source heterogeneous dataset using natural language processing based on a hierarchical keyword library and semantic filtering rules to obtain multi-source data features, includes: Step S021: Collect core symptoms, disease names, and behavioral keywords related to the target infectious disease, and filter the core symptoms, disease names, and behavioral keywords by word frequency-inverse document frequency to obtain infectious disease keywords; Step S022: Semantically expand the infectious disease keywords using a word vector model to obtain expanded keywords; Step S023: Construct a hierarchical keyword library based on the expanded keywords and infectious disease keywords; Step S024: Use the hierarchical keyword library to perform keyword matching on the self-media article data to obtain matching results, and use the BERT pre-trained model to perform binary classification and filtering on the matching results to obtain the self-media article proportion feature. Step S025: Combine keywords in the search engine query data using the hierarchical keyword library to obtain combined keywords, and perform a weighted summation of the combined keywords based on the search frequency of the search engine query data to obtain a comprehensive search index feature; Step S026: Classify the social media data into texts using the hierarchical keyword library to obtain classification results, and calculate the proportion of the classification results to obtain the social media proportion feature.

[0051] First, collect initial keywords related to the target infectious disease, divided into three levels: primary keywords (core symptoms) such as "fever", "cough", "fatigue", "headache", etc.; secondary keywords (disease names) such as "influenza" and "dengue fever", etc.; and tertiary keywords (related behaviors) such as "vaccine", "isolation", "testing", "hospital", etc.

[0052] A batch of articles was randomly selected from historical data, and the importance of each word was calculated using the term frequency-inverse document frequency (TF-IDF) algorithm. High-frequency and representative words were selected to obtain infectious disease keywords.

[0053] Subsequently, the Word2Vec word vector model was used to calculate words that are semantically similar to infectious disease keywords. For example, "fever" can be expanded to "high fever" or "low fever", and "cough" can be expanded to "dry cough" or "sputum". This expands the keyword library and yields extended keywords.

[0054] The infectious disease keywords obtained from the above steps are merged with the expanded keywords, and epidemiological experts are invited to review the candidate keywords, eliminating irrelevant or ambiguous words, and finally a multi-level keyword library is constructed.

[0055] Keyword matching is performed on the preprocessed self-media article titles and texts. If the article contains at least one word from the keyword library, it is marked as an "infectious disease related article". To further improve accuracy, a BERT pre-trained model is used to perform binary classification (relevant / irrelevant) on the marked articles. The model input is the article text and the output is the disease relevance probability. Articles with a probability greater than 0.7 are selected as the final relevant articles.

[0056] Count the number of infectious disease-related articles on social media each week (or day). and the total number of articles collected in the same period Calculate the characteristics of the proportion of self-media articles :

[0057] Logarithmic transformation is used to stabilize variance, making the data more consistent with the normal distribution assumption.

[0058] For the target infectious disease, multiple keyword combinations are constructed (such as "flu symptoms", "flu treatment", "flu vaccine"), and the search frequency of each is extracted. Let the search frequency of the j-th keyword in week t be . Calculate the overall search index:

[0059] Where M is the total number of keywords, The weight of the j-th keyword (determined through relevance analysis).

[0060] A BERT pre-trained model is used to classify preprocessed social media posts to determine whether they are related to the target infectious disease. During model fine-tuning, a labeled dataset of disease-related posts is used. Let the number of posts classified as related in week t be . The total number of posts collected during the same period was Calculate the social media share characteristics:

[0061] Furthermore, step S03 above, which involves fusing and predicting the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain the predicted trend of infectious disease activity, includes: Step S031: The multi-source data features and historical multi-source data features are predicted using the autoregressive exogenous prediction model to obtain initial prediction values; Step S032: Resample the initial predicted value using a bootstrap sampling method to obtain the confidence interval of the initial predicted value; Step S033: Determine the prediction result of infectious disease activity trend based on the initial predicted value and the confidence interval.

[0062] The multi-source data features of the current and historical times (including the proportion of self-media articles) Comprehensive search index characteristics Social media proportion characteristics Input the completed autoregressive exogenous prediction model. The model is based on the formula:

[0063] Perform calculations and output the predicted value of the target variable at the current time. Then, the initial predicted value is obtained through the inverse logit transformation.

[0064] The Bootstrap method (resampled 1000 times) was then used to resample the initial predicted values. Each time, samples with replacement were drawn from the original data with the same sample size as the original samples. The model was refitted and the predicted values ​​were calculated, resulting in 1000 predicted value samples. These 1000 predicted value samples were sorted from smallest to largest, and the 2.5 percentile and 97.5 percentile were taken as the lower and upper limits of the 95% confidence interval, respectively, to obtain the confidence interval of the initial predicted values.

[0065] Finally, the initial predicted values ​​are combined with confidence intervals to determine the final prediction result of infectious disease activity trends. This result includes point predicted values ​​(initial predicted values) and interval predicted values ​​(confidence intervals), which are used for subsequent visualization output and early warning judgment.

[0066] This embodiment, through the above-described scheme, specifically uses a distributed web crawler tool to collect data from an internet platform, obtaining a multi-source heterogeneous dataset; based on a hierarchical keyword library and semantic filtering rules, it extracts features from the multi-source heterogeneous dataset using natural language processing to obtain multi-source data features; it then uses a pre-built autoregressive exogenous prediction model to fuse and predict the multi-source data features, obtaining a predicted trend of infectious disease activity; the predicted trend of infectious disease activity is visualized and output, and a warning notification is sent when the predicted value of the predicted trend of infectious disease activity exceeds a safety threshold. Therefore, multi-source heterogeneous data is collected through distributed crawlers, and then NLP feature extraction is performed using a hierarchical keyword library and semantic filtering rules. Compared with traditional keyword matching, this significantly improves accuracy and solves the problem of inaccurate feature extraction. Then, a pre-built autoregressive exogenous prediction model is used to fuse multi-source features as exogenous variables with historical autoregressive terms to achieve mature multi-source data fusion. Finally, based on the visualization output of the prediction results and threshold warning, timely warning is achieved. This solves the problem that the multi-source fusion model is immature and the feature extraction is inaccurate in infectious disease monitoring technology, which makes it difficult to achieve timely warning. This improves the efficiency of infectious disease monitoring based on multi-source heterogeneous data.

[0067] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 In step S04, the step of visualizing the predicted trend of infectious disease activity, the infectious disease monitoring method based on multi-source heterogeneous data further includes steps S041-S044: Step S041: Based on the historical official monitoring data and the predicted trend of infectious disease activity, a comparison chart is generated through time series curves, and the comparison chart is distinguished by color to obtain a trend comparison chart; Step S042: The confidence interval of the predicted trend of the infectious disease activity is displayed by shaded areas to obtain the error range; Step S043: Perform trend stage analysis based on the change curve of the infectious disease activity trend prediction result to obtain the trend analysis result; Step S044: Visualize the trend comparison chart, error range, and trend analysis results.

[0068] Historical official monitoring data (such as the actual percentage of influenza-like illness cases) and the infectious disease activity trend prediction results (including point predictions and interval predictions) output by the model are plotted on the same time series curve. The horizontal axis represents time (weekly or daily), and the vertical axis represents the level of infectious disease activity. The blue solid line represents the measured value of historical official monitoring data, the red solid line represents the point prediction value of the model, and the light red shaded area represents the 95% confidence interval of the prediction value, thus obtaining a trend comparison chart.

[0069] Based on the trend comparison chart, the confidence interval of the predicted value is specially displayed in the form of a shaded area. The 95% confidence interval calculated by the Bootstrap method is covered by a semi-transparent shaded area on the time series curve, which intuitively reflects the uncertainty range of the prediction result and obtains the error range visualization result.

[0070] Based on the continuous changes in the prediction results, the system automatically determines the stage of the epidemic. If the epidemic increases for two consecutive weeks and the growth rate is greater than 5%, it is determined to be in the rising phase. If the epidemic remains at a high level for three consecutive weeks and the fluctuation is less than 10%, it is determined to be in the peak phase. If the epidemic declines for two consecutive weeks and the decline rate is greater than 5%, it is determined to be in the declining phase. The trend stage analysis results are marked with text labels in the corresponding positions on the trend comparison chart to obtain the trend analysis results.

[0071] The generated trend comparison charts, error range visualizations, and trend analysis results are integrated and output uniformly through a visualization interface. Output formats include web page display, PDF report export, or large-screen monitoring panel, for reference in public health decision-making.

[0072] Specifically, step S04 above, the step of sending an early warning notification when the predicted value of the infectious disease activity trend prediction result exceeds the safety threshold, includes: Step S045: Divide the thresholds according to the historical safety average to obtain the first safety threshold, the second safety threshold and the third safety threshold, and determine the duration condition according to the first safety threshold, the second safety threshold and the third safety threshold. Step S046: The predicted value is used to make an alarm judgment based on the first security threshold, the second security threshold, and the third security threshold to obtain a judgment result; Step S047: When the judgment result is that the predicted value exceeds the safety threshold and the duration condition is met, an early warning notification is generated and the early warning notification is sent to the corresponding early warning processing terminal.

[0073] Thresholds are defined based on historical safety averages (i.e., the average level of infectious disease activity during the same historical period): the first safety threshold is set at 1.5 times the historical safety average, the second safety threshold is set at 2 times the historical safety average, and the third safety threshold is set at 3 times the historical safety average. At the same time, corresponding duration conditions are determined for each safety threshold: the first safety threshold requires the lower limit of the confidence interval of the predicted value to be higher than the historical safety average, the second safety threshold requires the predicted value to maintain an upward trend for two consecutive weeks, and the third safety threshold requires the predicted value to exceed the historical peak value.

[0074] The determined infectious disease activity trend prediction results (including point prediction values ​​and confidence intervals) are compared with the first safety threshold, the second safety threshold, and the third safety threshold, respectively.

[0075] If the predicted value exceeds the first safety threshold and the lower limit of the confidence interval is higher than the historical safety average, a Level 1 warning (yellow) is triggered. If the predicted value exceeds the second safety threshold and continues to rise for two consecutive weeks, a Level 2 warning (orange) is triggered. If the predicted value exceeds the third safety threshold or exceeds the historical peak, a Level 3 warning (red) is triggered. The judgment result includes the warning level and the triggering conditions.

[0076] When the judgment result meets the triggering conditions of the corresponding level, the system automatically generates an early warning notification. The early warning notification includes detailed information such as the early warning level (Level 1 / Level 2 / Level 3), early warning time, predicted value, safety threshold, confidence interval, and trend analysis results. The early warning notification is sent to designated contacts (such as staff of the CDC, public health decision-makers, etc.) through various means such as email, SMS, and system pop-ups, so that corresponding prevention and control measures can be taken in a timely manner.

[0077] More specifically, after step S04 above, which involves visualizing the infectious disease activity trend prediction results and sending an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold, the method further includes: Step S05: Collect newly added official monitoring data and Internet data to obtain an incremental dataset; Step S06: Incrementally learn the autoregressive exogenous prediction model using the incremental dataset to obtain the updated autoregressive exogenous prediction model.

[0078] Over time, new official surveillance data (such as the latest influenza-like illness percentage data) and internet data (including new self-media articles, search engine query data, and social media data) are collected periodically (e.g., quarterly or monthly) and aggregated into an incremental dataset.

[0079] An incremental learning strategy is adopted. While retaining the original model parameters, the obtained incremental dataset is used to fine-tune the autoregressive exogenous prediction model to adapt to changes in the epidemic and data distribution drift, and to maintain the model's prediction accuracy. In this embodiment, the new data is merged with the original training data and the parameters are re-estimated, but the original parameters are used as the initial values ​​for iterative optimization to reduce the computational overhead of full retraining. The updated model will be used for subsequent fusion prediction to form a closed loop of continuous optimization.

[0080] This embodiment, through the above-described scheme, specifically generates a comparison chart using time series curves based on historical official monitoring data and the predicted trend of infectious disease activity. The comparison chart is then color-coded to obtain a trend comparison chart. The confidence interval of the predicted trend of infectious disease activity is shaded to obtain the error range. Trend stage analysis is performed based on the change curve of the predicted trend of infectious disease activity to obtain trend analysis results. The trend comparison chart, error range, and trend analysis results are then visualized. Thus, multi-source heterogeneous data is collected through distributed crawling. Subsequently, NLP feature extraction is performed using a hierarchical keyword library and semantic filtering rules, significantly improving accuracy compared to traditional keyword matching and solving the problem of inaccurate feature extraction. Then, a pre-built autoregressive exogenous prediction model is used to fuse multi-source features as exogenous variables with historical autoregressive terms, achieving mature multi-source data fusion. Finally, based on the visualized output of the prediction results and threshold warnings, timely warnings are achieved. This solves the problem of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which makes timely warnings difficult to achieve, thus improving the efficiency of infectious disease monitoring based on multi-source heterogeneous data.

[0081] For example, to help understand the implementation process of the infectious disease monitoring method based on multi-source heterogeneous data obtained by combining this embodiment with the above embodiment one, please refer to... Figure 3 , Figure 3 A simplified flowchart of an infectious disease surveillance method based on multi-source heterogeneous data is provided, specifically: First, a multi-source heterogeneous corpus is constructed by collecting multi-source heterogeneous data from internet platforms in real time using distributed web crawling tools, including self-media article data, search engine query data, and social media data.

[0082] Then, differentiated preprocessing strategies were designed based on the characteristics of different data sources. Self-media articles were cleaned and structured for language standardization, social media data were subjected to stronger noise filtering and semantic disambiguation strategies, and search engine data were focused on extracting the fluctuation features of query word frequency.

[0083] Next, natural language processing technology is used in conjunction with a pre-set hierarchical keyword database related to infectious diseases and semantic filtering rules to extract feature indicators related to the target infectious disease from the pre-processed multi-source data, including the proportion of self-media articles, the comprehensive search index, and the proportion of social media.

[0084] Subsequently, based on historical official monitoring data, the extracted multi-source data features were used as exogenous input variables to construct an autoregressive exogenous prediction model. The parameters were estimated by the least squares method and the optimal lag order was determined by cross-validation to make real-time predictions of the activity trend of infectious diseases.

[0085] Finally, the prediction results are output through a visualization interface, including time series curves, confidence interval shading, and trend stage analysis. When the predicted value exceeds the safety threshold, a tiered early warning notification is automatically triggered and sent to a designated contact via email, SMS, or system pop-up. At the same time, the system regularly collects new data to incrementally learn and update the model, forming a closed loop of continuous optimization.

[0086] The overall implementation architecture of this embodiment is as follows: Figure 4 As shown, the architecture is divided into four layers from bottom to top: data acquisition layer, data processing layer, model prediction layer, and application output layer.

[0087] The data acquisition layer collects data from self-media articles, search engine queries, and social media in real time from internet platforms through a distributed web crawler system, while also receiving historical monitoring data from official health agencies.

[0088] The data processing layer implements differentiated preprocessing strategies for different data sources, including language standardization and cleaning of self-media articles, strong noise filtering and semantic disambiguation of social media data, missing value interpolation and trend smoothing of search engine data, and uses a hierarchical keyword library combined with a BERT pre-trained model for feature extraction to obtain multi-source data features.

[0089] The model prediction layer uses historical official monitoring data as the target variable and multi-source data features as exogenous input variables to construct an autoregressive exogenous prediction model. The parameters are solved by the least squares method and the optimal lag order is determined by cross-validation. The model outputs the prediction results and confidence intervals of infectious disease activity trends.

[0090] The application output layer displays the prediction results through a visualization interface, including time series comparison charts, error range shading, and trend stage analysis. It also automatically triggers tiered warning notifications when the predicted value exceeds the safety threshold. In addition, it supports online model updates, regularly collecting new data to incrementally learn and fine-tune the model to maintain prediction accuracy.

[0091] The entire architecture automates the entire process from data collection, processing, modeling to early warning output, providing timely and accurate digital support for public health decision-making.

[0092] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the infectious disease monitoring method based on multi-source heterogeneous data in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0093] This application also provides an infectious disease monitoring device based on multi-source heterogeneous data. Please refer to [link / reference]. Figure 5 The infectious disease monitoring device based on multi-source heterogeneous data includes: The data collection module 10 is used to collect data from the Internet platform using a distributed web crawler tool to obtain a multi-source heterogeneous dataset. Extraction module 20 is used to extract features from the multi-source heterogeneous dataset through natural language processing based on the hierarchical keyword library and semantic filtering rules to obtain multi-source data features; Prediction module 30 is used to fuse and predict the features of the multi-source data through a pre-built autoregressive exogenous prediction model to obtain the prediction result of infectious disease activity trend. The sending module 40 is used to visualize the infectious disease activity trend prediction results and send an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold.

[0094] The infectious disease monitoring device based on multi-source heterogeneous data provided in this application, employing the infectious disease monitoring method based on multi-source heterogeneous data in the above embodiments, can solve the technical problems of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which makes it difficult to achieve timely early warning. Compared with the prior art, the beneficial effects of the infectious disease monitoring device based on multi-source heterogeneous data provided in this application are the same as those of the infectious disease monitoring method based on multi-source heterogeneous data provided in the above embodiments, and other technical features in the infectious disease monitoring device based on multi-source heterogeneous data are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0095] This application provides an infectious disease monitoring device based on multi-source heterogeneous data. The infectious disease monitoring device based on multi-source heterogeneous data includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the infectious disease monitoring method based on multi-source heterogeneous data in the above embodiment 1.

[0096] The following is for reference. Figure 6This document illustrates a structural schematic diagram of an infectious disease monitoring device based on multi-source heterogeneous data, suitable for implementing embodiments of this application. The infectious disease monitoring device based on multi-source heterogeneous data in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and vehicle-mounted terminals (e.g., vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 6 The infectious disease monitoring device based on multi-source heterogeneous data shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0097] like Figure 6 As shown, an infectious disease monitoring device based on multi-source heterogeneous data may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the infectious disease monitoring device based on multi-source heterogeneous data. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the infectious disease surveillance device based on multi-source heterogeneous data to communicate wirelessly or wiredly with other devices to exchange data. While the figure shows infectious disease surveillance devices based on multi-source heterogeneous data with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0098] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0099] The infectious disease monitoring device based on multi-source heterogeneous data provided in this application, employing the infectious disease monitoring method based on multi-source heterogeneous data in the above embodiments, can solve the technical problems of immature multi-source fusion modeling and inaccurate feature extraction in infectious disease monitoring technology, which makes it difficult to achieve timely early warning. Compared with the prior art, the beneficial effects of the infectious disease monitoring device based on multi-source heterogeneous data provided in this application are the same as those of the infectious disease monitoring method based on multi-source heterogeneous data provided in the above embodiments, and other technical features in this infectious disease monitoring device based on multi-source heterogeneous data are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0100] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0102] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the infectious disease monitoring method based on multi-source heterogeneous data in the above embodiments.

[0103] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0104] The aforementioned computer-readable storage medium may be included in an infectious disease surveillance device based on multi-source heterogeneous data; or it may exist independently and not be assembled into an infectious disease surveillance device based on multi-source heterogeneous data.

[0105] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by an infectious disease monitoring device based on multi-source heterogeneous data, the device performs the following actions: collects data from an internet platform using a distributed web crawler to obtain a multi-source heterogeneous dataset; extracts features from the multi-source heterogeneous dataset using natural language processing based on a hierarchical keyword library and semantic filtering rules to obtain multi-source data features; fuses and predicts the multi-source data features using a pre-built autoregressive exogenous prediction model to obtain an infectious disease activity trend prediction result; visualizes the infectious disease activity trend prediction result; and sends an early warning notification when the predicted value of the infectious disease activity trend prediction result exceeds a safety threshold.

[0106] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0107] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0108] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0109] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described infectious disease monitoring method based on multi-source heterogeneous data. This addresses the technical problem in infectious disease monitoring technology where immature multi-source fusion modeling and inaccurate feature extraction lead to difficulties in achieving timely early warning. Compared with existing technologies, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the infectious disease monitoring method based on multi-source heterogeneous data provided in the above embodiments, and will not be elaborated upon here.

[0110] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the infectious disease monitoring method based on multi-source heterogeneous data as described above.

[0111] The computer program product provided in this application can solve the technical problem that infectious disease surveillance technology suffers from immature multi-source fusion modeling and inaccurate feature extraction, making it difficult to achieve timely early warning. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the infectious disease surveillance method based on multi-source heterogeneous data provided in the above embodiments, and will not be repeated here.

[0112] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for infectious disease surveillance based on multi-source heterogeneous data, characterized in that, The infectious disease surveillance method based on multi-source heterogeneous data includes: Data was collected from internet platforms using distributed web crawler tools to obtain multi-source heterogeneous datasets; Based on the hierarchical keyword library and semantic filtering rules, feature extraction is performed on the multi-source heterogeneous dataset using natural language processing to obtain multi-source data features; The multi-source data features are fused and predicted using a pre-constructed autoregressive exogenous prediction model to obtain the prediction results of infectious disease activity trends. The infectious disease activity trend prediction results are visualized and output, and an early warning notification is sent when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold.

2. The infectious disease monitoring method based on multi-source heterogeneous data as described in claim 1, characterized in that, Before the step of fusing and predicting the multi-source data features using a pre-constructed autoregressive exogenous prediction model to obtain the infectious disease activity trend prediction result, the method further includes: Obtain historical official monitoring data and perform logarithmic probability transformation on the historical official monitoring data to obtain the target variable; The optimal lag order of the historical official monitoring data and Internet data sources is determined by cross-validation, resulting in a set of lag orders. The historical multi-source data features are converted into exogenous input variables, and the historical autoregressive term of the target variable and the exogenous input variables are combined according to the set of lag orders to obtain the autoregressive exogenous prediction model formula. The parameters of the autoregressive exogenous prediction model formula are solved by the least squares method to obtain the autoregressive exogenous prediction model.

3. The infectious disease monitoring method based on multi-source heterogeneous data as described in claim 1, characterized in that, The multi-source data features include self-media article proportion features, search index features, and social media proportion features. The step of extracting features from the multi-source heterogeneous dataset using natural language processing based on a hierarchical keyword library and semantic filtering rules to obtain multi-source data features includes: Collect core symptoms, disease names, and behavioral keywords related to the target infectious disease, and filter these core symptoms, disease names, and behavioral keywords using term frequency-inverse document frequency to obtain infectious disease keywords; The semantics of the infectious disease keywords were expanded using a word vector model to obtain expanded keywords; A hierarchical keyword library was constructed based on the expanded keywords and infectious disease keywords. The hierarchical keyword library is used to perform keyword matching on the self-media article data to obtain matching results. The matching results are then subjected to binary classification and filtering using a BERT pre-trained model to obtain the self-media article proportion feature. The hierarchical keyword library is used to combine keywords in the search engine query data to obtain combined keywords. The combined keywords are then weighted and summed according to the search frequency of the search engine query data to obtain a comprehensive search index feature. The social media data is classified into texts using the hierarchical keyword library to obtain classification results. The proportion of the classification results is then calculated to obtain the social media proportion feature.

4. The infectious disease monitoring method based on multi-source heterogeneous data as described in claim 2, characterized in that, The step of fusing and predicting the multi-source data features using a pre-constructed autoregressive exogenous prediction model to obtain the predicted trend of infectious disease activity includes: The initial predicted value is obtained by predicting the features of the multi-source data and the features of historical multi-source data using the autoregressive exogenous prediction model. The confidence interval of the initial predicted value is obtained by resampling the initial predicted value using a bootstrap sampling method. Based on the initial predicted value and the confidence interval, the predicted trend of infectious disease activity is determined.

5. The infectious disease monitoring method based on multi-source heterogeneous data as described in claim 2, characterized in that, The step of visualizing the results of the infectious disease activity trend prediction includes: Based on the historical official monitoring data and the predicted trend of infectious disease activity, a comparison chart is generated using time series curves, and the comparison chart is differentiated by color to obtain a trend comparison chart. The confidence interval of the predicted trend of the infectious disease activity is displayed by shaded areas to obtain the error range; Based on the change curve of the predicted trend of infectious disease activity, a trend stage analysis is performed to obtain the trend analysis results. The trend comparison chart, error range, and trend analysis results are visualized and output.

6. The infectious disease monitoring method based on multi-source heterogeneous data as described in claim 1, characterized in that, The step of sending an early warning notification when the predicted value of the infectious disease activity trend prediction result exceeds a safety threshold includes: Thresholds are divided based on historical safety averages to obtain a first safety threshold, a second safety threshold, and a third safety threshold. Duration conditions are then determined based on the first safety threshold, the second safety threshold, and the third safety threshold. The predicted value is used to make an alarm judgment based on the first security threshold, the second security threshold, and the third security threshold to obtain a judgment result. When the judgment result indicates that the predicted value exceeds the safety threshold and the duration condition is met, an early warning notification is generated and sent to the corresponding early warning processing terminal.

7. The infectious disease surveillance method based on multi-source heterogeneous data as described in claim 1, characterized in that, After the steps of visualizing the infectious disease activity trend prediction results and sending an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold, the method further includes: New official monitoring data and internet data are collected to obtain an incremental dataset; The autoregressive exogenous prediction model is incrementally learned using the incremental dataset to obtain an updated autoregressive exogenous prediction model.

8. An infectious disease monitoring device based on multi-source heterogeneous data, characterized in that, The infectious disease monitoring device based on multi-source heterogeneous data includes: The data collection module is used to collect data from Internet platforms using distributed web crawler tools to obtain multi-source heterogeneous datasets; The extraction module is used to extract features from the multi-source heterogeneous dataset through natural language processing based on the hierarchical keyword library and semantic filtering rules to obtain multi-source data features; The prediction module is used to fuse and predict the features of the multi-source data through a pre-built autoregressive exogenous prediction model to obtain the prediction results of infectious disease activity trends. The sending module is used to visualize the infectious disease activity trend prediction results and send an early warning notification when the predicted value of the infectious disease activity trend prediction results exceeds a safety threshold.

9. An infectious disease monitoring device based on multi-source heterogeneous data, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the infectious disease surveillance method based on multi-source heterogeneous data as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the infectious disease monitoring method based on multi-source heterogeneous data as described in any one of claims 1 to 7.