Emergency public health event early warning method and system based on big data

Abstract

The invention discloses an emergency public health event early warning method and system based on big data, and relates to the technical field of public health. The method comprises the following steps: preliminarily judging infectivity through disease information, performing similarity calculation on monitoring data of non-infectious diseases and historical data in a to-be-judged object set, and judging that the non-infectious diseases have public harmfulness when the matching number exceeds a set threshold value; in the judgment process, a basic object is compared with a basic object in a dynamically-updated to-be-judged object set, the early recognition effect on a hidden propagation event is especially remarkable, and the problem of early warning failure caused by neglecting the event development trend is effectively avoided through dynamically-updated set matching and trend analysis; according to the invention, the target object is determined according to the feature data, then the target object is divided into a plurality of category object sets, each category object set corresponds to one emergent public health event, and simultaneous monitoring of various types of emergent public health events can be realized.

Description

Technical Field

[0001] The present invention belongs to the field of public health, and specifically relates to a method and system for early warning of public health emergencies based on big data. Background Art

[0002] Public health emergencies are sudden, complex and highly dangerous, such as major infectious disease outbreaks and mass unexplained diseases. Their rapid spread may cause serious damage to public health. Therefore, it is very important to establish an early warning mechanism for public health emergencies.

[0003] Existing early warning technologies have a relatively limited mechanism for determining public hazards, often relying solely on direct assessments of the infectiousness of symptoms. They lack effective means for identifying non-infectious public health events. When faced with unexplained illnesses or symptoms without clear infectious characteristics, the lack of effective monitoring mechanisms often leads to missed reports or misjudgments, delaying early response.

[0004] The present invention provides a big data-based early warning method and system for public health emergencies to solve the above-mentioned technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method and system for early warning of public health emergencies based on big data. After collecting the monitoring data and symptom information of the basic objects, the present invention first preliminarily judges the infectiousness through the symptom information. For non-infectious diseases, the similarity between the monitoring data and the historical data in the set of objects to be determined is calculated. When the number of matches exceeds the set threshold, it is determined to be publicly harmful. In the judgment process, the basic objects are compared with the basic objects in the dynamically updated set of objects to be determined, which is particularly effective in early identification of hidden transmission events. Through dynamically updated set matching and trend analysis, the problem of early warning failure caused by ignoring the development trend of the event is effectively avoided, and the early warning accuracy of public health emergencies is improved.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for early warning of public health emergencies based on big data, comprising:

[0007] Collecting characteristic data of the basic object; wherein the characteristic data includes monitoring data and corresponding disease information;

[0008] Verify whether the disease of the basic object is of public danger based on the feature data; if yes, mark the basic object as the target object; if no, include the basic object in the set of objects to be determined;

[0009] Divide a number of target objects into a number of category object sets based on feature data; construct a basic heat map based on the position information of basic objects in the category object sets;

[0010] The basic heat map is predicted using a thermal prediction model to obtain a predicted heat map; early warning measures for public health emergencies are set based on the predicted heat map; the thermal prediction model is built based on the LSTM model.

[0011] Preferably, verifying whether the disease of the basic subject is of public danger based on the characteristic data includes:

[0012] Determine whether the corresponding disease is contagious based on the symptom information in the characteristic data; if so, determine that it is a public hazard; if not, analyze the monitoring data in the characteristic data;

[0013] Extracting a set of objects to be determined; matching basic objects in the set of objects to be determined that are similar to the monitoring data in the feature data;

[0014] When the number of matched basic objects is greater than the set number, all matched basic objects are determined to be public hazards; otherwise, they are not public hazards.

[0015] Preferably, matching basic objects similar to the monitoring data in the feature data in the set of objects to be determined includes:

[0016] Calculate the similarity between the basic object and the monitoring data of each basic object in the set of objects to be determined;

[0017] When the similarity is greater than a similarity threshold of one, the two basic objects are determined to be similar.

[0018] Preferably, the target objects are divided into several sets of category objects based on the feature data, including:

[0019] Extracting several target objects and corresponding feature data, and marking any target object as a reference object;

[0020] Calculate the similarity of feature data of other target objects and the reference object; when the similarity of feature data is greater than similarity threshold 2, classify the corresponding target object and the reference object into the same category object set.

[0021] Preferably, calculating the similarity of feature data of other target objects with respect to the reference object includes:

[0022] Set the weight coefficients of monitoring data and disease information in feature data;

[0023] The similarity between the monitoring data and symptom information of the benchmark object and other target objects is calculated, and the similarity of the feature data is obtained by weighted summation combined with the weight coefficient.

[0024] Preferably, when the attribute of the category object set is infectious, its basic heat map is modified, including:

[0025] constructing an individual contact network based on the mobile signaling data of the target object, and screening a number of suspected objects according to the individual contact network; wherein, nodes in the individual contact network are people, and edges are contact relationships;

[0026] modifying the basic heat map according to the number of suspected objects.

[0027] Preferably, when the attribute of the category object set is infectious, the basic heat map is modified, including:

[0028] determining a residence point based on the mobile signaling data of the target object, and matching the passenger flow of the residence point; wherein, the residence point refers to a position where the target object is likely to infect others with the disease;

[0029] determining the number of suspected objects of the residence point based on the passenger flow; and modifying the basic heat map based on the number of suspected objects of the residence point.

[0030] Preferably, the basic heat map is predicted by using a heat prediction model to obtain a predicted heat map, including:

[0031] constructing the heat prediction model based on an LSTM model;

[0032] inputting the time-series heat data of each position in the basic heat map into the heat prediction model to obtain heat prediction data corresponding to the position;

[0033] constructing the predicted heat map based on the heat prediction data of each position.

[0034] Preferably, the outbreak public health event early warning measures are set according to the predicted heat map, including:

[0035] extracting the heat data of each position at a future time through the predicted heat map; wherein, the heat data is converted based on the number of target objects;

[0036] determining the corresponding early warning level according to the heat data, and matching the early warning measures according to the early warning level.

[0037] The second aspect of the present application provides an outbreak public health event early warning system based on big data, including an early warning analysis module, and a data acquisition module connected thereto;

[0038] The early warning analysis module: is used to verify whether the disease suffered by the basic object has public harmfulness according to the characteristic data collected by the data acquisition module; if yes, the basic object is marked as a target object; if no, the basic object is included in a to-be-judged object set; and,

[0039] Used to divide several target objects into several category object sets based on feature data; construct a basic heat map according to the location information of basic objects in the category object set; use the thermal prediction model to predict the basic heat map to obtain a predicted heat map; set early warning measures for public health emergencies based on the predicted heat map.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. After collecting the monitoring data and symptom information of the basic objects, the present invention first preliminarily judges the infectiousness through the symptom information. For non-infectious diseases, the similarity between the monitoring data and the historical data in the set of objects to be determined is calculated. When the number of matches exceeds the set threshold, it is determined to be publicly harmful. During the judgment process, the basic objects are compared with the basic objects in the dynamically updated set of objects to be determined, which is particularly effective in early identification of hidden transmission events. Through dynamically updated set matching and trend analysis, the problem of early warning failure caused by ignoring the development trend of the event is effectively avoided, and the early warning accuracy of public health emergencies is improved. The present invention first determines the target object based on the feature data, and then divides the target object into several category object sets. Each category object set corresponds to a public health emergency, which can realize the simultaneous monitoring of multiple types of public health emergencies.

[0042] 2. When constructing a basic heat map, the present invention uses a GIS platform to form a basic heat map reflecting the current risk distribution based on the location information of the target objects in the category object set. For infectious categories, mobile phone signaling data is introduced to determine the target object's residence point, and the number of suspected objects is dynamically estimated by combining the flow of people during the residence point period and the infection coefficient. Through the superposition correction mechanism of "basic heat value + suspected risk increment", the problem that traditional static heat maps cannot capture hidden transmission risks is solved. When constructing a predictive heat map, an LSTM model is used to train the time series data of the basic heat map, and a predictive heat map is deduced by learning the time series laws of data development. This present invention breaks through the limitation of traditional heat maps that can only display the current status, and realizes the upgrade from "spatial risk visualization" to "spatiotemporal trend prediction", providing public health departments with a full-dimensional risk decision-making basis covering "current situation-future trend". BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1Schematic diagram of the method steps of the method for early warning of public health emergencies in an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of the principle of the method for determining whether there is public harm in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] See also Figure 1 The first embodiment of the present invention provides a method for early warning of public health emergencies based on big data.

[0048] S100: Collecting characteristic data of a basic object; wherein the characteristic data includes monitoring data and corresponding symptom information;

[0049] S200: Verify whether the disease of the basic object is of public danger based on the feature data; if yes, mark the basic object as a target object; if no, include the basic object in the set of objects to be determined;

[0050] S300: Dividing a plurality of target objects into a plurality of category object sets based on feature data; constructing a basic heat map according to position information of basic objects in the category object sets;

[0051] S400: Use a thermal prediction model to predict the basic thermal map to obtain a predicted thermal map; set early warning measures for public health emergencies based on the predicted thermal map; wherein the thermal prediction model is constructed based on the LSTM model.

[0052] In S100 , feature data of a basic object is collected.

[0053] A public health emergency is a sudden outbreak of a major infectious disease, a mass outbreak of an unexplained illness, a major food or occupational poisoning, or any other event that severely impacts public health. Early warning and response to public health emergencies focuses on data collection, and the subject of data collection is the primary object. Therefore, the primary object primarily refers to humans, but in some special cases, it can also include infectious animals.

[0054] Data collection methods include: hospitals collecting data when diagnosing and treating basic subjects. For example, if a basic subject (patient) is diagnosed with an infectious disease during a hospital visit, the hospital will privacy-enforce the relevant monitoring data and diagnostic information of the basic subject before uploading it. It also includes basic subjects (excluding animals) uploading data themselves through smart terminals. For example, if a basic subject (user) uploads relevant monitoring data through a smartphone and is diagnosed with (possibly) an infectious disease, the uploaded monitoring data and diagnostic information will be privacy-enforced before uploading. Basic subjects refer to those objects whose data needs to be collected when monitoring and warning of public health emergencies.

[0055] As mentioned above, public health emergencies include many types, including major infectious disease outbreaks and some group diseases caused by various reasons. Moreover, with the development of Internet technology, online consultation has become one of the options for users. Users can send monitoring data to the Internet at any time through their smartphones to seek help. When online hospitals or doctors analyze monitoring data, if the monitoring data or corresponding disease information uploaded by the user is publicly harmful, the object corresponding to the monitoring data still needs to be marked as a target object. Diversified data collection methods can improve the source of monitoring data, so that the development trend of sudden and urgent public health events can be accurately grasped, so that targeted solutions can be proposed later.

[0056] It should be noted that all data collection processes in the present invention that involve private data require authorization from the underlying object (patient or user) during collection and processing, and privacy processing is required during the data analysis process to avoid data leakage of the underlying object and ensure data security.

[0057] Within the feature data of basic objects, symptom information is determined by monitoring data and then integrated with the monitoring data to form feature data. Symptom information includes both the disease diagnosed based on monitoring data and the possible causes analyzed based on the monitoring data. This information can be either accurate disease information or a general description of the disease. For some rare or difficult diseases, the likelihood of a direct diagnosis is low, but a description based on monitoring data can be provided initially, with further symptom information subsequently supplemented through in-depth examination.

[0058] It is understandable that if the disease information clearly shows that the basic object is not suffering from the disease, there is no need to verify the public hazard; if the monitoring data of the basic object is not within the normal range, but accurate disease information cannot be given, it is still necessary to verify the public hazard.

[0059] The public hazard in this embodiment includes at least two meanings: the first is whether the cause of the disease in the symptom information corresponding to the basic object will cause public hazard, such as food poisoning. People who eat the same thing may also suffer from food poisoning. If many people suffer from food poisoning, it will cause public hazard. On the contrary, if the basic object accidentally causes a fracture, it is considered not to have public hazard; the second is whether the disease in the symptom information corresponding to the basic object will cause public hazard, such as a highly contagious infectious disease. If the basic object carries the virus in and out of public places, it is possible that many people will become sick, thereby causing public hazard.

[0060] When determining whether a disease poses a public hazard based on monitoring data and symptom information, the subject of the judgment is not the underlying object, but the disease corresponding to the underlying object's characteristic data. The judgment process requires combining monitoring data and symptom information to determine whether a disease poses a public hazard.

[0061] In a preferred embodiment, see Figure 2 , based on the characteristic data, verify whether the illness of the basic object is of public danger, including:

[0062] S210: Determine whether the corresponding disease is contagious based on the symptom information in the characteristic data; if so, determine that it is a public hazard; if not, analyze the monitoring data in the characteristic data;

[0063] S220: extracting a set of objects to be determined; matching basic objects similar to the monitoring data in the feature data in the set of objects to be determined;

[0064] S230: When the number of basic objects obtained by matching is greater than the set number, it is determined that all the basic objects matched are public hazards; otherwise, it is determined that they are not public hazards.

[0065] When determining whether a disease is contagious based on the symptom information in the feature data, the primary focus is on determining whether the disease possesses infectious characteristics based on the diagnosis results. If so, it can be determined to be a public health hazard. If the disease information does not correspond to an infectious symptom, such as food poisoning, subsequent analysis will require monitoring data from the feature data.

[0066] When analyzing monitoring data, it is necessary to extract a set of objects to be determined. A similarity analysis is performed between the monitoring data and the monitoring data of the basic objects in the set of objects to be determined. The Euclidean distance between the monitoring data can be calculated, and the reciprocal of the Euclidean distance is used as the similarity. When the similarity exceeds a set similarity threshold, the basic objects are considered similar to the basic objects in the set of objects to be determined.

[0067] In a preferred embodiment, matching the basic objects similar to the monitoring data in the feature data in the set of objects to be determined includes:

[0068] S221: Calculating the similarity between the basic object and the monitoring data of each basic object in the set of objects to be determined;

[0069] S222: When the similarity is greater than a similarity threshold of one, it is determined that the two basic objects are similar.

[0070] When the number of basic objects with similar monitoring data is greater than the set number, it can be determined to be a public hazard, and all matched basic objects will be marked as target objects. For example, the monitoring data of a certain basic object shows food poisoning. The similarity calculation is performed between the monitoring data of the basic object and the monitoring data of the basic objects in the set of objects to be determined. If multiple basic objects are found to be food poisoning, it can be determined to be a public hazard, and these multiple basic objects will be marked as target objects. If the number of matched basic objects is insufficient, such as no match, it should be determined that the food poisoning was caused by the basic object's own improper handling of food, which of course does not constitute a public hazard.

[0071] It should be noted that if the monitoring data in the characteristic data does not have matching disease information, the subsequent analysis of the monitoring data will be carried out directly; for example, if the monitoring data of a basic object is abnormal, but there is no matching disease diagnosis result, it is necessary to analyze whether there is a public hazard based on the monitoring data.

[0072] The characteristic data of several basic objects in the set of objects to be determined requires verification of public hazards. When the characteristic data of a basic object requires verification of public hazards, but the method of first analyzing the symptom information and then the monitoring data cannot determine whether it is a public hazard, the corresponding basic object and characteristic data will be temporarily included in the set of objects to be determined. Therefore, the set of objects to be determined may be empty.

[0073] The set of pending objects includes several basic objects that require a public hazard assessment, but whose public hazard cannot be determined based on the current data volume. For example, if the monitoring data for a basic object indicates food poisoning, but there are no basic objects with similar monitoring data in the set of pending objects, the basic object will naturally be determined to be not a public hazard. However, according to the above determination rules, this does not guarantee that the basic object will not be a public hazard. If the number of basic objects with similar monitoring data increases significantly in the next few hours, the basic object will also be determined to be a public hazard. In this case, the basic object can be directly included in the set of pending objects.

[0074] The coverage of the set of pending objects depends on the scope of the public health emergency warning. For example, if a public health emergency warning is needed for a prefecture-level city, the coverage of the set of pending objects could cover the entire province or just the corresponding prefecture-level city. Furthermore, the basic objects in the set of pending objects must be regularly updated to avoid distorting the public health risk assessment results.

[0075] S300: Divide a number of target objects into a number of category object sets based on feature data; and construct a basic heat map according to position information of basic objects in the category object sets.

[0076] After determining that the feature data corresponding to a base object is publicly hazardous, the base object and similar base objects are marked as target objects. The target objects are classified based on the feature data to obtain a set of class objects. Within each class object set, the target objects have similar feature data and are publicly hazardous, pointing to the same cause.

[0077] The location information of all target objects in the category set object is extracted and visualized using a GIS system to obtain a basic heat map, which shows the distribution of the target objects. It should be noted that to protect the privacy of the target objects, the accuracy of the location information does not need to be high enough. The street where the target object resides can be selected as the location information. This location information also forms the basis for subsequent corrections to the basic heat map and the generation of time-series thermal data. For example, extracting time-series thermal data for each location in the basic heat map means extracting time-series thermal data for each street.

[0078] In a preferred embodiment, the target objects are divided into several sets of category objects based on the feature data, including:

[0079] S310: extracting several target objects and corresponding feature data, and marking any target object as a reference object;

[0080] S320: Calculate the similarity of feature data of other target objects and the reference object; when the similarity of the feature data is greater than a similarity threshold 2, classify the corresponding target object and the reference object into the same category object set.

[0081] It is worth noting that when comparing the basic object with the basic objects in the set of objects to be determined, a similarity threshold of one is set. Its purpose is to determine whether it is a public hazard based on the number of similar basic objects. When dividing the target objects, a similarity threshold of two is set. Its purpose is to divide the target objects with public hazards into different sets to analyze the development trends of different causes. Therefore, similarity threshold one is greater than similarity threshold two. When the basic object has a certain similarity with the basic objects in the set of objects to be determined, the number can be counted to determine whether it is a public hazard. At this time, it is better to make a misjudgment than to miss it. When dividing the target objects, it is necessary to divide the target objects with higher similarity into one set to ensure that the reasons why the basic objects are public hazards are unified, so as to improve the development trend and prediction accuracy of public health emergencies under a single cause.

[0082] In another preferred embodiment, during the target object classification process, the monitoring data of the target object and the corresponding symptom information can be clustered and analyzed using a clustering algorithm, thereby dividing several target objects into different categories to obtain several category object sets.

[0083] In a preferred embodiment, calculating the similarity between the feature data of the other target objects and the reference object includes:

[0084] S321: Setting weight coefficients of monitoring data and symptom information in the feature data;

[0085] S322: Calculate the similarity between the monitoring data and symptom information of the benchmark object and other target objects, and perform weighted summation based on the weight coefficient to obtain the similarity of the feature data.

[0086] The weight coefficients for monitoring data and symptom information within the feature data can be determined based on the purpose of calculating similarity, namely, calculating similarity to categorize target objects with the same or similar public hazards into the same set of objects. For example, some diseases are contagious and their symptom information is similar, but the monitoring data varies across patients. In this case, the weight coefficient for the symptom information is greater than the weight coefficient for the monitoring data. Similarly, if a food-related mass poisoning outbreak occurs, the monitoring data may have similar monitoring data but different symptom information. In this case, the weight coefficient for the monitoring data is greater than the weight coefficient for the symptom information. In some extreme cases, where no corresponding symptom information is available, the weight coefficient for the symptom information is set to 0.

[0087] When calculating the similarity between the monitoring data and symptom information of the benchmark object and other target objects, the Euclidean distance of the monitoring data and the Euclidean distance of the symptom information can be calculated first. The larger the Euclidean distance, the smaller the similarity, and the smaller the Euclidean distance, the greater the similarity. Therefore, the reciprocal of the Euclidean distance (if meaningful) can be used as the similarity. When performing weighted summation, the weight coefficient is multiplied by the corresponding similarity and then added to obtain the similarity of the feature data. It should be noted that in order to avoid the reciprocal of the Euclidean distance being meaningless, a parameter is introduced. This parameter is smaller than the Euclidean distance and will not change significantly after adding it to the Euclidean distance. For example, if the Euclidean distance is 1, the parameter can be taken as 0.00000001, and the reciprocal of the sum of the Euclidean distance and the parameter is used as the similarity.

[0088] S400: Use a thermal prediction model to predict the basic thermal map to obtain a predicted thermal map; set early warning measures for public health emergencies based on the predicted thermal map; wherein the thermal prediction model is constructed based on the LSTM model.

[0089] After obtaining several sets of category objects, it is necessary to determine whether the basic heat map needs to be modified based on the attributes of the category object sets. The attributes of the category object sets include infectiousness and non-infectiousness. Infectiousness means that the disease suffered by the target object is contagious and the target object may infect others. Non-infectiousness means that the disease suffered by the target object is not contagious (or the conditions for infection are harsh), and the target object will not infect others even if they have normal contact with others.

[0090] If the attribute of the category object set is infectious, the basic heat map needs to be modified based on the mobile phone signaling data of the target object. In other words, the basic heat map needs to be modified to take into account the people who have been infected by the target object. If the attribute of the category object set is non-infectious, no modification is required.

[0091] When the attribute of the category object set is non-infectious, the residence of each target object in the category object set is identified and extracted, and a basic heat map is generated based on the residence of several target objects. As the number of target objects in the category object set increases or decreases, the basic heat map changes dynamically, showing the changes in the target objects in the monitoring area at the current and past time points.

[0092] When the attribute of the category object set is infectious, the residence and mobile phone signaling data (obtained through mobile phone signaling data) of each target object in the category object set are identified and extracted. A basic heat map is first established based on the residence of several target objects, and then the basic heat map is modified based on the mobile phone signaling data of several target objects.

[0093] In a preferred embodiment, when the attribute of the category object set is infectious, its basic heat map is modified, including:

[0094] S411: constructing an individual contact network based on the target object's mobile phone signaling data, and obtaining the number of suspected objects based on the individual contact network screening; wherein the nodes in the individual contact network are people and the edges are contact relationships;

[0095] S412: Modify the basic heat map according to the number of suspected objects.

[0096] When building an individual contact network based on mobile phone signaling data, it's necessary to obtain the signaling data of non-target individuals. The more signaling data there is, the more complete the individual contact network will be. However, since signaling data contains a large amount of personal privacy, obtaining signaling data on non-target individuals is difficult. Therefore, the number of suspected individuals can also be estimated based on crowd density.

[0097] In another preferred embodiment, when the attribute of the category object set is infectious, the basic heat map is modified, including:

[0098] S421: Determine a dwelling point based on the target subject's mobile phone signaling data and match the flow of people at the dwelling point; wherein the dwelling point refers to a location where the target subject may transmit the disease to others;

[0099] S422: Determine the number of suspected objects at the residence point based on the flow of people; and modify the basic heat map based on the number of suspected objects at the residence point.

[0100] When calculating the number of suspected subjects at a location based on foot traffic, we first extract the infection coefficient based on the number of target subjects at the location during that period. This infection coefficient is then multiplied by the foot traffic to determine the number of suspected subjects for that period. Suspected subjects are the estimated number of people who could have been infected by the target subject, based on foot traffic and the infection coefficient.

[0101] Dwelling points are determined based on the infectiousness of the disease corresponding to the basic objects in the category object set. For example, if an infectious disease can spread within one minute of contact, mobile phone signaling data can be used to identify locations where the target object stayed for more than one minute and these locations can be considered dwelling points. Foot traffic refers to the total number of people who visited the dwelling point during the time the target object stayed there. This traffic can be measured in real time or estimated based on historical data.

[0102] In S400, the basic thermal map is predicted using the thermal prediction model to obtain a predicted thermal map.

[0103] In a preferred embodiment, the basic thermal map is predicted using a thermal prediction model to obtain a predicted thermal map, including:

[0104] S431: Constructing thermal prediction model based on LSTM model;

[0105] S432: Inputting the time series thermal data of each location in the basic thermal map into a thermal prediction model to obtain thermal prediction data for the corresponding location;

[0106] S433: Construct a prediction heat map based on the thermal prediction data of each location.

[0107] The LSTM model is a time series prediction model that can predict future trends of data based on its temporal changes. Of course, other time series prediction models can also be used to build a thermal prediction model.

[0108] A basic heat map displays the changes in target objects (including suspected objects) at each location in the monitoring area. Using time constraints, we can determine the status of target objects at each location at a specific moment in the past. Training samples can be extracted from the basic heat map. Each training sample includes the number of target objects at a specific moment in time. The data from the first few moments in each training sample serves as the input to an LSTM model, while the data from the last few moments serves as the output. The LSTM model can then be trained to produce a thermal prediction model. Alternatively, the LSTM model can be trained using data from similar public health emergencies.

[0109] It is worth noting that when the attribute of the category object set is contagious, the data at each moment includes determined target objects and suspected objects. The target objects and suspected objects are input into the LSTM model as one-dimensional vectors respectively. The output of the LSTM model also includes the number of target objects and suspected objects at future moments.

[0110] In a preferred embodiment, setting early warning measures for public health emergencies based on the prediction heat map includes:

[0111] S441: extracting thermal data of each location at a future time through a predicted thermal map; wherein the thermal data is converted based on the number of target objects;

[0112] S442: Determine the corresponding warning level based on the thermal data, and match warning measures according to the warning level.

[0113] After obtaining the predicted heat map, thermal data for any future moment can be extracted from the map. This thermal data can be the number of target objects or a conversion calculation based on the number of target objects. The thermal data is used to determine the warning level for the corresponding location, and the warning level is then used to determine the warning measures to be taken at the corresponding moment. Warning measures include initiating an emergency response at the corresponding level, implementing regional risk control, and restricting the movement of people in the area.

[0114] The second embodiment of the present invention provides a public health emergency early warning system based on big data, including an early warning analysis module and a data acquisition module connected thereto;

[0115] Early warning analysis module: used to verify whether the disease of the basic object is of public danger based on the characteristic data collected by the data collection module; if yes, mark the basic object as the target object; if not, include the basic object in the set of objects to be determined; and

[0116] Used to divide several target objects into several category object sets based on feature data; construct a basic heat map according to the location information of basic objects in the category object set; use the thermal prediction model to predict the basic heat map to obtain a predicted heat map; set early warning measures for public health emergencies based on the predicted heat map.

[0117] The data acquisition module connects to various monitoring devices or databases to collect characteristic data from subjects within the monitoring area. This data, including various physical and physiological data, is primarily used to assess the subject's physical condition and determine whether they pose a public health hazard. The early warning analysis module, which interacts with the data acquisition module and database, integrates various data to analyze the possibility of a public health emergency and to predict and warn of trends in already occurring public health emergencies.

[0118] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A public health emergency early warning method based on big data, characterized in that: include: Collecting characteristic data of the basic object; wherein the characteristic data includes monitoring data and corresponding disease information; Verify whether the disease of the basic object is of public danger based on the characteristic data; if yes, mark the basic object as a target object; if no, include the basic object in the set of objects to be determined; Dividing a plurality of target objects into a plurality of category object sets based on the feature data; constructing a basic heat map according to the position information of basic objects in the category object sets; The basic heat map is predicted using a thermal prediction model to obtain a predicted heat map; early warning measures for public health emergencies are set based on the predicted heat map; the thermal prediction model is built based on the LSTM model.

2. The method for early warning of public health emergencies based on big data according to claim 1, characterized in that: Verifying whether the illness suffered by the basic subject is of public danger based on the characteristic data includes: Determine whether the corresponding disease is contagious based on the symptom information in the characteristic data; if so, determine that it is a public hazard; if not, analyze the monitoring data in the characteristic data; Extracting a set of objects to be determined; matching basic objects similar to the monitoring data in the feature data in the set of objects to be determined; When the number of basic objects obtained by matching is greater than a set number, it is determined that all the matched basic objects have public hazards; otherwise, they do not have public hazards.

3. The method for early warning of public health emergencies based on big data according to claim 2, characterized in that: Matching basic objects similar to the monitoring data in the feature data in the set of objects to be determined includes: Calculating the similarity between the basic object and the monitoring data of each basic object in the set of objects to be determined; When the similarity is greater than a similarity threshold of one, the two basic objects are determined to be similar.

4. The method for early warning of public health emergencies based on big data according to claim 1, characterized in that: The target objects are divided into a plurality of category object sets based on the feature data, including: Extracting several target objects and corresponding feature data, and marking any target object as a reference object; Calculate the similarity of feature data between the other target objects and the reference object; when the similarity of feature data is greater than a similarity threshold 2, classify the corresponding target object and the reference object into the same category object set.

5. The method for early warning of public health emergencies based on big data according to claim 4, characterized in that: Calculating the similarity between the feature data of the other target objects and the reference object includes: Setting weight coefficients of monitoring data and symptom information in the feature data; The similarity between the monitoring data and symptom information of the reference object and other target objects is calculated, and the similarity of the feature data is obtained by weighted summation combined with the weight coefficient.

6. The method for early warning of public health emergencies based on big data according to claim 1, characterized in that: When the attribute of the class object set is infectious, the basic heat map thereof is modified, including: Constructing an individual contact network based on the target object's mobile phone signaling data, and obtaining the number of suspected objects based on the individual contact network; wherein the nodes in the individual contact network are people and the edges are contact relationships; The basic heat map is modified according to the number of the suspected objects.

7. The method for early warning of public health emergencies based on big data according to claim 1, characterized in that: When the attribute of the class object set is infectious, the basic heat map is modified, including: Determine a dwelling point based on the target subject's mobile phone signaling data and match the traffic volume at the dwelling point; a dwelling point is a location where the target subject may transmit the disease to others; The number of suspected objects at the residence point is determined based on the human flow; and the basic heat map is modified based on the number of suspected objects at the residence point.

8. The method for early warning of public health emergencies based on big data according to claim 1, characterized in that: The thermal prediction model is used to predict the basic thermal map to obtain the predicted thermal map, including: Build a thermal prediction model based on the LSTM model; Inputting the time series thermal data of each position in the basic thermal map into the thermal prediction model to obtain thermal prediction data of the corresponding position; A predicted heat map is constructed based on the thermal prediction data for each location.

9. The method for early warning of public health emergencies based on big data according to claim 8, characterized in that: Setting early warning measures for public health emergencies based on the predicted heat map, including: Extract thermal data of each location at a future time through a predicted thermal map; where the thermal data is converted based on the number of target objects; A corresponding warning level is determined according to the thermal data, and warning measures are matched according to the warning level.

10. A public health emergency early warning system based on big data, used to implement the public health emergency early warning method based on big data according to any one of claims 1 to 9, characterized in that: It includes an early warning analysis module and a data acquisition module connected thereto; Early warning analysis module: used to verify whether the disease of the basic subject is a public hazard based on the characteristic data collected by the data collection module; If yes, mark the base object as the target object; if no, include the base object in the set of objects to be determined; and Used to divide a number of target objects into a number of category object sets based on the feature data; construct a basic heat map according to the location information of the basic objects in the category object set; use a thermal prediction model to predict the basic heat map to obtain a predicted heat map; set early warning measures for public health emergencies according to the predicted heat map.

Images