WO2022178947A1 - 基于多维度的监测预警方法、装置、设备及存储介质 - Google Patents

基于多维度的监测预警方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2022178947A1
WO2022178947A1 PCT/CN2021/084538 CN2021084538W WO2022178947A1 WO 2022178947 A1 WO2022178947 A1 WO 2022178947A1 CN 2021084538 W CN2021084538 W CN 2021084538W WO 2022178947 A1 WO2022178947 A1 WO 2022178947A1
Authority
WO
WIPO (PCT)
Prior art keywords
factor
dimension
early warning
factors
specified
Prior art date
Application number
PCT/CN2021/084538
Other languages
English (en)
French (fr)
Inventor
唐蕊
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2022178947A1 publication Critical patent/WO2022178947A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/80ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for detecting, monitoring or modelling epidemics or pandemics, e.g. flu
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/22Matching criteria, e.g. proximity measures
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/088Non-supervised learning, e.g. competitive learning
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02ATECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • the present application relates to the technical field of intelligent customer service, and in particular, to a multi-dimensional monitoring and early warning method, device, equipment and storage medium.
  • infectious disease surveillance and early warning has become the focus of infectious disease prevention and control.
  • the existing methods of infectious disease monitoring and early warning are mainly based on the number of confirmed cases of infectious diseases, and early warning of abnormal growth and changes in the number of confirmed cases of infectious diseases.
  • the inventor realized that the infectious disease early warning based on the number of confirmed cases of infectious diseases relies on a single data source, which leads to technical problems such as limited monitoring scope, delayed early warning, and low early warning push accuracy.
  • the main purpose of this application is to provide a multi-dimensional monitoring and early warning method, which aims to solve the problem that the infectious disease early warning based on the number of confirmed cases of infectious diseases relies on a single data source, resulting in limited monitoring scope, early warning delay, and early warning push accuracy. Low technical issues.
  • This application proposes a multi-dimensional monitoring and early warning method, including:
  • the dimension factor is the name of a specified node in the knowledge graph that has an associated edge with the node corresponding to the object to be monitored, and the specified node includes at least two ;
  • the specified dimension factor is used as an early warning feedback result corresponding to the specified data set.
  • the designated data set includes electronic case sets corresponding to all monitoring users
  • the to-be-monitored object includes designated infectious diseases
  • statistics corresponding to each of the dimension factors respectively steps for sample data including:
  • the standard sub-factor set includes a plurality of sub-factors, and the first dimension factor is any one of all dimension factors
  • the sample set corresponding to each of the dimension factors is selected from the electronic case set as the sample data corresponding to each of the dimension factors.
  • This application also provides
  • a computer device comprising a memory and a processor, wherein computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the processor is made to execute a multi-dimensional monitoring and early warning method. step:
  • the dimension factor is the name of a specified node in the knowledge graph that has an associated edge with the node corresponding to the object to be monitored, and the specified node includes at least two ;
  • the specified dimension factor is used as an early warning feedback result corresponding to the specified data set.
  • the present application also provides a storage medium storing computer-readable instructions.
  • the computer-readable instructions are executed by one or more processors, the one or more processors execute the multi-dimensional-based monitoring and early warning method.
  • the dimension factor is the name of a specified node in the knowledge graph that has an associated edge with the node corresponding to the object to be monitored, and the specified node includes at least two ;
  • the specified dimension factor is used as an early warning feedback result corresponding to the specified data set.
  • the present application proposes an infectious disease monitoring and early warning system based on multi-dimensional information, which uses multi-dimensional information related to infectious diseases to construct multiple early warning information, and calculates early warning signals from multiple dimensions. Form a set of early warning signals composed of multiple early warning signals to focus on the occurrence of infectious diseases and improve the sensitivity and accuracy of monitoring and early warning.
  • FIG. 1 is a schematic flowchart of a multi-dimensional monitoring and early warning method according to an embodiment of the present application
  • FIG. 2 is a schematic structural diagram of a deep neural network according to an embodiment of the present application.
  • FIG. 3 is a schematic flowchart of a multi-dimensional monitoring and early warning system according to an embodiment of the present application
  • FIG. 4 is a schematic diagram of an internal structure of a computer device according to an embodiment of the present application.
  • a multi-dimensional monitoring and early warning method includes:
  • S1 Determine the dimension factor corresponding to the object to be monitored from the knowledge graph, where the dimension factor is the name of a specified node in the knowledge graph that has an associated edge with the node corresponding to the object to be monitored, and the specified node at least includes two;
  • the above-mentioned knowledge graph is a graph corresponding to pre-formed disease data.
  • it can be generated by expert knowledge, and on the other hand, it can be automatically generated by artificial intelligence methods such as big data mining, natural language processing, and machine learning.
  • the knowledge graph consists of nodes and associated edges between nodes, where each node represents an entity or concept, and the associated edges between nodes represent the relationship between entities and entities, or between entities and concepts.
  • the entities or concepts of the embodiments of the present application are symptoms, diseases, and the like. For example, influenza is a node (representing an entity as a disease), fever is also a node (representing a concept as a symptom), and cough is a node (representing a concept as a symptom).
  • the above-mentioned objects to be monitored are infectious diseases such as influenza, and the above-mentioned dimension factors include symptoms, diagnoses, medicines, tests, and inspections associated with influenza nodes.
  • the above designated data set refers to the influenza data set in a specific geographical area.
  • the above specific geographical area includes a hospital or a specific area, etc.
  • the influenza data set includes the number of influenza patients.
  • the above sample data refers to the symptoms of each person during the treatment process. , diagnostics, medicines, inspections and inspections and other aspects of data statistics.
  • the concepts associated with influenza in the knowledge graph also include diagnosis, medicines, tests, and examinations.
  • Symptom dimension factors include, for example, fever, epistaxis, sore throat, thick sputum, and wheezing
  • diagnostic dimension factors include, for example, acute pharyngitis, acute tonsillitis, acute bronchitis, unspecified acute lower respiratory tract infection, streptococcal pneumonia
  • drug dimension factors such as acetylsalicylic acid, dextromethorphan, peramivir, aminocaxanthine, phenol aminocaffeine, etc.
  • test dimension factors such as blood routine, cerebrospinal fluid, viral nucleic acid, blood Biochemical and arterial blood gas analysis divide factors, check dimension factors, such as chest CT, chest X-ray, and build multi-dimensional data by building relationships between concepts associated with influenza.
  • the specified disease node in addition to the specified disease node to the symptom node with edge connection, that is, to find the associated symptoms for the disease.
  • the specified disease node is connected to the edge-connected inspection node, that is, to find the associated test for the disease; to find the specified disease node to the edge-connected inspection node, that is, to find the associated inspection for the disease, by introducing multiple infectious diseases and influenza.
  • the dimensional information is used for comprehensive early warning, which improves the sensitivity rate of the model.
  • the present application proposes an infectious disease monitoring and early warning system based on multi-dimensional information, which uses multi-dimensional information related to infectious diseases to construct multiple early warning information, and calculates early warning signals from multiple dimensions. Form a set of early warning signals composed of multiple early warning signals to focus on the occurrence of infectious diseases such as influenza, and improve the sensitivity and accuracy of monitoring and early warning.
  • the designated data set includes electronic case sets corresponding to all monitoring users, the objects to be monitored include designated infectious diseases, and in the designated data sets corresponding to the objects to be monitored, statistics corresponding to each of the dimension factors are Step S2 of the sample data, including:
  • S21 Acquire a standard sub-factor set corresponding to the first dimension factor of the designated infectious disease, wherein the standard sub-factor set includes a plurality of sub-factors, and the first dimension factor is any one of all dimension factors;
  • S22 Extract the actual factor corresponding to the first dimension factor from the designated electronic case corresponding to the designated user to form a measured factor set, wherein the designated user is any one of all monitoring users, and the designated electronic case be any electronic case in the electronic case set;
  • Influenza infectious diseases take influenza as an example.
  • Influenza infectious diseases have the following five data sources, including "influenza-related symptoms”, “influenza-related diagnoses”, “influenza-related medicines”, “ Influenza-Associated Tests", “Influenza-Associated Tests”.
  • the data of each of the above data sources can be referred to as the data of a dimension factor.
  • a standard sub-factor set consisting of multiple sub-factors is included, such as ⁇ symptom 1, symptom 2, symptom 3, ...symptom n ⁇ .
  • the measured factor set is formed, for example, the actual factor set is ⁇ symptom 1, symptom 2, symptom 3 ⁇ .
  • the ratio of the actual factor to the standard sub-factor set the matching degree between the measured factor set and the standard factor set is determined, and it is determined whether the electronic case of the current patient belongs to the sample data of the dimension factor of symptoms.
  • step S23 of calculating the matching degree between the measured factor set and the standard factor set includes:
  • S231 Acquire a keyword corresponding to a designated measured factor in the measured factor set, wherein the designated measured factor is any factor in the measured factor set;
  • S235 Calculate the matching degree between the measured factor set and the standard factor set according to the number of factors and the total amount of factors in the standard factor set.
  • statistics are separately performed on the data of each dimension factor from a specified data set.
  • flu-related symptoms first, apply deep learning technology and natural language processing technology, such as named entity recognition technology, etc., to obtain symptoms and symptom-related attributes from the chief complaint of each patient's electronic medical record, such as the occurrence of symptoms. location, duration, severity, etc.; then, according to the keyword matching technology, the keywords of the patient's symptom information and the keywords of the influenza-related symptom items are matched, and the successfully matched patients are marked as influenza symptom-related cases.
  • deep learning technology and natural language processing technology such as named entity recognition technology, etc.
  • matching degree Intersection Symptoms/Influenza Associated Symptoms.
  • flu-related symptoms ⁇ symptom 1, symptom 2, symptom 3 ⁇
  • patient 1's symptom ⁇ symptom 1, symptom 2, symptom 3, symptom 4 ⁇
  • patient 2's symptom ⁇ symptom 3, symptom 4, symptom 5 ⁇ .
  • the matching process is also similar for other dimension factor data of other diseases, and will not be described in detail.
  • step S23 of calculating the matching degree between the measured factor set and the standard factor set includes:
  • S2303 Input the vectors corresponding to the measured factor set and the standard factor set respectively into a deep learning network to obtain a first low-dimensional vector corresponding to the measured factor set and a second low-dimensional vector corresponding to the standard factor set ;
  • S2304 Calculate the similarity between the first low-dimensional vector and the second low-dimensional vector
  • S2305 Use the similarity between the first low-dimensional vector and the second low-dimensional vector as the matching degree between the measured factor set and the standard factor set.
  • the keywords are not matched successfully, thereby causing errors.
  • the deep learning network-based representation learning model For patients who have not been successfully matched, continue to apply the deep learning network-based representation learning model, perform fuzzy matching between the vector represented by the patient's structured information and the definition vector of each dimension factor data, and mark the patients with higher matching scores as influenza symptoms related cases.
  • the above-mentioned consequential information refers to the symptom data obtained from the chief complaint of the electronic medical record of each patient and the textual description of the attributes related to the symptom for the dimensional factor data of symptoms.
  • a multi-hot encoding vector is constructed by constructing the resulting information.
  • the dimension of the vector is the number of all symptom-related attributes.
  • Each dimension factor corresponds to an attribute of a symptom. A value of 1 indicates that the symptom occurs, and a value of 0 indicates that there is no such symptom.
  • a representation learning model of flu-related symptoms is obtained.
  • the input of the representation learning model is the above multi-hot encoding vector, and the output is also a multi-hot encoding vector.
  • the input and output multi-hot encoding vectors are high-dimensional sparse vectors.
  • the high-dimensional sparse vectors are converted into low-dimensional dense vectors in the middle hidden layer through a network of three-layer fully connected hidden layers with an autoencoder structure.
  • the low-dimensional dense vector output by the middle hidden layer is used as the definition vector B of each dimension factor data associated with influenza.
  • the similarity is calculated by formula (1) through the vector A output by the middle hidden layer of the above-mentioned representation learning model, that is, the matching degree is obtained.
  • the electronic cases of patients with a matching degree exceeding a threshold are marked as cases related to influenza symptoms, and the threshold can generally be set to 0.8.
  • the expression of formula (1) is: Among them, n represents the vector dimension of vectors A and B, and i represents the ith vector dimension.
  • step S3 of calculating the early warning signal set corresponding to each of the dimension factors according to the sample data corresponding to each of the dimension factors includes:
  • S31 Acquire a time unit amount of a preset length and a statistical time period corresponding to the specified data set;
  • S34 Obtain sub-sample data corresponding to the second dimension factor in each of the sub-time periods, wherein the second dimension factor is any one of all the dimension factors, and the sample data of the second dimension factor is determined by The sub-sample data corresponding to each of the sub-time periods are combined to obtain;
  • the statistics of the number of influenza-related cases are separately performed on the data of each dimension factor, and the statistics are carried out according to the time unit.
  • the number of influenza drug-related cases corresponding to the time unit is sorted out; for influenza-related inspection and inspection items, the number of influenza inspection-related cases corresponding to the time unit is sorted, and finally each time unit (such as days, hours) is formed. etc.) of influenza symptom-related cases.
  • early warning signals based on time and indicator dimensions, and calculate early warning signal values.
  • early warning signals can be constructed in various ways, such as including the week-on-week growth rate in days as the time unit, the week-on-year growth rate, etc.
  • week-on-week growth rate According to the formula "(the sum of the values of the last seven days - the sum of the values of the previous seven days) / the value of the previous seven days Calculated by "sum”; week-on-year growth rate: calculated according to the formula "(the sum of the values of the last seven days – the sum of the values of the seven days in the same period last year)/the sum of the values of the seven days in the same period last year", for example, this year is 2021 January 1, 2021 to January 7, 2021 (sum of the last seven days), for "the sum of the seven days of the same period last year” means January 1, 2020 to January 7, 2020; the most recent Three-day growth rate: Average the daily growth rate of the last three days; Last seven-day growth rate: Average the daily growth rate of the last seven days, etc.
  • the early warning signal is calculated separately. For example, for influenza-related symptoms, according to the value of the symptoms in a certain time unit, calculate the week-on-week growth rate of symptoms, the year-on-year growth rate of symptoms, etc.; for influenza-related symptoms For drugs, according to the value of the drug in a certain time unit, calculate the weekly growth rate of the drug, the year-on-year growth rate of the drug, and so on.
  • the early warning signal value When calculating the early warning signal value, it is necessary to calculate the historical data in order to obtain the week-on-week growth rate and the week-on-year growth rate.
  • the symptom dimension factor for the historical data of the number of cases related to influenza symptoms (generally the historical data of the past five years), historical values are calculated separately for multiple early warning signals.
  • a threshold value is determined according to the percentage based on its historical value, for example, 95% of the historical value is taken as the threshold value of the early warning signal.
  • a threshold value is determined according to the percentage based on its historical value, for example, 95% of the historical value is taken as the threshold value of the early warning signal.
  • the week-to-week growth rate of symptoms based on the data in 2020 was obtained. Calculated with days as the time unit, and obtained the value of the weekly growth rate of 365 symptoms. Sort, and take the value at the 95% position as the threshold for the week-to-week growth rate of symptoms. Similarly, under the symptom dimension factor, for other early warning signals, "the threshold for the weekly growth rate of symptoms", “the threshold for the growth rate of symptoms in the last three days” and so on can also be obtained.
  • the thresholds of the early warning signals corresponding to each dimension factor can be obtained, such as “threshold of the growth rate of drugs in the last three days", “the growth rate of drugs in the last three days” threshold” and so on.
  • step S36 of forming the early warning signals corresponding to each of the sub-time periods respectively to form the early warning signal set corresponding to the second dimension factor it includes:
  • S362 Determine whether the correlation coefficient between the first early warning signal and the second early warning signal is greater than a preset correlation threshold, wherein the first early warning signal and the second early warning signal are in the early warning signal set corresponding to the second dimension factor any two early warning signals;
  • S364 Use the early warning signal set of the second dimension factor that has undergone the selection and retention process as a new early warning signal set corresponding to the second dimension factor.
  • multiple early warning signals are designed to be constructed from time and data indicators to provide early warning of influenza.
  • time and data indicators can include year-on-year, month-on-month, historical percentiles, and growth rates.
  • the multiple early warning signals are analyzed.
  • Correlation filter The specific steps are: firstly apply the correlation test method, for example, calculate the correlation coefficient between any two early warning signals through the Pearson correlation coefficient; then screen the early warning signals based on the correlation coefficient threshold, for example, the correlation coefficient threshold is set to 0.3 , that is, 0.3 and below indicate no or weak correlation. If the correlation coefficient of the two early warning signals exceeds the threshold, it means that the changes of the two early warning signals are similar.
  • Another early warning signal such as deleting the early warning signal with a smaller value to ensure the early warning sensitivity; the correlation between the final early warning signals does not exceed the threshold to highlight different data analysis angles, and use the early warning after the selection and retention process. signal for comprehensive early warning.
  • each early warning signal in this application are the same, and the goal is to have each early warning signal represent abnormal changes in the changing trend of infectious diseases from different data perspectives, so as to better conduct comprehensive early warning and reduce false alarms.
  • a comprehensive early warning is performed, as long as any early warning signal in the early warning signal set used for the comprehensive early warning exceeds the threshold, the early warning will be triggered.
  • step S6 of using the specified dimension factor as the early warning feedback result corresponding to the specified data set it includes:
  • the early warning in the embodiment of the present application is a hierarchical structure with two layers of early warning.
  • the calculation of the first level early warning is completed, and the comprehensive early warning signal result is obtained according to the data of each dimension factor, and the dimension factor that triggers the early warning is determined; then, according to the first level early warning
  • the result of the second-level warning is calculated, that is, the proportion of the number of dimension factors that trigger the warning to the total number of dimension factors is calculated, the warning result of the second-level warning is obtained, and the severity level of the warning is determined.
  • the first level of early warning takes the dimension factor of influenza symptoms as an example, and calculates historical values for multiple early warning signals for the historical data of the number of cases related to influenza symptoms. Thresholds are determined for each early warning signal as a percentage based on its historical value. After the threshold is determined for each early warning signal, the values of multiple early warning signals are integrated to carry out early warning. If the value of multiple early warning signals are below their respective thresholds, no warning is issued; if the value of any one early warning signal exceeds its threshold, an early warning is issued.
  • the second-level early warning is the early warning result of comprehensive analysis of the data of each dimension factor. For example, after it is determined that the dimension factor of symptoms triggers an early warning, then it is determined whether the dimension factor related to influenza is an early warning; and/or whether the data source of drugs related to influenza is early warning; and/or to determine whether the inspection and inspection related to influenza Whether the data source is early warning, etc., according to the following rules for early warning risk level early warning, and give early warning information.
  • the warning signals corresponding to all dimension factors do not trigger the warning, the risk is low; if the warning signal of one dimension factor triggers the warning, the risk is medium, and the details of the data corresponding to the dimension factor that triggers the warning are displayed as warning information; If there are two or more dimensional factors that trigger an early warning, the risk is high, and the details of the data corresponding to all the dimensional factors that trigger the warning will be displayed as early warning information. For example, in the data source of symptoms, if the early warning signals of all symptoms do not exceed the threshold, then the dimension factor of symptoms has no early warning information; in the dimension factor of drugs, the early warning signal of the week-on-week growth rate of drugs exceeds the threshold. Threshold, then the dimension factor of drugs is marked as an early warning, and will be processed as the information of the second-level early warning.
  • the present application builds an infectious disease early warning scheme based on multi-dimensional factor data through a two-layer early warning mechanism.
  • a hierarchical early warning mechanism and a wealth of influenza-related early warning signals By designing a hierarchical early warning mechanism and a wealth of influenza-related early warning signals, the sensitivity of early warning can be improved.
  • early warning signals with low correlation are used for early warning, and anomalies are captured from different data analysis angles to further improve the accuracy of early warning.
  • the dimension factor of symptoms triggers an early warning.
  • the two early warning signals corresponding to the symptoms actually exceed their thresholds.
  • the early warning signal value of the week-on-week growth rate is 200%, which exceeds the early warning signal threshold for the growth rate of the week-on-month period by 20%
  • the early warning signal value for the growth rate of the last three days is 300%, which exceeds the early warning signal threshold for the growth rate of the last three days by 50%
  • the displayed warning details are:
  • Early warning For medium risk, the flu-related symptom dimension factor triggers an early warning, specifically including the week-on-week growth rate of flu-related symptoms is 200%, exceeding the threshold; the growth rate of flu-related symptoms in the last three days is 300%, exceeding the threshold.
  • a multi-dimensional monitoring and early warning device includes:
  • the determination module 1 is used to determine the dimension factor corresponding to the object to be monitored from the knowledge graph, wherein the dimension factor is the name of the specified node in the knowledge graph that has an associated edge with the node corresponding to the object to be monitored, and the The specified node includes at least two;
  • a statistical module 2 configured to count the sample data corresponding to each of the dimension factors in the designated data set corresponding to the object to be monitored;
  • the first calculation module 3 is configured to calculate the early warning signal sets corresponding to each of the dimension factors according to the sample data corresponding to each of the dimension factors respectively;
  • Judging module 4 for judging whether there is a designated early warning signal exceeding the threshold in the early warning signal sets corresponding to each of the dimension factors respectively;
  • the first acquisition module 5 is configured to acquire the specified dimension factor to which the specified early warning signal belongs if there is a specified early warning signal exceeding the threshold;
  • module 6 it is used to use the specified dimension factor as the early warning feedback result corresponding to the specified data set.
  • the designated data set includes electronic case sets corresponding to all monitoring users
  • the to-be-monitored object includes designated infectious diseases
  • the statistical module 2 includes:
  • a first obtaining unit configured to obtain a standard sub-factor set corresponding to the first dimension factor of the designated infectious disease, wherein the standard sub-factor set includes a plurality of sub-factors, and the first dimension factor is one of all dimension factors. any of the;
  • the extraction unit is configured to extract the actual factor corresponding to the first dimension factor from the designated electronic case corresponding to the designated user to form a measured factor set, wherein the designated user is any one of all monitoring users, and the Designate an electronic case as any electronic case in the electronic case set;
  • a first computing unit configured to calculate the degree of matching between the measured factor set and the standard factor set
  • an induction unit configured to summarize the designated electronic case into a first sample set corresponding to the first dimension factor if a preset matching degree is reached;
  • a screening unit configured to screen the sample sets corresponding to each of the dimension factors from the electronic case set according to the formation method of the first sample set corresponding to the first dimension factor, as the sample sets corresponding to each of the dimension factors respectively sample data.
  • the computing unit includes:
  • a first obtaining subunit configured to obtain a keyword corresponding to a designated measured factor in the measured factor set, wherein the designated measured factor is any factor in the measured factor set;
  • a judging subunit for judging whether the keyword corresponding to the specified measured factor exists in the standard factor set
  • a marking subunit used for marking the specified measured factor as a matching factor if it exists in the standard factor set
  • a statistical subunit used to count the number of factors marked as matching factors in the measured factor set
  • a first calculation subunit configured to calculate the matching degree between the measured factor set and the standard factor set according to the number of factors and the total amount of factors in the standard factor set.
  • the computing unit includes:
  • a second acquiring subunit configured to acquire the structured information corresponding to the measured factor set and the standard factor set respectively
  • a conversion subunit used to convert the structured information corresponding to the measured factor set and the standard factor set respectively into a vector
  • the input subunit is used to input the vectors corresponding to the measured factor set and the standard factor set respectively into the deep learning network to obtain the first low-dimensional vector corresponding to the measured factor set and the first low-dimensional vector corresponding to the standard factor set.
  • a second calculation subunit configured to calculate the similarity between the first low-dimensional vector and the second low-dimensional vector
  • calculation module 3 includes:
  • a second obtaining unit configured to obtain a time unit amount of a preset length and a statistical time period corresponding to the specified data set
  • a determining unit configured to determine the corresponding start time and end time of the statistical time period
  • a truncation unit configured to sequentially truncate the statistical time period into a plurality of sub-periods according to the natural time sequence with the start time as a starting point according to the time unit amount;
  • a third acquiring unit configured to acquire sub-sample data corresponding to the second dimension factor in each of the sub-time periods, wherein the second dimension factor is any one of all dimension factors, and the second dimension factor
  • the sample data of are obtained by combining the sub-sample data corresponding to each of the sub-time periods respectively;
  • a second calculation unit configured to calculate the early warning signals corresponding to each of the sub-time periods according to the sub-sample data of the second dimension factor corresponding to each of the sub-time periods;
  • a forming unit configured to form the early warning signals corresponding to each of the sub-time periods respectively into a set of early warning signals corresponding to the second dimension factor
  • the third calculation unit is configured to calculate the early warning signal set corresponding to each of the dimension factors according to the calculation method of the early warning signal set corresponding to the second dimension factor.
  • calculation module 3 includes:
  • a fourth calculation unit configured to calculate the correlation coefficient between the early warning signals in pairs in the early warning signal set corresponding to the second dimension factor
  • a judgment unit used for judging whether the correlation coefficient of the first early warning signal and the second early warning signal is greater than a preset correlation threshold, wherein the first early warning signal and the second early warning signal are the early warning corresponding to the second dimension factor Any two early warning signals in the signal set;
  • a selection unit configured to select and reserve the first warning signal and the second warning signal if it is greater than a preset correlation threshold, and delete the first warning signal or the second warning signal;
  • the set of early warning signals of the second dimension factor that has undergone the selection and retention process is used as a new set of early warning signals corresponding to the second dimension factor.
  • the multi-dimensional monitoring and early warning device includes:
  • a second obtaining module configured to obtain the number of the specified dimension factors
  • the second calculation module is used to calculate the ratio range of the number of the specified dimension factors to all the dimension factors
  • a third obtaining module configured to obtain the preset warning level corresponding to the scale range
  • the adding module is used for adding the preset warning level to the warning feedback result.
  • an embodiment of the present application further provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in FIG. 4 .
  • the computer device includes a processor, memory, a network interface, and a database connected by a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the nonvolatile storage medium stores an operating system, a computer program, and a database.
  • the memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used to store all the data required for the multi-dimensional monitoring and early warning process.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a monitoring and early warning method based on multiple dimensions is realized.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Public Health (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Epidemiology (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Investigating Or Analysing Biological Materials (AREA)

Abstract

本申请涉及智慧医疗领域,揭示了基于多维度的监测预警方法,包括:从知识图谱中确定待监测对象对应的维度因子,维度因子为所述知识图谱中与待监测对象对应节点具有关联边的指定节点的名称,指定节点至少包括两个;在待监测对象对应的指定数据集中,统计各维度因子分别对应的样例数据;根据各维度因子分别对应的样例数据,分别计算各维度因子对应的预警信号集合;判断各维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;若是,则获取指定预警信号所属的指定维度因子;将指定维度因子作为指定数据集对应的预警反馈结果。应用多维度构建多个预警信息,形成由多个预警信号组成的预警信号集合,提升监测预警的灵敏度和准确度。

Description

基于多维度的监测预警方法、装置、设备及存储介质
本申请要求于2021年02月25日提交中国专利局、申请号为2021102128828,发明名称为“基于多维度的监测预警方法、装置、设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及智能客服技术领域,特别涉及一种基于多维度的监测预警方法、装置、设备及存储介质。
背景技术
近年来,在公共卫生健康领域,传染病监测预警已成为传染病防控的重点。已有的传染病监测预警的方法主要是基于传染病的确诊病例数,对传染病确诊病例数异常的增长变化进行预警。但发明人意识到基于传染病的确诊病例数的传染病预警依赖的数据源单一,导致监测预警存在监测范围局限、预警延迟、预警推送准确率低等技术问题。
技术问题
本申请的主要目的为提供基于多维度的监测预警方法,旨在解决基于传染病的确诊病例数的传染病预警依赖的数据源单一,导致监测预警存在监测范围局限、预警延迟、预警推送准确率低的技术问题。
技术解决方案
本申请提出一种基于多维度的监测预警方法,包括:
从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
若是,则获取所述指定预警信号所属的指定维度因子;
将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
优选地,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,所述在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据的步骤,包括:
获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
计算所述实测因子集与所述标准因子集的匹配度;
判断所述匹配度是否达到预设匹配度;
若是,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
本申请还提供了
一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述计算机可读指令被所述处理器执行时,使得所述处理器执行基于多维度的监测预警方法的步骤:
从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
若是,则获取所述指定预警信号所属的指定维度因子;
将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
本申请还提供了一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器执行所述基于多维度的监测预警方法的步骤:
从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
若是,则获取所述指定预警信号所属的指定维度因子;
将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
有益效果
本申请基于已有监测预警方法的缺陷,提出了一种基于多维度信息的传染病监测预警系统,应用传染病相关的多维度信息构建多个预警信息,并从多个维度分别计算预警信号,形成由多个预警信号组成的预警信号集合,以集中预警传染病的发生情况,提升监测预警的灵敏度和准确度。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分,示出了符合本申请的实施例,并与说明书一起用于解释本申请的原理。
图1本申请一实施例的基于多维度的监测预警方法流程示意图;
图2本申请一实施例的深度神经网络结构示意图;
图3本申请一实施例的基于多维度的监测预警系统流程示意图;
图4本申请一实施例的计算机设备内部结构示意图。
本申请的最佳实施方式
参照图1,本申请一实施例的基于多维度的监测预警方法,包括:
S1:从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
S2:在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
S3:根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
S4:判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
S5:若是,则获取所述指定预警信号所属的指定维度因子;
S6:将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
上述知识图谱是预先形成的疾病数据对应的图谱,一方面可由专家知识生成,另一方面可应用大数据挖掘、自然语言处理、机器学习等人工智能方法自动生成。知识图谱由节点和节点之间的关联边组成,其中每个节点表示一个实体或者概念,节点之间的关联边表示实体与实体之间,或者实体与概念之间的关联关系。本申请实施例的实体或者概念,为症状和疾病等。例如流感是一个节点(表示实体为疾病),发热也是一个节点(表示概念为症状),咳嗽是一个节点(表示概念为症状)。“流感”节点和“发热”节点之间有边连接,表示“发热”是“流感”的一个症状,具有关联关系,“流感”节点和“咳嗽”节点之间有边连接,表示“咳嗽”是“流感”的一个症状,也具有关联关系。
上述待监测对象为流感等传染病,上述维度因子包括与流感节点关联的症状、诊断、药品、检验和检查。上述指定数据集指特定地理区域的流感数据集合,上述特定地理区域包括某个医院或某个具体地区等,流感数据集合包括流感人员数量,上述样例数据指每个人员的治疗过程中的症状、诊断、药品、检验和检查等方面的数据统计。
通过在知识图谱中构建疾病与关联症状的关联关系,即找到疾病节点有症状关系边连接的症状节点,将所有连接的症状作为该疾病的相关联的症状。知识图谱中与流感相关联的概念除了症状,还包括诊断、药品、检验和检查等。症状维度因子例如包括发热、鼻出血、咽痛、浓痰、喘息等分因子,诊断维度因子例如包括急性咽炎、急性扁桃体炎、急性支气管炎、未特指的急性下呼吸道感染、链球菌性肺炎等分因子,药品维度因子例如包括乙酰水杨酸、右美沙芬、帕拉米韦、氨咖黄敏、酚氨咖敏等分因子,检验维度因子例如包括血常规、脑脊液、病毒核酸、血液生化、动脉血气分析等分因子,检查维度因子例如包括胸部CT、胸部X光等分因子,通过构建与流感相关联的概念之间的关系,组建多维度数据。即在知识图谱中,除了指定的疾病节点到有边连接的症状节点,即对疾病找到相关联的症状。类似的,还能找到指定的疾病节点到有边连接的诊断节点,即对疾病找到相关联的诊断;找到指定的疾病节点到有边连接的药品节点,即对疾病找到相关联的药品;找到指定的疾病节点到有边连接的检验节点,即对疾病找到相关联的检验;找到指定的疾病节点到有边连接的检查节点,即对疾病找到相关联的检查,通过引入传染病流感的多维度信息进行综合预警,提高了模型的灵敏率。
本申请基于已有监测预警方法的缺陷,提出了一种基于多维度信息的传染病监测预警系统,应用传染病相关的多维度信息构建多个预警信息,并从多个维度分别计算预警信号,形成由多个预警信号组成的预警信号集合,以集中预警流感等传染病的发生情况,提升监测预警的灵敏度和准确度。
进一步地,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,所述在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据的步骤S2,包括:
S21:获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
S22:从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
S23:计算所述实测因子集与所述标准因子集的匹配度;
S24:判断所述匹配度是否达到预设匹配度;
S25:若是,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
S26:根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
本申请实施例的指定传染病以流感为例,流感传染病,有以下五个数据源,包括“流感相关联的症状”,“流感相关联的诊断”,“流感相关联的药品”,“流感相关联的检验”,“流感相关联的检查”。上述每一个数据源的数据可被称为一个维度因子的数据。对于维度因子‘症状’,包括多个分因子组成的标准分因子集,比如为{症状1,症状2,症状3,...症状n}。根据患者实际电子病例中记载的症状实际出现的实际因子,形成实测因子集,比如实际因子集为{症状1,症状2,症状3}。然后通过计算实际因子占比标准分因子集的比例,确定实测因子集与标准因子集的匹配度,并确定当前患者的电子病例是否属于症状这个维度因子的样例数据。
进一步地,所述计算所述实测因子集与所述标准因子集的匹配度的步骤S23,包括:
S231:获取所述实测因子集中的指定实测因子对应的关键字,其中,所述指定实测因子为所述实测因子集中的任一因子;
S232:判断所述指定实测因子对应的关键字是否存在于所述标准因子集中;
S233:若是,则将所述指定实测因子标记为匹配因子;
S234:统计所述实测因子集中标记为匹配因子的因子数量;
S235:根据所述因子数量以及所述标准因子集中的因子总量,计算所述实测因子集与所述标准因子集的匹配度。
本申请实施例通过对每个维度因子的数据从指定数据集中分别进行统计。例如,对流感相关的症状,首先,应用深度学习技术和自然语言处理技术,比如命名实体识别技术等,从每个患者的电子病历的主诉中获取症状以及与症状相关的属性,比如症状发生的部位、持续时间、严重程度等;然后根据关键词匹配技术,将患者症状信息的关键词和流感相关联的症 状项目的关键词进行匹配,将匹配成功的患者标记为流感症状相关病例。比如将从患者的电子病例的主诉中抽取出的症状作为患者症状信息的关键词,与知识图谱中流感相关联的症状名称通过关键词识别一一进行匹配计算匹配度,计算公式为:匹配度=交集的症状/流感相关联的症状。比如流感相关联的症状:{症状1,症状2,症状3},患者1的症状:{症状1,症状2,症状3,症状4},患者2的症状:{症状3,症状4,症状5}。则对于患者1,流感相关联的症状都出现在患者1的症状列表中,所以匹配度为1=3/3。对于患者2,流感相关联的症状只有“症状3”出现在患者1的症状列表中,所以匹配度为0.33=1/3。对其它疾病的其它维度因子数据,也是类似的匹配过程,不赘述。
进一步地,所述计算所述实测因子集与所述标准因子集的匹配度的步骤S23,包括:
S2301:分别获取所述实测因子集和所述标准因子集分别对应的结构化信息;
S2302:将所述实测因子集和所述标准因子集分别对应的结构化信息转换成向量;
S2303:将所述实测因子集和所述标准因子集分别对应的向量,输入深度学习网络,得到所述实测因子集对应的第一低维向量以及所述标准因子集对应的第二低维向量;
S2304:计算所述第一低维向量和所述第二低维向量的相似度;
S2305:将所述第一低维向量和所述第二低维向量的相似度,作为所述实测因子集与所述标准因子集的匹配度。
本申请实施例中,为了避免症状存在多个同义词造成关键词,导致匹配不成功从而带来误差。对没有匹配成功的患者,继续应用基于深度学习网络的表示学习模型,将患者的结构化信息表示的向量和各个维度因子数据的定义向量进行模糊匹配,将匹配得分较高的患者标记为流感症状相关病例。
上述结果化信息指对于症状这一维度因子数据而言,从每个患者的电子病历的主诉中获取的症状以及与症状相关的属性的文字描述。通过将结果化信息构造multi-hot encoding向量,向量维度为所有与症状相关的属性的数目,每一个维度因子对应一个症状的属性,赋值为1表示出现这个症状,赋值为0表示没有这个症状。
如图2所示,通过构建有三层全连接隐藏层的自动编码器结构的深度学习网络,通过无监督学习方式训练网络,得到流感相关联的症状的表示学习模型。表示学习模型的输入为上述的multi-hot encoding向量,输出也是multi-hot encoding向量。输入和输出的multi-hot encoding向量是高维稀疏的向量,通过三层全连接隐藏层的自动编码器结构的网络,将高维稀疏的向量在中间隐藏层转换为低维密实的向量,并将中间隐藏层输出的低维密实的向量作为流感相关联的各个维度因子数据的定义向量B。对患者的症状的症状通过上述表示学习模型的中间隐藏层输出的向量A,通过公式(1)计算相似度similarity,即得到匹配度。将匹配度超过阈值的患者电子病例标记为流感症状相关病例,一般可将阈值设为0.8。公式(1)的表达式为:
Figure PCTCN2021084538-appb-000001
其中,n表示向量A和B的向量维度,i表示第i个向量维度。
进一步地,所述根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合的步骤S3,包括:
S31:获取预设长度的时间单位量,以及所述指定数据集对应的统计时间段;
S32:确定所述统计时间段对应的开始时间和结束时间;
S33:以所述开始时间为起点按照自然时序,按照所述时间单位量将所述统计时间段依次截分成多个子时间段;
S34:获取各所述子时间段内第二维度因子分别对应的子样例数据,其中,所述第二维度因子为所有维度因子中的任一个,所述第二维度因子的样例数据由各所述子时间段分别对应的子样例数据组合得到;
S35:根据各所述子时间段内分别对应的所述第二维度因子的子样例数据,计算各所述子时间段分别对应的预警信号;
S36:将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号 集合;
S37:根据所述第二维度因子对应的预警信号集合的计算方式,计算各所述维度因子对应的预警信号集合。
本申请实施例通过对每个维度因子的数据分别进行流感相关病例数的统计,并按照时间单位进行统计,对于流感相关的诊断项目,整理按照时间单位对应的流感诊断相关病例数;对于流感相关的药品项目,整理按照时间单位对应的流感药品相关病例数;对于流感相关的检验检查项目,整理按照时间单位对应的流感检验检查相关病例数等等,最后形成每个时间单位(例如天、小时等)的流感症状相关病例数。一一对应整理各维度因子中时间单位对应的数值,例如对于流感的症状,时间单位为天,得到整理后的数据如下表1:
表1
时间 流感症状相关病例数
…… ……
2020年1月1日 10
2020年1月2日 26
…… ……
对于流感的药品,时间单位同样为天,得到整理后的数据如表2:
表2
时间 流感药品相关病例数
…… ……
2020年1月1日 156
2020年1月2日 234
…… ……
基于时间和指标维度构造预警信号,并计算预警信号值。预警信号有多种,可以通过各种方式来构建,比如包括以天为时间单位的周环比增长率、周同比增长比率等。在这里举出以下四个例子来具体描述预警信号值的计算过程:周环比增长率:根据公式“(最近七天的值之和–再上一个七天的值之和)/再上一个七天的值之和”计算得到;周同比增长比率:根据公式“(最近七天的值之和–去年同一个时期七天的值之和)/去年同一个时期七天的值之和”计算得到,例如今年是2021年1月1日到2021年1月7日(最近七天的值之和),对于“去年同一个时期七天的值之和”是指2020年1月1日到2020年1月7日;最近三天增长率:对最近三天的日增长率求平均;最近七天增长率:对最近七天的日增长率求平均,等等。对于每个维度因子都分别计算预警信号,例如对于流感相关联的症状,根据某个时间单位内的症状的值,计算症状周环比增长率,症状周同比增长率等等;对于流感相关联的药品,根据某个时间单位内的药品的值,计算药品周环比增长率,药品周同比增长率等等。
计算预警信号值时,需要计算历史数据,才能得到周环比增长率、周同比增长率等。以症状维度因子为例,对于流感症状相关病例数的历史数据(一般为过去五年的历史数据)对多个预警信号分别计算历史值。对每个预警信号,按照基于其历史值的百分比确定阈值,比如取历史值的95%作为预警信号的阈值。例如计算症状维度因子中一个预警信号“周环比增长率”的预警信号值,用去年(即2020年)的数据计算预警信号值,“2020年的数据”就是“历史数据”。根据2020年的数据得到基于2020年数据的症状周环比增长率,以天为时间单位计算,得到365个症状周环比增长率的值,对365个症状周环比增长率的值按照大小从小到大排序,取95%位置上的值作为症状的周环比增长率的阈值。类似的,在症状维度因子下,对于其他预警信号,还可得到“症状的周同比增长率的阈值”,“症状的最近三天增长率的阈值”等等。除此之外,对于药品等其他维度因子,按照上述的步骤,可得到各维度因子分别对应的预警信号的阈值,比如“药品的周环比增长率的阈值”,“药品的最近三天增长率的阈值”等等。
进一步地,所述将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合的步骤S36之后,包括:
S361:在所述第二维度因子对应的预警信号集合中,两两计算预警信号之间的相关性系数;
S362:判断第一预警信号和第二预警信号的相关性系数是否大于预设相关性阈值,其中, 所述第一预警信号和第二预警信号为所述第二维度因子对应的预警信号集合中的任意两个预警信号;
S363:若是,则对所述第一预警信号和第二预警信号进行取舍保留,删除所述第一预警信号或第二预警信号;
S364:将经过取舍保留过程的所述第二维度因子的预警信号集合,作为所述第二维度因子对应的新预警信号集合。
本申请实施例中,为了从不同数据分析角度捕获预警信号,设计从时间和数据指标构造多个预警信号对流感进行预警。其中,基于时间因素可分为天、周、月等不同的时间单位,数据指标可包括同比、环比、历史百分位、增长率等。从时间和数据指标两个因素构建的多个预警信号之间可能存在较强的相关性,不仅影响计算效率且无法更简洁地凸显不同数据分析角度。本申请中用多个预警信号进行综合预警时,如果预警信号之间的存在较强的相关性,为了避免预警信号之间相关性较强对综合预警结果的影响,对多个预警信号进行了相关性筛选。具体的步骤是:首先应用相关性检验方法,例如通过皮尔森相关系数对任意两个预警信号之间计算相关性系数;然后基于相关性系数阈值进行预警信号筛选,比如相关性系数阈值设为0.3,即0.3及以下表示无相关性或相关性较弱。如果两个预警信号的相关性系数超过该阈值,表示两个预警信号的变化是类似的,所以要将两个相关性较强的预警信号进行取舍保留,则只保留其中一个预警信号,移除另一个预警信号,比如删除预警信号值较小者,以确保预警灵敏度;最后得到的预警信号之间的相关性都不超过该阈值,以凸显不同数据分析角度,用经过取舍保留过程后的预警信号进行综合预警。
本申请中各个预警信号的权重是相同的,目标是要各个预警信号从不同数据角度表征传染病变化趋势中的异常变化,从而更好的进行综合预警,减少误报。在进行综合预警时,用于综合预警的预警信号集合中只要有任一个预警信号超过阈值,则触发预警。
进一步地,所述将所述指定维度因子作为所述指定数据集对应的预警反馈结果的步骤S6之后,包括:
S61:获取所述指定维度因子的数量;
S62:计算所述指定维度因子的数量占比所有维度因子的比例范围;
S63:获取所述比例范围对应的预设预警等级;
S64:将所述预设预警等级添加至所述预警反馈结果中。
本申请实施例的预警是两层预警的层级式结构,首先完成第一层预警的计算,根据每个维度因子的数据得到综合预警信号结果,确定触发预警的维度因子;然后根据第一层预警的结果,再进行第二层预警的计算,即计算触发预警的维度因子数量占比总维度因子数量的占比,得到第二层预警的预警结果,确定预警的严重等级。
第一层预警以流感症状这一维度因子为例,对于流感症状相关病例数的历史数据对多个预警信号分别计算历史值。对每个预警信号按照基于其历史值的百分比确定阈值。在对每个预警信号确定阈值后,综合多个预警信号的值进行预警。如果多个预警信号的值都低于它们各自的阈值,则不进行预警;如果有任何一个预警信号的值超过其阈值,则进行预警。
第二层预警是对每个维度因子的数据进行综合分析的预警结果。比如在判定症状这一维度因子触发预警后,再判断流感相关联的诊断这一维度因子是否预警;和/或判断流感相关联的药品数据源是否预警;和/或判断流感相关联的检验检查数据源是否预警等等,按照如下规则进行预警风险等级预警,并给出预警信息。比如所有维度因子对应的预警信号都没有触发预警,则显示低风险;出现一个维度因子的预警信号触发预警,则为中风险,并将这个触发预警的维度因子对应数据详细情况作为预警信息显示;出现两个及以上维度因子的预警信号触发预警,则为高风险,并将所有触发预警的维度因子对应数据详细情况作为预警信息显示。例如在症状这一数据源中,所有症状的预警信号都没有超过阈值,那么症状这一维度因子没有预警信息;在药品这一维度因子中,有药品的周环比增长率这一预警信号超过其阈值,那么药品这一维度因子被标记为预警,并将作为第二层预警的信息被处理。
本申请通过两层预警机制构建了基于多维度因子数据的传染病预警方案,通过设计层级预警机制以及丰富的流感相关联的预警信号,可以提升预警的灵敏度。并且通过对多种预警信号进行相关性筛选,用相关性较低的预警信号进行预警,从不同数据分析角度捕获异常,进一步提升预警的精准度。例如,对流感这一传染病进行预警,症状这一维度因子触发预警,而在症状维度因子中,实际是症状对应的两个预警信号超过了其阈值,例如周环比增长率的 预警信号值为200%,超过周环比增长率的预警信号阈值20%,而且最近三天增长率的预警信号值为300%,超过最近三天增长率的预警信号阈值50%,所以显示的预警详情为:预警中风险,流感相关联的症状维度因子触发预警,具体包括流感相关联的症状的周环比增长率为200%,超过阈值;流感相关联的症状的最近三天增长率为300%,超过阈值。
参照图3,本申请一实施例的基于多维度的监测预警装置,包括:
确定模块1,用于从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
统计模块2,用于在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
第一计算模块3,用于根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
判断模块4,用于判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
第一获取模块5,用于若存在超过阈值的指定预警信号,则获取所述指定预警信号所属的指定维度因子;
作为模块6,用于将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
本申请实施例的相关解释适用对应方法部分的解释,不赘述。
进一步地,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,统计模块2,包括:
第一获取单元,用于获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
抽取单元,用于从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
第一计算单元,用于计算所述实测因子集与所述标准因子集的匹配度;
判断单元,用于判断所述匹配度是否达到预设匹配度;
归纳单元,用于若是达到预设匹配度,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
筛选单元,用于根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
进一步地,所述计算单元,包括:
第一获取子单元,用于获取所述实测因子集中的指定实测因子对应的关键字,其中,所述指定实测因子为所述实测因子集中的任一因子;
判断子单元,用于判断所述指定实测因子对应的关键字是否存在于所述标准因子集中;
标记子单元,用于若是存在于所述标准因子集中,则将所述指定实测因子标记为匹配因子;
统计子单元,用于统计所述实测因子集中标记为匹配因子的因子数量;
第一计算子单元,用于根据所述因子数量以及所述标准因子集中的因子总量,计算所述实测因子集与所述标准因子集的匹配度。
进一步地,所述计算单元,包括:
第二获取子单元,用于分别获取所述实测因子集和所述标准因子集分别对应的结构化信息;
转换子单元,用于将所述实测因子集和所述标准因子集分别对应的结构化信息转换成向量;
输入子单元,用于将所述实测因子集和所述标准因子集分别对应的向量,输入深度学习网络,得到所述实测因子集对应的第一低维向量以及所述标准因子集对应的第二低维向量;
第二计算子单元,用于计算所述第一低维向量和所述第二低维向量的相似度;
作为子单元,用于将所述第一低维向量和所述第二低维向量的相似度,作为所述实测因子集与所述标准因子集的匹配度。
进一步地,计算模块3,包括:
第二获取单元,用于获取预设长度的时间单位量,以及所述指定数据集对应的统计时间段;
确定单元,用于确定所述统计时间段对应的开始时间和结束时间;
截分单元,用于以所述开始时间为起点按照自然时序,按照所述时间单位量将所述统计时间段依次截分成多个子时间段;
第三获取单元,用于获取各所述子时间段内第二维度因子分别对应的子样例数据,其中,所述第二维度因子为所有维度因子中的任一个,所述第二维度因子的样例数据由各所述子时间段分别对应的子样例数据组合得到;
第二计算单元,用于根据各所述子时间段内分别对应的所述第二维度因子的子样例数据,计算各所述子时间段分别对应的预警信号;
组成单元,用于将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合;
第三计算单元,用于根据所述第二维度因子对应的预警信号集合的计算方式,计算各所述维度因子对应的预警信号集合。
进一步地,计算模块3,包括:
第四计算单元,用于在所述第二维度因子对应的预警信号集合中,两两计算预警信号之间的相关性系数;
判断单元,用于判断第一预警信号和第二预警信号的相关性系数是否大于预设相关性阈值,其中,所述第一预警信号和第二预警信号为所述第二维度因子对应的预警信号集合中的任意两个预警信号;
取舍单元,用于若大于预设相关性阈值,则对所述第一预警信号和第二预警信号进行取舍保留,删除所述第一预警信号或第二预警信号;
作为单元,用于将经过取舍保留过程的所述第二维度因子的预警信号集合,作为所述第二维度因子对应的新预警信号集合。
进一步地,基于多维度的监测预警装置,包括:
第二获取模块,用于获取所述指定维度因子的数量;
第二计算模块,用于计算所述指定维度因子的数量占比所有维度因子的比例范围;
第三获取模块,用于获取所述比例范围对应的预设预警等级;
添加模块,用于将所述预设预警等级添加至所述预警反馈结果中。
参照图4,本申请实施例中还提供一种计算机设备,该计算机设备可以是服务器,其内部结构可以如图4所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中,该计算机设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储基于多维度的监测预警过程需要的所有数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现基于多维度的监测预警方法。

Claims (20)

  1. 一种基于多维度的监测预警方法,其中,包括:
    从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
    在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
    根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
    判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
    若是,则获取所述指定预警信号所属的指定维度因子;
    将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
  2. 根据权利要求1所述的基于多维度的监测预警方法,其中,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,所述在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据的步骤,包括:
    获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
    从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
    计算所述实测因子集与所述标准因子集的匹配度;
    判断所述匹配度是否达到预设匹配度;
    若是,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
    根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
  3. 根据权利要求2所述的基于多维度的监测预警方法,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    获取所述实测因子集中的指定实测因子对应的关键字,其中,所述指定实测因子为所述实测因子集中的任一因子;
    判断所述指定实测因子对应的关键字是否存在于所述标准因子集中;
    若是,则将所述指定实测因子标记为匹配因子;
    统计所述实测因子集中标记为匹配因子的因子数量;
    根据所述因子数量以及所述标准因子集中的因子总量,计算所述实测因子集与所述标准因子集的匹配度。
  4. 根据权利要求2所述的基于多维度的监测预警方法,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    分别获取所述实测因子集和所述标准因子集分别对应的结构化信息;
    将所述实测因子集和所述标准因子集分别对应的结构化信息转换成向量;
    将所述实测因子集和所述标准因子集分别对应的向量,输入深度学习网络,得到所述实测因子集对应的第一低维向量以及所述标准因子集对应的第二低维向量;
    计算所述第一低维向量和所述第二低维向量的相似度;
    将所述第一低维向量和所述第二低维向量的相似度,作为所述实测因子集与所述标准因子集的匹配度。
  5. 根据权利要求1所述的基于多维度的监测预警方法,其中,所述根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合的步骤,包括:
    获取预设长度的时间单位量,以及所述指定数据集对应的统计时间段;
    确定所述统计时间段对应的开始时间和结束时间;
    以所述开始时间为起点按照自然时序,按照所述时间单位量将所述统计时间段依次截分成多个子时间段;
    获取各所述子时间段内第二维度因子分别对应的子样例数据,其中,所述第二维度因子为所有维度因子中的任一个,所述第二维度因子的样例数据由各所述子时间段分别对应的子样例数据组合得到;
    根据各所述子时间段内分别对应的所述第二维度因子的子样例数据,计算各所述子时间段分别对应的预警信号;
    将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合;
    根据所述第二维度因子对应的预警信号集合的计算方式,计算各所述维度因子对应的预警信号集合。
  6. 根据权利要求5所述的基于多维度的监测预警方法,其中,所述将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合的步骤之后,包括:
    在所述第二维度因子对应的预警信号集合中,两两计算预警信号之间的相关性系数;
    判断第一预警信号和第二预警信号的相关性系数是否大于预设相关性阈值,其中,所述第一预警信号和第二预警信号为所述第二维度因子对应的预警信号集合中的任意两个预警信号;
    若是,则对所述第一预警信号和第二预警信号进行取舍保留,删除所述第一预警信号或第二预警信号;
    将经过取舍保留过程的所述第二维度因子的预警信号集合,作为所述第二维度因子对应的新预警信号集合。
  7. 根据权利要求1所述的基于多维度的监测预警方法,其中,所述将所述指定维度因子作为所述指定数据集对应的预警反馈结果的步骤之后,包括:
    获取所述指定维度因子的数量;
    计算所述指定维度因子的数量占比所有维度因子的比例范围;
    获取所述比例范围对应的预设预警等级;
    将所述预设预警等级添加至所述预警反馈结果中。
  8. 一种基于多维度的监测预警装置,其中,包括:
    确定模块,用于从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
    统计模块,用于在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
    第一计算模块,用于根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
    判断模块,用于判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
    第一获取模块,用于若存在超过阈值的指定预警信号,则获取所述指定预警信号所属的指定维度因子;
    作为模块,用于将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
  9. 一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述计算机可读指令被所述处理器执行时,使得所述处理器执行基于多维度的监测预警方法的步骤:
    从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
    在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
    根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
    判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
    若是,则获取所述指定预警信号所属的指定维度因子;
    将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
  10. 根据权利要求9所述的计算机设备,其中,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,所述在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据的步骤,包括:
    获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
    从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
    计算所述实测因子集与所述标准因子集的匹配度;
    判断所述匹配度是否达到预设匹配度;
    若是,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
    根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
  11. 根据权利要求10所述的计算机设备,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    获取所述实测因子集中的指定实测因子对应的关键字,其中,所述指定实测因子为所述实测因子集中的任一因子;
    判断所述指定实测因子对应的关键字是否存在于所述标准因子集中;
    若是,则将所述指定实测因子标记为匹配因子;
    统计所述实测因子集中标记为匹配因子的因子数量;
    根据所述因子数量以及所述标准因子集中的因子总量,计算所述实测因子集与所述标准因子集的匹配度。
  12. 根据权利要求10所述的计算机设备,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    分别获取所述实测因子集和所述标准因子集分别对应的结构化信息;
    将所述实测因子集和所述标准因子集分别对应的结构化信息转换成向量;
    将所述实测因子集和所述标准因子集分别对应的向量,输入深度学习网络,得到所述实测因子集对应的第一低维向量以及所述标准因子集对应的第二低维向量;
    计算所述第一低维向量和所述第二低维向量的相似度;
    将所述第一低维向量和所述第二低维向量的相似度,作为所述实测因子集与所述标准因子集的匹配度。
  13. 根据权利要求9所述的计算机设备,其中,所述根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合的步骤,包括:
    获取预设长度的时间单位量,以及所述指定数据集对应的统计时间段;
    确定所述统计时间段对应的开始时间和结束时间;
    以所述开始时间为起点按照自然时序,按照所述时间单位量将所述统计时间段依次截分成多个子时间段;
    获取各所述子时间段内第二维度因子分别对应的子样例数据,其中,所述第二维度因子为所有维度因子中的任一个,所述第二维度因子的样例数据由各所述子时间段分别对应的子样例数据组合得到;
    根据各所述子时间段内分别对应的所述第二维度因子的子样例数据,计算各所述子时间段分别对应的预警信号;
    将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合;
    根据所述第二维度因子对应的预警信号集合的计算方式,计算各所述维度因子对应的预警信号集合。
  14. 根据权利要求13所述的计算机设备,其中,所述将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合的步骤之后,包括:
    在所述第二维度因子对应的预警信号集合中,两两计算预警信号之间的相关性系数;
    判断第一预警信号和第二预警信号的相关性系数是否大于预设相关性阈值,其中,所述第一预警信号和第二预警信号为所述第二维度因子对应的预警信号集合中的任意两个预警信号;
    若是,则对所述第一预警信号和第二预警信号进行取舍保留,删除所述第一预警信号或第二预警信号;
    将经过取舍保留过程的所述第二维度因子的预警信号集合,作为所述第二维度因子对应的新预警信号集合。
  15. 根据权利要求9所述的计算机设备,其中,所述将所述指定维度因子作为所述指定数据集对应的预警反馈结果的步骤之后,包括:
    获取所述指定维度因子的数量;
    计算所述指定维度因子的数量占比所有维度因子的比例范围;
    获取所述比例范围对应的预设预警等级;
    将所述预设预警等级添加至所述预警反馈结果中。
  16. 一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器执行所述基于多维度的监测预警方法的步骤:
    从知识图谱中确定待监测对象对应的维度因子,其中,所述维度因子为所述知识图谱中与所述待监测对象对应节点具有关联边的指定节点的名称,所述指定节点至少包括两个;
    在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据;
    根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合;
    判断各所述维度因子分别对应的预警信号集合中是否存在超过阈值的指定预警信号;
    若是,则获取所述指定预警信号所属的指定维度因子;
    将所述指定维度因子作为所述指定数据集对应的预警反馈结果。
  17. 根据权利要求16所述的存储介质,其中,所述指定数据集包括所有监测用户对应的电子病例集,所述待监测对象包括指定传染病,所述在所述待监测对象对应的指定数据集中,统计各所述维度因子分别对应的样例数据的步骤,包括:
    获取所述指定传染病的第一维度因子对应的标准分因子集,其中,所述标准分因子集包括多个分因子,所述第一维度因子为所有维度因子中的任一个;
    从指定用户对应的指定电子病例中,抽取与所述第一维度因子对应的实际因子,形成实测因子集,其中,所述指定用户为所有监测用户中的任一个,所述指定电子病例为所述电子病例集中的任一电子病例;
    计算所述实测因子集与所述标准因子集的匹配度;
    判断所述匹配度是否达到预设匹配度;
    若是,则将所述指定电子病例归纳至所述第一维度因子对应的第一样例集中;
    根据所述第一维度因子对应的第一样例集的形成方式,从所述电子病例集中筛选各所述维度因子分别对应的样例集,作为各所述维度因子分别对应的样例数据。
  18. 根据权利要求17所述的存储介质,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    获取所述实测因子集中的指定实测因子对应的关键字,其中,所述指定实测因子为所述实测因子集中的任一因子;
    判断所述指定实测因子对应的关键字是否存在于所述标准因子集中;
    若是,则将所述指定实测因子标记为匹配因子;
    统计所述实测因子集中标记为匹配因子的因子数量;
    根据所述因子数量以及所述标准因子集中的因子总量,计算所述实测因子集与所述标准因子集的匹配度。
  19. 根据权利要求17所述的存储介质,其中,所述计算所述实测因子集与所述标准因子集的匹配度的步骤,包括:
    分别获取所述实测因子集和所述标准因子集分别对应的结构化信息;
    将所述实测因子集和所述标准因子集分别对应的结构化信息转换成向量;
    将所述实测因子集和所述标准因子集分别对应的向量,输入深度学习网络,得到所述实测因子集对应的第一低维向量以及所述标准因子集对应的第二低维向量;
    计算所述第一低维向量和所述第二低维向量的相似度;
    将所述第一低维向量和所述第二低维向量的相似度,作为所述实测因子集与所述标准因子集的匹配度。
  20. 根据权利要求16所述的存储介质,其中,所述根据各所述维度因子分别对应的样例数据,分别计算各所述维度因子对应的预警信号集合的步骤,包括:
    获取预设长度的时间单位量,以及所述指定数据集对应的统计时间段;
    确定所述统计时间段对应的开始时间和结束时间;
    以所述开始时间为起点按照自然时序,按照所述时间单位量将所述统计时间段依次截分成多个子时间段;
    获取各所述子时间段内第二维度因子分别对应的子样例数据,其中,所述第二维度因子为所有维度因子中的任一个,所述第二维度因子的样例数据由各所述子时间段分别对应的子样例数据组合得到;
    根据各所述子时间段内分别对应的所述第二维度因子的子样例数据,计算各所述子时间段分别对应的预警信号;
    将各所述子时间段分别对应的预警信号,组成所述第二维度因子对应的预警信号集合;
    根据所述第二维度因子对应的预警信号集合的计算方式,计算各所述维度因子对应的预警信号集合。
PCT/CN2021/084538 2021-02-25 2021-03-31 基于多维度的监测预警方法、装置、设备及存储介质 WO2022178947A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110212882.8 2021-02-25
CN202110212882.8A CN112951441B (zh) 2021-02-25 2021-02-25 基于多维度的监测预警方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2022178947A1 true WO2022178947A1 (zh) 2022-09-01

Family

ID=76246264

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/084538 WO2022178947A1 (zh) 2021-02-25 2021-03-31 基于多维度的监测预警方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN112951441B (zh)
WO (1) WO2022178947A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116389223A (zh) * 2023-04-26 2023-07-04 福芯高照(上海)科技有限公司 一种基于大数据的人工智能可视化预警系统及方法
CN117405177A (zh) * 2023-12-15 2024-01-16 国网山东省电力公司济南供电公司 电缆隧道有害气体泄漏预警方法、系统、设备及介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109448859A (zh) * 2018-11-09 2019-03-08 贵州医渡云技术有限公司 数据处理方法及装置、电子设备、存储介质
US20200253547A1 (en) * 2017-10-31 2020-08-13 Apple Inc. Monitoring System for Assessing Control of a Disease State
CN112037925A (zh) * 2020-07-29 2020-12-04 郑州大学第一附属医院 一种基于lstm算法的新发重大传染病预警方法
CN112117010A (zh) * 2020-07-13 2020-12-22 北京大瑞集思技术有限公司 一种传染病智能预警系统及管理平台

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114141378A (zh) * 2017-01-19 2022-03-04 京东方科技集团股份有限公司 数据分析方法和设备
CN106897273B (zh) * 2017-04-12 2018-02-06 福州大学 一种基于知识图谱的网络安全动态预警方法
CN109711155A (zh) * 2018-11-13 2019-05-03 同盾控股有限公司 一种预警确定方法和装置
CN109767842B (zh) * 2018-12-13 2023-08-22 平安科技(深圳)有限公司 一种疾病预警方法、疾病预警装置及计算机可读存储介质
CN110853744A (zh) * 2019-09-04 2020-02-28 太原市中心医院 一种大数据下医院质控管理系统
CN111785374A (zh) * 2020-06-15 2020-10-16 山东省玖玖医养健康产业有限公司 一种基于大数据的健康状况分析预测方法及系统
CN111916215B (zh) * 2020-06-17 2022-04-29 扬州大学附属医院 一种群集性/慢性疾病早期自动报警、初步特征分析与风险评估系统
CN111785392B (zh) * 2020-07-01 2024-02-09 医渡云(北京)技术有限公司 人口数量的预警方法、装置、电子设备及计算机可读介质
CN111967761B (zh) * 2020-08-14 2024-04-02 国网数字科技控股有限公司 一种基于知识图谱的监控预警方法、装置及电子设备
CN112100498A (zh) * 2020-09-16 2020-12-18 北京百度网讯科技有限公司 疾病舆情的监控方法、装置
CN112185566B (zh) * 2020-10-14 2021-08-13 上海玺翎智能科技有限公司 一种基于机器学习预测预警感染性疾病就医人数突增的方法
CN112185586A (zh) * 2020-11-17 2021-01-05 北京嘉和海森健康科技有限公司 一种传染病监测预警方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200253547A1 (en) * 2017-10-31 2020-08-13 Apple Inc. Monitoring System for Assessing Control of a Disease State
CN109448859A (zh) * 2018-11-09 2019-03-08 贵州医渡云技术有限公司 数据处理方法及装置、电子设备、存储介质
CN112117010A (zh) * 2020-07-13 2020-12-22 北京大瑞集思技术有限公司 一种传染病智能预警系统及管理平台
CN112037925A (zh) * 2020-07-29 2020-12-04 郑州大学第一附属医院 一种基于lstm算法的新发重大传染病预警方法

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116389223A (zh) * 2023-04-26 2023-07-04 福芯高照(上海)科技有限公司 一种基于大数据的人工智能可视化预警系统及方法
CN116389223B (zh) * 2023-04-26 2023-12-22 郑州数智科技集团有限公司 一种基于大数据的人工智能可视化预警系统及方法
CN117405177A (zh) * 2023-12-15 2024-01-16 国网山东省电力公司济南供电公司 电缆隧道有害气体泄漏预警方法、系统、设备及介质
CN117405177B (zh) * 2023-12-15 2024-04-26 国网山东省电力公司济南供电公司 电缆隧道有害气体泄漏预警方法、系统、设备及介质

Also Published As

Publication number Publication date
CN112951441A (zh) 2021-06-11
CN112951441B (zh) 2023-05-30

Similar Documents

Publication Publication Date Title
CN109935336B (zh) 一种儿童呼吸科疾病的智能辅助诊断系统
CN113871003B (zh) 一种基于含因果性医学知识图谱的疾病辅助鉴别诊断系统
CN111292853B (zh) 基于多参数的心血管疾病风险预测网络模型及其构建方法
CN104766259A (zh) 一种基于单病种模型的医疗临床质量监测与评价系统
WO2022178947A1 (zh) 基于多维度的监测预警方法、装置、设备及存储介质
CN110299209B (zh) 相似病历查找方法、装置、设备及可读存储介质
CN109378066A (zh) 一种基于特征向量实现疾病预测的控制方法及控制装置
CN109801687B (zh) 一种面向健康领域的因果关系知识库的构建方法和系统
CN112435756B (zh) 基于多数据集差异互证的肠道菌群关联疾病风险预测系统
WO2024008043A1 (zh) 一种基于因果关系挖掘的临床数据自动化生成方法及系统
CN109508336A (zh) 基于医疗资源事实库的检索方法、存储介质及计算机设备
US20230401466A1 (en) Method for temporal knowledge graph reasoning based on distributed attention
CN114091677A (zh) 一种因果关系判定系统
CN113889265A (zh) 一种基于表型可视化的罕见病辅助推理的方法及系统
CN112820400B (zh) 基于医疗知识图谱知识推理的疾病诊断装置、设备
Spreco et al. Influenza detection and prediction algorithms: comparative accuracy trial in Östergötland county, Sweden, 2008–2012
CN114191665A (zh) 机械通气过程中人机异步现象的分类方法和分类装置
CN116864062B (zh) 一种基于互联网的健康体检报告数据分析管理系统
Qiu et al. An online test for goodness-of-fit in logistic regression model
WO2022141929A1 (zh) 一种健康自测系统、服务器及健康检测系统
CN113822420B (zh) 基于容积二氧化碳图的死腔分数的模型建立方法及系统
CN114300158A (zh) 一种药品不良反应的识别方法及装置
CN110010231A (zh) 一种数据处理系统及计算机可读存储介质
Zhu et al. Research on data mining of college students’ physical health for physical education reform
CN115312196A (zh) 一种新型筛选压力性损伤危险因素的模型构建评价方法及其应用

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21927381

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21927381

Country of ref document: EP

Kind code of ref document: A1