WO2021135429A1 - 基于知识图谱的健康信息推荐方法、装置、设备及介质 - Google Patents

基于知识图谱的健康信息推荐方法、装置、设备及介质 Download PDF

Info

Publication number
WO2021135429A1
WO2021135429A1 PCT/CN2020/117090 CN2020117090W WO2021135429A1 WO 2021135429 A1 WO2021135429 A1 WO 2021135429A1 CN 2020117090 W CN2020117090 W CN 2020117090W WO 2021135429 A1 WO2021135429 A1 WO 2021135429A1
Authority
WO
WIPO (PCT)
Prior art keywords
sample
preset
health
health information
information
Prior art date
Application number
PCT/CN2020/117090
Other languages
English (en)
French (fr)
Inventor
左磊
赵惟
徐卓扬
孙行智
胡岗
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 WO2021135429A1 publication Critical patent/WO2021135429A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/60—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to nutrition control, e.g. diets

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device, device, and medium for recommending health information based on a knowledge graph.
  • carbohydrate intake needs to be controlled so as not to increase blood sugar, and appropriate exercise is required; however, without the guidance of professional doctors, patients should make appropriate judgments on carbohydrate intake and exercise
  • the accuracy is low, and its practicability is also uneven, and it is difficult to self-regulate blood sugar levels; under the condition of blindly using the wrong adjustment method, it may even lead to aggravation of the disease.
  • the embodiments of the present application provide a method, device, device, and medium for recommending health information based on a knowledge graph to solve the problems of low practicability and low accuracy in judging health information such as diet and exercise.
  • a method for recommending health information based on knowledge graphs including:
  • the target object is a user requesting to push health information
  • the target characteristic information refers to the individual characteristic information of the target object
  • the sample triad is composed of sample classification labels, sample health information, and The recommended value association structure associated with the sample health information and the sample classification label;
  • a health information recommendation device based on a knowledge graph including:
  • the characteristic information acquisition module is configured to acquire target characteristic information of a target object; the target object is a user requesting to push health information; the target characteristic information refers to the individual characteristic information of the target object;
  • the classification label determination module is configured to input the target feature information into a preset health feature similarity model to obtain a health classification label corresponding to the target feature information;
  • the sample triple acquisition module is used to acquire all sample triples with the same sample classification label as the health classification label from a preset health recommendation database constructed based on the knowledge graph; the sample triple is It is composed of sample classification label, sample health information, and recommended value associations associated with both sample health information and sample classification label;
  • the health information recommendation module is configured to extract sample health information and recommended values associated with the sample health information from all the acquired triples, and push the associated health information to the mobile terminal of the target object according to the recommended values The sample health information.
  • a computer device includes a memory, a processor, and computer-readable instructions that are stored in the memory and can run on the processor, and the processor implements the following steps when the processor executes the computer-readable instructions:
  • the target object is a user requesting to push health information
  • the target characteristic information refers to the individual characteristic information of the target object
  • the sample triad is composed of sample classification labels, sample health information, and The recommended value association structure associated with the sample health information and the sample classification label;
  • One or more readable storage media storing computer readable instructions, when the computer readable instructions are executed by one or more processors, the one or more processors execute the following steps:
  • the target object is a user requesting to push health information
  • the target characteristic information refers to the individual characteristic information of the target object
  • the sample triad is composed of sample classification labels, sample health information, and The recommended value association structure associated with the sample health information and the sample classification label;
  • the above-mentioned method, device, device and medium for recommending health information based on the knowledge graph by inputting the target characteristic information of the target object into the preset health characteristic similarity model, the health classification label corresponding to the target characteristic information is obtained;
  • the preset health recommendation database constructed by the Knowledge Graph all sample triads with the same sample classification label as the health classification label are obtained; the sample health information and the health information associated with the sample health information are extracted from all the obtained sample triads
  • the recommended value based on the recommended value, pushes the sample health information associated with the target object to the target object.
  • this application recommends matching sample health information to the target object to improve the accuracy of the sample health information recommendation; at the same time, the push is determined according to the recommendation value and user recommendation needs Sample health information can provide users with more options, and these options have corresponding recommended values, so as to improve the user's choice under guaranteed conditions.
  • This application belongs to the field of smart medical care and relates to digital medical care related to health management. Through this application, the construction of smart cities can be promoted.
  • FIG. 1 is a schematic diagram of an application environment of a health information recommendation method based on a knowledge graph in an embodiment of the present application
  • FIG. 2 is a flowchart of a method for recommending health information based on a knowledge graph in an embodiment of the present application
  • FIG. 3 is another flowchart of a method for recommending health information based on a knowledge graph in an embodiment of the present application
  • step S24 is a flowchart of step S24 in the method for recommending health information based on the knowledge graph in an embodiment of the present application
  • step S24 is another flowchart of step S24 in the method for recommending health information based on the knowledge graph in an embodiment of the present application
  • Fig. 6 is a functional block diagram of a health information recommendation device based on a knowledge graph in an embodiment of the present application
  • FIG. 7 is another principle block diagram of a health information recommendation device based on a knowledge graph in an embodiment of the present application.
  • FIG. 8 is a functional block diagram of a database construction module in a health information recommendation device based on a knowledge graph in an embodiment of the present application
  • FIG. 9 is another principle block diagram of the database construction module in the health information recommendation device based on the knowledge graph in an embodiment of the present application.
  • Fig. 10 is a schematic diagram of a computer device in an embodiment of the present application.
  • the health information recommendation method based on the knowledge graph provided by the embodiment of the present application can be applied to the application environment as shown in FIG. 1.
  • the health information recommendation method based on the knowledge graph is applied to the health information recommendation system based on the knowledge graph.
  • the health information recommendation system based on the knowledge graph includes the client and the server as shown in FIG.
  • the network communicates to solve the problems of low practicability and low accuracy in the judgment of health information such as diet and exercise.
  • the client is also called the client, which refers to the program that corresponds to the server and provides local services to the client.
  • the client can be installed on, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.
  • the server can be implemented as an independent server or a server cluster composed of multiple servers.
  • a method for recommending health information based on a knowledge graph is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:
  • the target object is a user requesting to push health information; the target characteristic information refers to individual characteristic information of the target object.
  • the health information recommendation method based on the knowledge graph provided in this application can be applied to some applications, and the target object at this time can be a user requesting to push health information.
  • the target characteristic information refers to the individual characteristic information of the target object, for example, the age, height, or weight of the target object.
  • S12 Input the target feature information into a preset health feature similarity model to obtain a health classification label corresponding to the target feature information.
  • the preset health feature similarity model is used to determine the health classification label corresponding to the target feature information.
  • the preset health feature similarity model includes multiple sets of classification labels associated with the health information, and the classification label represents its corresponding According to the feature information of the object, the feature information similar to it in the health feature similarity model is determined according to the target feature information, and then the health classification label is determined.
  • the health classification label refers to the classification label corresponding to the target characteristic information.
  • the health classification label represents the target characteristic information of the target object.
  • the health classification label may be a BMI (Body Mass Index) index, that is According to the labels generated by the height and weight in the target feature information, different BMI indexes can be classified to obtain classification labels corresponding to the BMI index; the health classification labels can also be classified according to different age groups, such as 30-40 years old as a classification Label, 40-50 years old is a classification label; further, the health classification label can be obtained by combining the above multiple groups of classification labels, that is, the health classification label can be: age in the range of 40-50 years old and BMI index 19- Between 21.
  • BMI Body Mass Index
  • the target feature information of the target object After acquiring the target feature information of the target object, input the target feature information into a preset health feature similarity model, and determine the classification label corresponding to the target feature information according to the target feature information, that is, the health classification label .
  • S13 Obtain all sample triples with sample classification labels that are the same as the health classification labels from a preset health recommendation database constructed based on the knowledge graph.
  • the sample triad is composed of a sample classification label, sample health information, and a recommendation value association associated with both the sample health information and the sample classification label.
  • the preset health recommendation database stores multiple sets of sample triads, and the sample triads contain sample classification labels and corresponding sample health information. That is, the preset health recommendation database is used to provide information to those with the same label. Users provide a database of sample health information. Both the sample classification label and the health classification label are used to classify and identify the data.
  • the sample classification label is generated according to preset sample feature information of the sample object, and the sample classification label may contain the same label as the health classification label.
  • Sample triples refer to sample classification labels, sample health information, and recommended values associated with sample health information and sample classification labels (that is, a measure of the priority of recommending sample health information to objects with corresponding sample classification labels). For example, suppose the sample classification label is 45-50 years old, the sample health information is light diet + low-intensity exercise, and the recommended value is 85. Then the triplet is (45-50, light diet + Low-intensity exercise, 85).
  • the target feature information is input into the preset health feature similarity model, and the health classification labels corresponding to the target feature information are obtained, from all the sample classification labels of the preset health recommendation database constructed based on the knowledge graph , Match the sample classification label with the same health classification label, and then obtain all sample triples with the sample classification label.
  • the recommended value associated with each corresponding sample health information may also be different, there are multiple sets of different sample triples for each sample classification label.
  • S14 Extract the sample health information and the recommended value associated with the sample health information from all the acquired sample triples, and push the sample health information associated with the sample health information to the target object according to the recommended value.
  • the sample health information is health information recommended to the target object, and the sample health information may include, but is not limited to, information about healthy exercise and healthy diet.
  • the recommendation value is data that measures the priority of recommending sample health information to objects with corresponding sample classification labels.
  • the sample health information is extracted from all the obtained sample triads And the recommended value associated with the sample health information; the number of sample health information that needs to be recommended can be determined according to the recommendation needs of the target object, and then according to the recommended value from high to low, the number of health information corresponding to the recommended needs can be recommended to the target Object.
  • the recommendation requirement set by the target object is 2 sample health information
  • the recommendation requirement set by the target object is 2 sample health information
  • after extracting the sample health information and the recommended value associated with the sample health information from all sample triples select two recommendations from high to low Value corresponding to the sample health information, and send the two sample health information to the mobile terminal of the target object or push it to the target object through other effective sending methods; further, if the extraction is completed according to the method from high to low, there is When there are multiple sets of sample health information corresponding to the same recommended value, random selection or other methods can be adopted to extract the sample health information corresponding to the same recommended value, which is equivalent to the number of sample health information in the recommendation requirements set by the target object.
  • the target object is recommended to match the sample health information to improve the accuracy of the sample health information recommendation; at the same time according to the recommended value and user recommendation
  • the need to determine the health information of the push sample can provide users with more options, and these options have corresponding recommended values, so as to improve the user's choice under guaranteed conditions.
  • the preset health feature similarity model and the preset health recommendation database may be stored in the area.
  • Block chain the Blockchain is an encrypted and chained transaction storage structure formed by blocks.
  • step S12 the following steps are further included:
  • S21 Obtain a daily sample data set and a preset knowledge database constructed according to the knowledge graph; the daily sample data set contains at least one sample health information and sample object characteristics of the sample object associated with the sample health information in a one-to-one correspondence.
  • the daily sample data set collects the physical condition of a large number of sample objects (that is, the characteristics of the sample object, such as age, height, weight, body fat rate, heart rate, disease history, bone health status, etc.), real exercise status (such as sports Type, time, intensity, frequency, etc.) and dietary conditions (such as the ratio of three meals, the type of meals and the amount of meals, etc.).
  • the sample object is the object selected through a random survey.
  • the sample object can choose people of different occupations, ages, and physical health status, so as to make the data range of the daily sample data set wider, and then make the sample health information provided to other objects later. The scope is wider.
  • the characteristics of the sample object include, but are not limited to, the age, weight, or physical health of the sample object.
  • the preset knowledge database refers to a database constructed based on authoritative data collected from sports medicine books, diet books, literature, clinical guidelines, expert consensus and other knowledge data through the knowledge map framework.
  • the unreasonable information in the daily sample data set is cleaned according to the data in the preset knowledge database. For example, according to a preset knowledge database, it is determined whether the exercise type in the daily sample data set is reasonable for the characteristics of the corresponding sample object, or whether the exercise intensity is reasonable, or whether the diet combination is balanced, etc. If unreasonable information is found, it will be cleaned and deleted from the daily sample data set.
  • S23 Generate a sample classification label according to the characteristics of the sample object in the daily sample data set after the cleaning process, and associate the sample classification label with its corresponding sample health information.
  • the sample classification label is generated according to the characteristics of the sample object in the daily sample data set after the cleaning process, and the sample classification label is combined with it. Corresponding sample health information association.
  • multiple sets of different sample classification labels can be generated according to the characteristics of the sample object, such as 40-50 years old, 50-60 years old or classified according to the BMI index, and classified into sample classification labels such as normal BMI, low BMI, or high BMI ,
  • associate the sample classification label with the corresponding sample health information that is, each sample classification label has at least one set of corresponding sample health information, so that other servers can obtain the corresponding sample health information according to the sample classification label and recommend it to the target object.
  • S24 Construct a preset health recommendation database according to the sample classification label, the sample health information, and the preset recommendation algorithm.
  • the preset recommendation algorithm is used to determine the algorithm for the recommendation value corresponding to the sample health information associated with the sample classification label.
  • the sample classification label is generated according to the characteristics of the sample object in the daily sample data set after the cleaning process, and the sample classification label is associated with its corresponding sample health information
  • the sample classification label, the sample health information, and the sample health information are associated with each other.
  • the preset recommendation algorithm generates recommended values associated with sample classification labels and sample health information to construct sample triples based on sample classification labels, sample health information, and recommended values, and then build presets based on all sample triples Health recommendation database.
  • step S24 specifically includes the following steps:
  • S241 Construct a sample entity relationship according to the sample classification label and the corresponding sample health information.
  • the sample entity relationship refers to the relationship between each sample classification label and the corresponding sample health information.
  • the sample classification label A and sample health information B are a set of sample entity relationships
  • sample classification label A and sample health information C are another set of sample entity relationships.
  • sample classification label is generated according to the characteristics of the sample object in the daily sample data set after the cleaning process, and the sample classification label is associated with its corresponding sample health information
  • sample classification label and its corresponding sample Health information constructs sample entity relationships.
  • the support frequency refers to the frequency of each sample entity relationship in the preset health recommendation database, and the value range of the support frequency is any value from 0-1, such as 0.5, 0.6.
  • the support frequency of each sample entity relationship in the preset health recommendation database is obtained. Further, the support frequency corresponding to each sample entity relationship can be determined according to the following expression:
  • Support() is the support frequency function
  • X ⁇ Y represents the probability of Y occurring when X occurs or exists
  • X is any sample classification label
  • Y is any sample health information
  • X ⁇ Y is the preset health recommendation
  • the database contains both X and Y data (that is, the number of sample entity relationships that can be considered to include X and Y);
  • Z is the total number of data in the preset health recommendation database (that is, the total number of all sample entity relationships).
  • the essence of the support frequency is the total number of times the sample entity relationship appears in the preset health recommendation database.
  • the preset support threshold can be adjusted according to the number of sample entity relationships (for example, when the total number of sample entity relationships is small, the preset support threshold can be set to such as 0.3, 0.4, etc., to ensure the health of subsequent samples pushed for users
  • the amount of information when the total amount of sample entity relationships is large, the preset support threshold can be set to such as 0.6, 0.7, etc., to further screen the sample entity relationships and improve the accuracy of subsequent sample health information
  • the support frequency of entity relationships increases the recommended priority of sample entity relationships.
  • the support frequency of a sample entity relationship is 0.7
  • the preset support threshold value is 0.5
  • the basic recommendation priority of each sample entity relationship is 5, and the preset frequency priority rule is based on support
  • S244 Determine the recommended value of the sample health information in the sample entity relationship according to the improved recommendation priority and the preset recommendation algorithm.
  • the recommendation priority of the sample entity relationship is increased according to the support frequency of the entity relationship, and then according to the improved recommendation
  • the priority and the preset recommendation algorithm determine the recommended value of the sample health information of the sample entity relationship.
  • the preset recommendation algorithm may be to convert the number of levels corresponding to the promoted recommendation priority into a recommended value (such as the promoted recommendation priority If it is level 5, the corresponding recommended value may be 50), so the recommended value of the sample health information of the sample entity relationship is 70.
  • S245 After associating mutually corresponding recommended values, sample classification labels, and sample health information into a sample triad, construct a preset health recommendation database based on the sample triad.
  • the recommended value of the sample health information in the sample entity relationship is determined according to the improved recommendation priority and the preset recommendation algorithm, and the corresponding recommended value, sample classification label, and sample health information are associated into a sample ternary Group, that is, sample triples such as (sample classification label, sample health information, recommended value); build a preset health recommendation database based on all sample triples.
  • step S241 the method further includes:
  • S246 Obtain the confidence level of each sample entity relationship in a preset health recommendation database.
  • the confidence level of each sample entity relationship in the preset health recommendation database is obtained. Further, the confidence level corresponding to each sample entity relationship can be determined according to the following expression:
  • Conf() is the confidence function
  • X ⁇ Y represents the probability that Y occurs or exists when X occurs or exists
  • X is any sample classification label
  • Y is any sample health information
  • X ⁇ Y is the preset
  • the health recommendation database contains both X and Y data (that is, the number of sample entity relationships that can be considered to contain X and Y).
  • the confidence is the credibility of the sample entity relationship in the preset health recommendation database.
  • the preset reliability threshold can be adjusted and determined according to the total number of sample entity relationships in the preset health recommendation database.
  • the preset confidence priority rule After obtaining the confidence of each sample entity relationship in the preset health recommendation database, if there is a sample entity relationship with a confidence greater than or equal to the preset support threshold, according to the preset confidence priority rule, according to the sample entity relationship The confidence of the entity relationship increases the recommendation priority of the sample entity relationship.
  • the preset support threshold is 0.4
  • the basic recommendation priority of each sample entity relationship is 5, and the preset confidence priority rule is based on confidence
  • step S21 that is, before obtaining the daily sample data set and the preset knowledge database constructed according to the knowledge graph, the method further includes:
  • the knowledge sample data set including at least one knowledge sample data.
  • all knowledge sample data in the knowledge sample data set can be collected from data sources such as sports medicine books, diet books, literature, clinical guidelines, and expert consensus.
  • Extract all sample entities in the knowledge sample data and obtain the position coding vector associated with each sample entity according to the distance between each data in the knowledge sample data and each sample entity.
  • the sample entities include but are not limited to exercise entities (such as exercise classification, exercise intensity), diet entities (such as food types, nutritional ingredients), and so on.
  • the position encoding vector is generated by encoding the distance between each data in the knowledge sample data and each sample entity.
  • the knowledge sample data contains two sample entities of “jogging” and “medium-intensity”, with "
  • the position of the medium intensity in the knowledge sample data is coded as 0).
  • the encoding is performed in units of words.
  • the "motion" in the above knowledge sample data is a word position encoding
  • the "type” is a word position encoding. Therefore, when “medium intensity” is used as the sample entity, the position of "motion” is coded as 1, and the position of "type” is coded as 2.
  • the word segmentation rules can be segmented according to methods such as stammering.
  • the feature recognition of the knowledge sample data is not a single Character recognition, but to recognize a group of words, such as “jogging”, “exercise”, “method”, etc. in “jogging is a common exercise in daily life, belonging to the type of moderate-intensity exercise", and then obtain knowledge sample data The corresponding sample feature vector.
  • the feature recognition of the knowledge sample data is only performed from the perspective of words or characters, without the position code in the foregoing embodiment.
  • the sample feature vector and all the position coding vectors are input into a preset convolutional neural network to obtain a sample classification result, and the sample classification result represents the intimacy between at least two sample entities.
  • the sample classification result represents The intimacy between at least two of the sample entities is determined. Understandably, the sample classification result represents the relationship or the degree of association between the two entities. For example, in the above embodiment, "jogging is a common exercise in daily life and belongs to the type of moderate-intensity exercise”.
  • the final sample classification result is : There is a correlation between "jogging" and “medium intensity”, or there is a strong correlation between jogging and medium intensity (that is, jogging is a type of moderate intensity exercise).
  • the inputting the sample feature vector and the position coding vector into a preset convolutional neural network to obtain a sample classification result for cleaning the daily sample data set includes :
  • the sample splicing vector is input into a preset convolutional neural network, and the sample splicing vector is processed through the preset convolutional neural network. Feature extraction to obtain at least one feature extraction vector.
  • the first splicing process refers to splicing the sample feature vector and the position coding vector together to represent the word vector, that is, the vector to be input into the preset convolutional neural network.
  • the feature extraction vector is obtained after feature extraction is performed on the sample stitching vector, and the feature extraction vector represents the feature information corresponding to each sample entity.
  • the sample feature vector corresponding to the knowledge sample data is obtained, and the position associated with each sample entity extracted is obtained according to the distance between each data in the knowledge sample data and each sample entity
  • the first splicing process on the sample feature vector and the position encoding vector as the word vector representation to be input to the preset convolutional neural network, that is, the sample splicing vector; input the sample splicing vector to the preset
  • feature extraction is performed on the sample stitching vector through the convolutional layer in the preset convolutional neural network to obtain at least one feature extraction vector.
  • multiple convolution cores of different sizes can be designed, and the number of convolution kernels of different sizes can be determined according to the number of sample entities
  • "jogging is a common exercise in daily life, belonging to the medium-intensity exercise type” includes two sample entities of "jogging” and "medium-intensity”, so two rolls of different sizes can be designed Integrate kernels (such as 3*3, 5*5, etc.) to perform more accurate feature extraction on the sample stitching vector to obtain a more comprehensive feature extraction vector.
  • each feature extraction vector after pooling is subjected to a second splicing process to obtain a feature splicing vector.
  • the pooling process is to reduce the number of feature extraction vectors, prevent data overfitting, and reduce subsequent calculation complexity.
  • each feature extraction vector is pooled through the pooling layer of the preset convolutional neural network, The number of parameters is further reduced, thereby reducing the computational complexity, and each feature extraction vector after the pooling process is subjected to the second splicing process to obtain the feature splicing vector.
  • the feature stitching vector is classified and recognized through the fully connected layer of the preset convolutional neural network, and the sample classification result is obtained.
  • each feature extraction vector after pooling is subjected to a second splicing process, and after the feature splicing vector is obtained, the The feature stitching vector is input to the fully connected layer (that is, the softmax layer), the feature stitching vector is classified and recognized, and the sample classification result is obtained.
  • the fully connected layer that is, the softmax layer
  • the cleaning processing of unreasonable information in the daily sample data set according to a preset knowledge database includes the following steps:
  • the health information of each sample in the daily sample data set and the matching degree between the characteristics of the corresponding sample object are obtained.
  • the sample health information in the daily sample data set and the corresponding Whether the characteristics of the sample objects match and then determine the degree of matching between the health information of each sample and the sample object characteristics of the sample object associated with it.
  • the corresponding sample health information includes exercise information, which is moderate-intensity exercise, or the diet information does not mention not eating too much and sugar.
  • the recommended sports recommendation is not the most suitable for the characteristics of the sample object, and the diet recommendation is not perfect, and the health information of the sample has a low degree of matching with the characteristics of the corresponding sample object.
  • the characteristics of the sample object and the health information of the sample whose matching degree is lower than the preset matching threshold are recorded as unreasonable information to clean up the unreasonable information.
  • the matching degree is lower than the preset matching threshold and is correlated with each other.
  • the characteristics of the sample object and the health information of the sample are recorded as unreasonable information, so as to clean up the unreasonable information.
  • the preset matching threshold may be 80%, 85%, or 90%, etc.
  • the preset matching threshold may be changed according to different groups of target objects.
  • a health information recommendation device based on a knowledge graph is provided, and the health information recommendation device based on the knowledge graph corresponds to the health information recommendation method based on the knowledge graph in the foregoing embodiment in a one-to-one correspondence.
  • the health information recommendation device based on the knowledge graph includes a feature information acquisition module 11, a classification label determination module 12, a sample triple acquisition module 13 and a health information recommendation module 14. The detailed description of each functional module is as follows:
  • the characteristic information acquisition module 11 is used to acquire target characteristic information of the target object.
  • the target object is a user requesting to push health information; the target characteristic information refers to individual characteristic information of the target object.
  • the classification label determination module 12 is configured to input the target feature information into a preset health feature similarity model to obtain a health classification label corresponding to the target feature information.
  • the sample triple acquisition module 13 is used to acquire all sample triples with the same sample classification label as the health classification label from a preset health recommendation database constructed based on the knowledge graph.
  • the sample triad is composed of a sample classification label, sample health information, and a recommendation value association associated with both the sample health information and the sample classification label.
  • the health information recommendation module 14 is configured to extract the sample health information and the recommended value associated with the sample health information from all the acquired triples, and push the associated value to the mobile terminal of the target object according to the recommended value Health information of the sample.
  • the health information recommendation device based on the knowledge graph further includes the following modules:
  • the data acquisition module 21 is used to acquire a daily sample data set and a preset knowledge database constructed according to the knowledge graph; the daily sample data set contains at least one sample health information and sample objects associated with the sample health information in a one-to-one correspondence The characteristics of the sample object.
  • the data cleaning module 22 is configured to clean the unreasonable information in the daily sample data set according to the preset knowledge database; the unreasonable information refers to the characteristics of the sample object and that are related to each other and do not match.
  • the sample health information is configured to clean the unreasonable information in the daily sample data set according to the preset knowledge database.
  • the classification label generation module 23 is configured to generate sample classification labels according to the characteristics of the sample objects in the daily sample data set after the cleaning process, and associate the sample classification labels with corresponding sample health information.
  • the database construction module 24 is configured to construct the preset health recommendation database according to the sample classification label, the sample health information, and a preset recommendation algorithm.
  • the database building module includes the following units:
  • the entity relationship construction unit 241 is configured to construct a sample entity relationship according to the sample classification label and the sample health information corresponding thereto.
  • the supporting frequency obtaining unit 242 is configured to obtain the supporting frequency of each sample entity relationship in the preset health recommendation database.
  • the first priority raising unit 243 is configured to increase the entity relationship according to the support frequency of the entity relationship according to the preset frequency priority rule when the support frequency of the entity relationship is greater than or equal to the preset support threshold. Recommended priority.
  • the recommended value determining unit 244 is configured to determine the recommended value of the sample health information in the entity relationship according to the improved recommendation priority and a preset recommendation algorithm.
  • the database construction unit 245 is configured to associate the recommended value, the sample classification label, and the sample health information corresponding to each other into a sample triad, and then construct the preset health according to the sample triad. Recommended database.
  • the database building module further includes the following units:
  • the confidence degree obtaining unit 246 is configured to obtain the confidence degree of each of the sample entity relationships in the preset health recommendation database.
  • the second priority raising unit 247 is configured to, when the confidence of the entity relationship is greater than or equal to the preset confidence threshold, according to the preset confidence priority rule, improve the entity relationship according to the confidence of the entity relationship Recommended priority.
  • the device for recommending health information based on the knowledge graph further includes the following modules:
  • the sample data set acquisition module is used to acquire a knowledge sample data set, the knowledge sample data set containing at least one knowledge sample data.
  • the sample entity extraction module is used to extract all sample entities in the knowledge sample data, and obtain and extract each of the sample entities according to the distance between each data in the knowledge sample data and each of the sample entities The associated position code vector.
  • the sample feature recognition module is used to perform feature recognition on the knowledge sample data to obtain the sample feature vector corresponding to the knowledge sample data.
  • the sample classification result generation module is used to input the sample feature vector and all the position coding vectors into a preset convolutional neural network to obtain a sample classification result, and the sample classification result represents at least two of the samples The intimacy between entities.
  • the sample classification result generation module includes:
  • the first splicing unit is configured to input the sample splicing vector into a preset convolutional neural network after performing the first splicing process on the sample feature vector and the position coding vector to obtain the sample splicing vector, and pass The preset convolutional neural network performs first splicing processing on the sample feature vector and each of the position coding vectors to obtain at least one sample splicing vector.
  • the feature extraction unit is configured to input the splicing vector into the preset convolutional neural network, and perform feature extraction on the sample splicing vector to obtain at least one feature extraction vector.
  • the second splicing unit is configured to perform a second pooling process on each of the feature extraction vectors through the pooling layer of the preset convolutional neural network, and then perform a second process on each of the feature extraction vectors after the pooling process.
  • the splicing process obtains the feature splicing vector.
  • the sample classification result generation unit is configured to classify and recognize the feature stitching vector through the fully connected layer of the preset convolutional neural network to obtain the sample classification result.
  • the data cleaning module further includes the following units:
  • the matching degree obtaining unit is configured to obtain the matching degree between the health information of each sample in the daily sample data set and the sample object characteristics of the sample object associated therewith according to the sample classification result;
  • the data cleaning unit is configured to record the characteristics of the sample object and the health information of the sample that are related to each other as unreasonable information with the matching degree lower than a preset matching threshold, so as to clean the unreasonable information.
  • the various modules in the above-mentioned health information recommendation device based on the knowledge graph can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a server, and its internal structure diagram may be as shown in FIG. 10.
  • the computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a readable storage medium and an internal memory.
  • the readable storage medium stores an operating system, computer readable instructions, and a database.
  • the internal memory provides an environment for the operation of the operating system and computer readable instructions in the readable storage medium.
  • the database of the computer device is used to store the data used in the health information recommendation method based on the knowledge graph in the foregoing embodiment.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer-readable instructions are executed by the processor, a method for recommending health information based on the knowledge graph is realized.
  • the readable storage medium provided in this embodiment includes a non-volatile readable storage medium and a volatile readable storage medium.
  • a computer device including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor.
  • the processor executes the computer-readable instructions, the following is achieved step:
  • the target object is a user requesting to push health information
  • the target characteristic information refers to the individual characteristic information of the target object
  • the sample triad is composed of sample classification labels, sample health information, and The recommended value association structure associated with the sample health information and the sample classification label;
  • one or more readable storage media storing computer readable instructions are provided.
  • the readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage. Medium; the readable storage medium stores computer readable instructions, and when the computer readable instructions are executed by one or more processors, the one or more processors perform the following steps:
  • the target object is a user requesting to push health information
  • the target characteristic information refers to the individual characteristic information of the target object
  • the sample triad is composed of sample classification labels, sample health information, and The recommended value association structure associated with the sample health information and the sample classification label;
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Primary Health Care (AREA)
  • Computing Systems (AREA)
  • Public Health (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Epidemiology (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Nutrition Science (AREA)
  • Physical Education & Sports Medicine (AREA)
  • Animal Behavior & Ethology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于知识图谱的健康信息推荐方法、装置、设备及介质，涉及人工智能技术领域，应用于智慧医疗中。该方法通过将获取目标对象的目标特征信息输入至预设的健康特征相似度模型中，得到与目标特征信息对应的健康分类标签；自基于知识图谱构建的预设的健康推荐数据库中，获取具有与健康分类标签相同的样本分类标签的所有样本三元组；自获取的所有样本三元组中提取样本健康信息以及与样本健康信息关联的推荐值，根据推荐值向目标对象推送与其关联的样本健康信息。根据基于知识图谱构建的健康推荐数据库中的样本三元组，给目标对象推荐与其匹配的样本健康信息，提高该样本健康信息推荐的准确率，能够推动智慧城市的建设。

Description

基于知识图谱的健康信息推荐方法、装置、设备及介质
本申请要求于2020年7月28日提交中国专利局、申请号为202010737217.6，发明名称为“基于知识图谱的健康信息推荐方法、装置、设备及介质”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
本申请涉及人工智能技术领域，尤其涉及一种基于知识图谱的健康信息推荐方法、装置、设备及介质。
背景技术
随着社会与经济的发展，人们对健康的意识越来越强烈，不管是健康人群还是亚健康人群，甚至患者，都可以从运动和饮食方面对自身身体状况进行调节。发明人意识到，大部分人对自己运动和饮食的规划源于网络上广为流传的方法，盲目地模仿别人的运动计划和饮食安排反而可能会造成不良后果。比如，对于糖尿病患者来说，需要控制碳水化合物摄入，以免使血糖升高，还需要适当运动；但是，在缺乏专业医生的指导下，患者对于碳水化合物的摄入量和运动量的适当的判断准确度低，其实用性也层次不齐，很难实现自我调节血糖水平；在盲目使用错误调节方式的状况下，甚至会导致病情加重。
申请内容
本申请实施例提供一种基于知识图谱的健康信息推荐方法、装置、设备及介质，以解决关于饮食和运动等健康信息的判断的实用性低以及准确率较低的问题。
一种基于知识图谱的健康信息推荐方法，包括：
获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
一种基于知识图谱的健康信息推荐装置，包括：
特征信息获取模块，用于获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
分类标签确定模块，用于将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
样本三元组获取模块，用于自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
健康信息推荐模块，用于自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上 运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现如下步骤：
获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
一个或多个存储有计算机可读指令的可读存储介质，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行如下步骤：
获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
上述基于知识图谱的健康信息推荐方法、装置、设备及介质，通过将获取目标对象的目标特征信息输入至预设的健康特征相似度模型中，得到与目标特征信息对应的健康分类标签；自基于知识图谱构建的预设的健康推荐数据库中，获取具有与健康分类标签相同的样本分类标签的所有样本三元组；自获取的所有样本三元组中提取样本健康信息以及与样本健康信息关联的推荐值，根据推荐值向目标对象推送与其关联的样本健康信息。本申请根据基于知识图谱构建的健康推荐数据库中的样本三元组，给目标对象推荐与其匹配的样本健康信息，提高该样本健康信息推荐的准确率；同时根据推荐值与用户推荐需求来确定推送样本健康信息，可以为用户提供更多选择方案，且这些方案均有相应的推荐值，从而在有保障的情况下，提高用户的可选性。本申请属于智慧医疗领域，且涉及与健康管理相关的数字医疗，通过本申请能够推动智慧城市的建设。
本申请的一个或多个实施例的细节在下面的附图和描述中提出，本申请的其他特征和优点将从说明书、附图以及权利要求变得明显。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对本申请实施例的描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请一实施例中基于知识图谱的健康信息推荐方法的一应用环境示意图；
图2是本申请一实施例中基于知识图谱的健康信息推荐方法的一流程图；
图3是本申请一实施例中基于知识图谱的健康信息推荐方法的另一流程图；
图4是本申请一实施例中基于知识图谱的健康信息推荐方法中步骤S24的一流程图；
图5是本申请一实施例中基于知识图谱的健康信息推荐方法中步骤S24的另一流程图；
图6是本申请一实施例中基于知识图谱的健康信息推荐装置的一原理框图；
图7是本申请一实施例中基于知识图谱的健康信息推荐装置的另一原理框图；
图8是本申请一实施例中基于知识图谱的健康信息推荐装置中数据库构建模块的一原理框图；
图9是本申请一实施例中基于知识图谱的健康信息推荐装置中数据库构建模块的另一原理框图；
图10是本申请一实施例中计算机设备的一示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请实施例提供的基于知识图谱的健康信息推荐方法，该基于知识图谱的健康信息推荐方法可应用如图1所示的应用环境中。具体地，该基于知识图谱的健康信息推荐方法应用在基于知识图谱的健康信息推荐系统中，该基于知识图谱的健康信息推荐系统包括如图1所示的客户端和服务器，客户端与服务器通过网络进行通信，用于解决关于饮食和运动等健康信息的判断的实用性低以及准确率较低的问题。其中，客户端又称为用户端，是指与服务器相对应,为客户提供本地服务的程序。客户端可安装在但不限于各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备上。服务器可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
在一实施例中，如图2所示，提供一种基于知识图谱的健康信息推荐方法，以该方法应用在图1中的服务器为例进行说明，包括如下步骤：
S11：获取目标对象的目标特征信息。所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息。
其中，本申请提供的基于知识图谱的健康信息推荐方法可以应用于一些应用程序中，则此时的目标对象可以为请求推送健康信息的用户。目标特征信息指的是目标对象的个体特征信息，示例性地，目标对象的年龄、身高或者体重等。
S12：将目标特征信息输入至预设的健康特征相似度模型中，得到与目标特征信息对应的健康分类标签。
其中，预设的健康特征相似度模型用于确定与目标特征信息对应的健康分类标签，预设的健康特征相似度模型中包括多组与健康信息关联的分类标签，该分类标签表征与其对应的对象的特征信息，根据目标特征信息确定在健康特征相似度模型中与其相似的特征信息，进而确定健康分类标签。健康分类标签指的是与目标特征信息对应的分类标签，该健康分类标签表征了目标对象的目标特征信息，示例性地，健康分类标签可以为BMI(Body Mass Index，体质指数)指数，也即根据目标特征信息中的身高以及体重生成的标签，可以将不同BMI指数进行分类以得到与BMI指数对应的分类标签；健康分类标签还可以根据不同年龄层进行划分，如30-40岁为一个分类标签，40-50岁为一个分类标签；进一步地，健康分类标签可以由上述多组分类标签共同组合得到，也即健康分类标签可以为：年龄在40-50岁范围内且BMI指数在19-21之间。
具体地，在获取目标对象的目标特征信息之后，将该目标特征信息输入至预设的健康特征相似度模型中，根据目标特征信息，确定该目标特征信息对应的分类标签，也即健康分类标签。
S13：自基于知识图谱构建的预设的健康推荐数据库中，获取具有与健康分类标签相同的样本分类标签的所有样本三元组。所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成。
其中，预设的健康推荐数据库中存储多组样本三元组，该样本三元组中包含样本分类标签以及与其对应的样本健康信息，也即预设的健康推荐数据库用于给具有相同标签的用户提供样本健康信息的数据库。样本分类标签与健康分类标签均是对数据进行分类标识，该样本分类标签是根据预设的样本对象的样本特征信息生成的，样本分类标签中可能包含与健康分类标签相同的标签。样本三元组指的是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值(也即衡量将样本健康信息推荐给具有对应的样本分类标签的对象的优先程度的数据)构成的，示例性地，假设样本分类标签为45-50岁，样本健康信息为清淡饮食+低强度运动，推荐值为85，则该三元组则为(45-50，清淡饮食+低强度运动，85)。
具体地，在将目标特征信息输入至预设的健康特征相似度模型中，得到与目标特征信息对应的健康分类标签之后，从基于知识图谱构建的预设的健康推荐数据库的所有样本分类标签中，匹配出具有与健康分类标签相同的样本分类标签，进而获取与该样本分类标签的所有样本三元组。其中，由于与样本分类标签对应的样本健康信息存在多组，且每一对应的样本健康信息关联的推荐值也可能不同，故每一样本分类标签存在多组不同的样本三元组。
S14：自获取的所有样本三元组中提取样本健康信息以及与样本健康信息关联的推荐值，根据推荐值向目标对象推送与其关联的样本健康信息。
其中，样本健康信息为向目标对象推荐的健康信息，该样本健康信息中可以包括但不限于关于健康运动以及关于健康饮食的信息。推荐值为衡量将样本健康信息推荐给具有对应的样本分类标签的对象的优先程度的数据。
具体地，在自基于知识图谱构建的预设的健康推荐数据库中，获取具有与健康分类标签相同的样本分类标签的所有样本三元组之后，从获取的所有样本三元组中提取样本健康信息以及与样本健康信息关联的推荐值；可以根据目标对象的推荐需求，确定需要推荐的样本健康信息的数量，进而可以根据推荐值从高到低，将与推荐需求对应数量的健康信息推荐给目标对象。示例性地，假设目标对象设置的推荐需求为2个样本健康信息，则在从所有样本三元组中提取样本健康信息以及与样本健康信息关联的推荐值后，从高到低选取两个推荐值对应的样本健康信息，并将这两个样本健康信息发送至目标对象的移动终端或者通过其它有效的发送方式推送给目标对象；进一步地，若根据从高到低的方法提取完毕之后，存在多组相同推荐值对应的样本健康信息时，则可以采取随机选取或其它方法，从这些相同推荐值对应的样本健康信息中提取，等同于目标对象设置的推荐需求中样本健康信息的数量。
在本实施例中，根据基于知识图谱构建的健康推荐数据库中的样本三元组，给目标对象推荐与其匹配的样本健康信息，提高该样本健康信息推荐的准确率；同时根据推荐值与用户推荐需求来确定推送样本健康信息，可以为用户提供更多选择方案，且这些方案均有相应的推荐值，从而在有保障的情况下，提高用户的可选性。
作为优选，为了保证上述实施例中预设的健康特征相似度模型以及预设的健康推荐数据库的私密以及安全性，可以将预设的健康特征相似度模型以及预设的健康推荐数据库存储在区块链中。其中，区块链(Blockchain)，是由区块(Block)形成的加密的、链式的交易的存储结构。
在一实施例中，如图3所示，步骤S12之前，还包括如下步骤：
S21：获取日常样本数据集以及根据知识图谱构建的预设的知识数据库；日常样本数据集中包含至少一个样本健康信息以及与样本健康信息一一对应关联的样本对象的样本对象特征。
其中，日常样本数据集是通过收集大量样本对象的身体状况(也即样本对象特征，如年龄，身高，体重，体脂率，心率，疾病史，骨骼健康状态等)、真实运动情况(如运动 类型、时间、强度及频率等)以及饮食情况(如三餐比例、膳食种类及膳食量等)等得到的。样本对象是通过随机调查选取的对象，该样本对象可以选择不同职业、年龄以及身体健康状态的人群，以令日常样本数据集中的数据范围更广，进而使得后续为其它对象提供的样本健康信息的范围更广。样本对象特征包括但不限于样本对象的年龄、体重或者身体健康状态等。预设的知识数据库指的是通过知识图谱框架，并根据从运动医学书籍、饮食书籍、文献、临床指南、专家共识等知识数据收集到的权威数据构建生成的数据库。
S22：根据预设的知识数据库，对日常样本数据集中的不合理信息进行清洗处理；所述不合理信息是指相互关联且不匹配的样本对象特征以及样本健康信息。
具体地，在获取日常样本数据集以及根据知识图谱构建的预设的知识数据库之后，根据预设的知识数据库中的数据，对日常样本数据集中的不合理信息进行清洗处理。示例性地，如根据预设的知识数据库，判断日常样本数据集中的运动类型对与其对应的样本对象特征来说是否合理、或者运动强度是否合理、亦或者饮食搭配是否均衡等。如果发现有不合理信息，将其从日常样本数据集中清洗删除。
S23：根据清洗处理之后的日常样本数据集中的样本对象特征生成样本分类标签，并将样本分类标签和与其对应的样本健康信息关联。
具体地，在根据预设的知识数据库，对日常样本数据集中的不合理信息进行清洗处理之后，根据清洗处理之后的日常样本数据集中的样本对象特征生成样本分类标签，并将样本分类标签和与其对应的样本健康信息关联。示例性地，可以根据样本对象特征生成多组不同的样本分类标签，如年龄40-50岁，50-60岁或者根据BMI指数分类，分为BMI正常、BMI偏低或者BMI偏高等样本分类标签，并将样本分类标签与对应的样本健康信息关联，也即每一样本分类标签至少一组对应的样本健康信息，以供其它服务器根据样本分类标签获取对应的样本健康信息并推荐给目标对象。
S24：根据样本分类标签、样本健康信息以及预设的推荐算法构建预设的健康推荐数据库。
其中，预设的推荐算法用于确定与样本分类标签关联的样本健康信息对应的推荐值的算法。
具体地，在根据清洗处理之后的所述日常样本数据集中的样本对象特征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联之后，根据样本分类标签、样本健康信息以及预设的推荐算法，生成与样本分类标签以及样本健康信息均关联的推荐值，以根据样本分类标签、样本健康信息以及推荐值构建样本三元组，进而根据所有样本三元组构建预设的健康推荐数据库。
在本实施例中，通过基于知识图谱构建的知识数据库，对日常样本数据集中不合理的信息进行清洗处理，提高日常样本数据集中与样本特征信息关联的样本健康信息的准确率。
在一实施例中，如图4所示，步骤S24中，具体包括如下步骤：
S241：根据样本分类标签以及与其对应的样本健康信息构建样本实体关系。
其中，样本实体关系指的是每一样本分类标签与对应的样本健康信息之间的关系，示例性地，假设样本分类标签A对应的有样本健康信息B以及样本健康信息C，则样本分类标签A与样本健康信息B为一组样本实体关系；样本分类标签A与样本健康信息C为另一组样本实体关系。
具体地，在根据清洗处理之后的所述日常样本数据集中的样本对象特征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联之后，根据样本分类标签以及与其对应的样本健康信息构建样本实体关系。
S242：获取每一样本实体关系在预设的健康推荐数据库的支持频率。
其中，支持频率指的是每一样本实体关系在预设的健康推荐数据库中出现的频率，支 持频率的取值范围为0-1中任意一个数值，如0.5,0.6。
具体地，在根据样本分类标签以及与其对应的样本健康信息构建样本实体关系之后，获取每一样本实体关系在预设的健康推荐数据库的支持频率。进一步地，可以根据如下表达式确定每一样本实体关系对应的支持频率：
Figure PCTCN2020117090-appb-000001
其中，Support()为支持频率函数；X→Y表示X发生或者存在时，Y发生的概率；X为任意一个样本分类标签；Y为任意一个样本健康信息；X∪Y为预设的健康推荐数据库中同时包含X和Y的数据(也即可以认为包含X和Y的样本实体关系的数量)；Z为预设的健康推荐数据库中数据的总数(也即所有样本实体关系的总数量)。
S243：在样本实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据实体关系的支持频率提升样本实体关系的推荐优先级。
其中，支持频率的实质为该样本实体关系在预设的健康推荐数据库中出现的总次数。预设支持度阈值可以根据样本实体关系的数量进行调整(如在样本实体关系的总量较小时，该预设支持度阈值可以设置为如0.3，0.4等，以保证后续为用户推送的样本健康信息的数量；在样本实体关系的总量较大时，该预设支持度阈值可以设置为如0.6,0.7等，以进一步对样本实体关系的筛选，提高后续推送的样本健康信息的准确率)，也可以通过用户需求进行设定。
具体地，在获取每一样本实体关系在预设的健康推荐数据库的支持频率之后，若存在支持频率大于或等于预设支持度阈值的样本实体关系，按照预设的频率优先级规则，根据样本实体关系的支持频率提升样本实体关系的推荐优先级。示例性地，假设某一样本实体关系的支持频率为0.7，预设支持度阈值为0.5，假设每一样本实体关系的基础推荐优先级均为5级，预设的频率优先级规则为根据支持频率与预设支持度阈值之差提升推荐优先级(示例性地，若支持频率与预设支持度阈值之差为0.1，则提升1级)，而该支持频率与预设支持度阈值之差为0.7-0.5＝0.2，进而该样本实体关系的推荐优先级从5级提升两个等级至7级。
S244：根据提升之后的推荐优先级以及预设的推荐算法确定样本实体关系中的样本健康信息的推荐值。
具体地，在样本实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据实体关系的支持频率提升样本实体关系的推荐优先级之后，根据提升之后的推荐优先级以及预设的推荐算法确定样本实体关系的样本健康信息的推荐值。示例性地，如上述样本实体关系的提升之后的推荐优先级为7级，预设的推荐算法可以为将提升后的推荐优先级对应的级数转换为推荐值(如提升后的推荐优先级为5级，则对应的推荐值可以为50)，因此该样本实体关系的样本健康信息的推荐值为70。
S245：将相互对应的推荐值、样本分类标签以及样本健康信息关联为一个样本三元组之后，根据样本三元组构建预设的健康推荐数据库。
具体地，在根据提升之后的推荐优先级以及预设的推荐算法确定样本实体关系中的样本健康信息的推荐值，将相互对应的推荐值、样本分类标签以及样本健康信息关联为一个样本三元组，也即如(样本分类标签，样本健康信息，推荐值)的样本三元组；根据所有样本三元组构建预设的健康推荐数据库。
在一实施例中，如图5所示，在步骤S241之后，还包括：
S246：获取每一样本实体关系在预设的健康推荐数据库的置信度。
具体地，在根据样本分类标签以及与其对应的样本健康信息构建样本实体关系之后， 获取每一样本实体关系在预设的健康推荐数据库的置信度。进一步地，可以根据如下表达式确定每一样本实体关系对应的置信度：
Figure PCTCN2020117090-appb-000002
其中，Conf()为置信度函数；X→Y表示X发生或者存在时，Y发生或者存在的概率；X为任意一个样本分类标签；Y为任意一个样本健康信息；X∪Y为预设的健康推荐数据库中同时包含X和Y的数据(也即可以认为包含X和Y的样本实体关系的数量)。
S247：在实体关系的置信度大于或等于预设置信度阈值时，按照预设的置信优先级规则，根据实体关系的置信度提升实体关系的推荐优先级。
其中，置信度为样本实体关系在预设的健康推荐数据库中的可信程度。预设置信度阈值可以根据预设的健康推荐数据库中的样本实体关系的总数进行调整确定。
具体地，在获取每一样本实体关系在预设的健康推荐数据库的置信度之后，若存在置信度大于或等于预设支持度阈值的样本实体关系，按照预设的置信优先级规则，根据样本实体关系的置信度提升样本实体关系的推荐优先级。示例性地，假设某一样本实体关系的置信度为0.8，预设支持度阈值为0.4，假设每一样本实体关系的基础推荐优先级均为5级，预设的置信优先级规则为根据置信度与预设置信度阈值之差提升推荐优先级(示例性地，若支持频率与预设支持度阈值之差为0.1，则提升1级)，而该支持频率与预设支持度阈值之差为0.8-0.4＝0.4，进而该样本实体关系的推荐优先级从5级提升四个等级至9级。
在一实施例中，步骤S21之前，也即在获取日常样本数据集以及根据知识图谱构建的预设的知识数据库之前，还包括：
获取知识样本数据集，所述知识样本数据集包含至少一个知识样本数据。
其中，知识样本数据集中的所有知识样本数据均可以从运动医学书籍、饮食书籍、文献、临床指南、专家共识等数据源中采集得到。
提取知识样本数据中的所有样本实体，并根据知识样本数据中各数据与每一样本实体之间的距离，获取与提取的各样本实体关联的位置编码向量。
其中，样本实体包括但不限于运动实体(如运动分类、运动强度)、饮食实体(如食品种类、营养成分)等。位置编码向量是根据知识样本数据中各数据与每一样本实体之间的距离进行编码生成的。
具体地，在获取知识样本数据集之后，提取知识样本数据集中每一知识样本数据的所有样本实体。并根据每一知识样本数据中各数据与对应的每一样本实体之间的距离进行位置编码，以获取与提取的各样本实体关联的位置编码向量。
示例性地，假设一个知识样本数据为“慢跑是日常生活中常见的运动方式，属于中等强度运动类型”，该知识样本数据中包含两个样本实体为“慢跑”以及“中等强度”，以“慢跑”为样本实体时，根据各数据与“慢跑”之间的距离进行位置编码后，得到的位置编码向量为pos_1＝[0,1,2,3,...](此时将慢跑在该知识样本数据中的位置编码为0)；以“中等强度”为样本实体时，根据各数据与“中等强度”之间的距离进行位置编码后，得到的位置编码向量为pos_2＝[...,-3,-2,-1,0,1,2](此时将中等强度在该知识样本数据中的位置编码为0)。作为优选，在对知识样本数据进行位置编码时，按照分词的规则，以词为单位进行编码，如上述知识样本数据中的“运动”为一个词位置编码，“类型”为一个词位置编码，故在以“中等强度”为样本实体时，“运动”位置编码为1，“类型”位置编码为2，该分词的规则可以根据如结巴分词等方法进行分词。
对知识样本数据进行特征识别，得到知识样本数据对应的样本特征向量。
作为优选，在获取知识样本数据集之后，对知识样本数据集中的知识样本数据进行特征识别，且在进行特征识别过程中，是按照词的识别方式，也即对知识样本数据进行特征 识别不是单个字符识别，而是对一组词进行识别，如“慢跑是日常生活中常见的运动方式，属于中等强度运动类型”中的“慢跑”，“运动”，“方式”等，进而得到知识样本数据对应的样本特征向量。可以理解地，对知识样本数据进行特征识别，只是从字词或者字符的角度进行识别，不带有上述实施例中的位置编码。
将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，所述样本分类结果表征了至少两个所述样本实体之间的亲密度。
在得到样本特征向量以及所有位置编码向量之后，将样本特征向量以及所有位置编码向量进行拼接后，输入至预设的卷积神经网络中进行特征提取，以得到样本分类结果，该样本分类结果表征了至少两个所述样本实体之间的亲密度。可以理解地，该样本分类结果表征了两个实体之间的关系或者关联程度，例如上述实施例中“慢跑是日常生活中常见的运动方式，属于中等强度运动类型”最后得到的样本分类结果为：“慢跑”与“中等强度”之间是关联的，或者说慢跑与中等强度之间存在较强的关联关系(也即慢跑是中等强度运动类型)。
在一实施例中，所述将所述样本特征向量以及所述位置编码向量输入至预设的卷积神经网络中，得到用于对所述日常样本数据集进行清洗处理的样本分类结果，包括：
在对样本特征向量以及所述位置编码向量进行第一拼接处理得到样本拼接向量之后，将样本拼接向量输入至预设的卷积神经网络中，通过预设的卷积神经网络对样本拼接向量进行特征提取，得到至少一个特征提取向量。
其中，第一拼接处理指的是将样本特征向量与位置编码向量拼接在一起，以作为词向量表示，也即待输入至预设的卷积神经网络中的向量。特征提取向量为对样本拼接向量进行特征提取之后得到的，该特征提取向量表征了每一样本实体对应的特征信息。
具体地，在对知识样本数据进行特征识别，得到知识样本数据对应的样本特征向量，以及根据知识样本数据中各数据与每一样本实体之间的距离，获取与提取的各样本实体关联的位置编码向量之后，对样本特征向量以及位置编码向量进行第一拼接处理，以作为待输入至预设的卷积神经网络的词向量表示，也即样本拼接向量；将该样本拼接向量输入至预设的卷积神经网络中，通过预设的卷积神经网络中的卷积层对样本拼接向量进行特征提取，得到至少一个特征提取向量。
作为优选，为了提取到的特征提取向量更加全面和精确，在对样本拼接向量进行特征提取时，可以设计多个不同尺寸的卷结核，不同尺寸的卷积核的数量可以根据样本实体的数量确定；示例性地，上述实施例中“慢跑是日常生活中常见的运动方式，属于中等强度运动类型”包含“慢跑”和“中等强度”两个样本实体，故可以设计两个不同尺寸大小的卷积核(如3*3,5*5等)，以对样本拼接向量进行更精确的特征提取，得到更全面的特征提取向量。
通过预设的卷积神经网络的池化层对各特征提取向量进行池化处理后，将池化处理后的各特征提取向量进行第二拼接处理，得到特征拼接向量。
其中，池化处理是为了减少特征提取向量的数量，防止数据过拟合，减少后续计算复杂度。
具体地，在通过预设的卷积神经网络对样本拼接向量进行特征提取，得到至少一个特征提取向量之后，通过预设的卷积神经网络的池化层对各特征提取向量进行池化处理，进一步降低参数数量，从而降低计算复杂度，并将池化处理后的各特征提取向量进行第二拼接处理，得到特征拼接向量。
通过预设的卷积神经网络的全连接层对特征拼接向量进行分类识别，得到样本分类结果。
具体地，在通过预设的卷积神经网络的池化层对各特征提取向量进行池化处理后，将池化处理后的各特征提取向量进行第二拼接处理，得到特征拼接向量之后，将特征拼接向 量输入至全连接层(也即softmax层)，对特征拼接向量进行分类识别，得到样本分类结果。
在一实施例中，所述根据预设的知识数据库，对日常样本数据集中的不合理信息进行清洗处理，包括如下步骤：
根据样本分类结果，获取日常样本数据集中每一样本健康信息以及与其对应的样本对象特征之间的匹配度。
具体地，在将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果之后，根据样本分类结果，确定日常样本数据集中样本健康信息与对应的样本对象特征之间是否匹配，进而确定每一样本健康信息和与其关联的样本对象的样本对象特征之间的匹配度。示例性地，假设一个样本对象的样本对象特征为50-60岁且患有糖尿病，而对应的样本健康信息中包括运动信息为中等强度运动，或者饮食信息中没有提及不能吃太多与糖相关的食物时，表征运动推荐的不是最适合该样本对象特征的，且饮食推荐不够完善，进而该样本健康信息与对应的样本对象特征的匹配度较低。
将匹配度低于预设匹配阈值的样本对象特征以及样本健康信息记录为不合理信息，以对不合理信息进行清洗处理。
具体地，在根据样本分类结果，获取日常样本数据集中每一样本健康信息和与其关联的样本对象的样本对象特征之间的匹配度之后，将匹配度低于预设匹配阈值，且相互关联的样本对象特征以及所述样本健康信息记录为不合理信息，以对不合理信息进行清洗处理。其中，预设匹配阈值可以为80％，85％或者90％等，该预设匹配阈值可以根据面向的目标对象的不同群体进行变换。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
在一实施例中，提供一种基于知识图谱的健康信息推荐装置，该基于知识图谱的健康信息推荐装置与上述实施例中基于知识图谱的健康信息推荐方法一一对应。如图6所示，该基于知识图谱的健康信息推荐装置包括特征信息获取模块11、分类标签确定模块12、样本三元组获取模块13和健康信息推荐模块14。各功能模块详细说明如下：
特征信息获取模块11，用于获取目标对象的目标特征信息。所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息。
分类标签确定模块12，用于将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签。
样本三元组获取模块13，用于自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组。所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成。
健康信息推荐模块14，用于自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
优选地，如图7所示，基于知识图谱的健康信息推荐装置还包括如下模块：
数据获取模块21，用于获取日常样本数据集以及根据知识图谱构建的预设的知识数据库；所述日常样本数据集中包含至少一个样本健康信息以及与所述样本健康信息一一对应关联的样本对象的样本对象特征。
数据清洗模块22，用于根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理；所述不合理信息是指相互关联且不匹配的所述样本对象特征以及所述样本健康信息。
分类标签生成模块23，用于根据清洗处理之后的所述日常样本数据集中的样本对象特 征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联。
数据库构建模块24，用于根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库。
优选地，如图8所示，数据库构建模块包括如下单元：
实体关系构建单元241，用于根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系。
支持频率获取单元242，用于获取每一所述样本实体关系在所述预设的健康推荐数据库的支持频率。
第一优先级提升单元243，用于在所述实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据所述实体关系的支持频率提升所述实体关系的推荐优先级。
推荐值确定单元244，用于根据提升之后的所述推荐优先级以及预设的推荐算法确定所述实体关系中的所述样本健康信息的所述推荐值。
数据库构建单元245，用于将相互对应的所述推荐值、所述样本分类标签以及所述样本健康信息关联为一个样本三元组之后，根据所述样本三元组构建所述预设的健康推荐数据库。
优选地，如图9所示，数据库构建模块还包括如下单元：
置信度获取单元246，用于获取每一所述样本实体关系在所述预设的健康推荐数据库的置信度。
第二优先级提升单元247，用于在所述实体关系的置信度大于或等于预设置信度阈值时，按照预设的置信优先级规则，根据所述实体关系的置信度提升所述实体关系的推荐优先级。
优选地，基于知识图谱的健康信息推荐装置还包括如下模块：
样本数据集获取模块，用于获取知识样本数据集，所述知识样本数据集包含至少一个知识样本数据。
样本实体提取模块，用于提取所述知识样本数据中的所有样本实体，并根据所述知识样本数据中各数据与每一所述样本实体之间的距离，获取与提取的各所述样本实体关联的位置编码向量。
样本特征识别模块，用于对所述知识样本数据进行特征识别，得到所述知识样本数据对应的样本特征向量。
样本分类结果生成模块，用于将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，所述样本分类结果表征了至少两个所述样本实体之间的亲密度。
优选地，样本分类结果生成模块包括：
第一拼接单元，用于在对将所述样本特征向量以及所述位置编码向量进行第一拼接处理得到样本拼接向量之后，将所述样本拼接向量输入至预设的卷积神经网络中，通过所述预设的卷积神经网络对所述样本特征向量以及各所述位置编码向量进行第一拼接处理，得到至少一个样本拼接向量。
特征提取单元，用于将所述拼接向量输入至通过所述预设的卷积神经网络中，对所述样本拼接向量进行特征提取，得到至少一个特征提取向量。
第二拼接单元，用于在通过所述预设的卷积神经网络的池化层对各所述特征提取向量进行池化处理后，将池化处理后的各所述特征提取向量进行第二拼接处理，得到特征拼接向量。
样本分类结果生成单元，用于通过所述预设的卷积神经网络的全连接层对所述特征拼接向量进行分类识别，得到所述样本分类结果。
优选地，数据清洗模块还包括如下单元：
匹配度获取单元，用于根据所述样本分类结果，获取所述日常样本数据集中每一样本健康信息和与其关联的样本对象的样本对象特征之间的匹配度；
数据清洗单元，用于将所述匹配度低于预设匹配阈值，且相互关联的所述样本对象特征以及所述样本健康信息记录为不合理信息，以对所述不合理信息进行清洗处理。
关于基于知识图谱的健康信息推荐装置的具体限定可以参见上文中对于基于知识图谱的健康信息推荐方法的限定，在此不再赘述。上述基于知识图谱的健康信息推荐装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图10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，该计算机设备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可读存储介质、内存储器。该可读存储介质存储有操作系统、计算机可读指令和数据库。该内存储器为可读存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的数据库用于存储上述实施例中的基于知识图谱的健康信息推荐方法中使用到的数据。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。该计算机可读指令被处理器执行时以实现一种基于知识图谱的健康信息推荐方法。本实施例所提供的可读存储介质包括非易失性可读存储介质和易失性可读存储介质。
在一个实施例中，提供了一种计算机设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令，所述处理器执行所述计算机可读指令时实现如下步骤：
获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
自获取的所有所述样本三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象推送与其关联的所述样本健康信息。
在一个实施例中，提供了一个或多个存储有计算机可读指令的可读存储介质，本实施例所提供的可读存储介质包括非易失性可读存储介质和易失性可读存储介质；该可读存储介质上存储有计算机可读指令，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行如下步骤：
获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
自获取的所有所述样本三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象推送与其关联的所述样本健康信息。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令可存储于一非易失性计算机可读取存储介质或者易失性计算机可读取存储介质中，该计算机可读指令在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。

Claims (20)

  1. 一种基于知识图谱的健康信息推荐方法，其中，包括：
    获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
    将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
    自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
    自获取的所有所述样本三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象推送与其关联的所述样本健康信息。
  2. 如权利要求1所述的基于知识图谱的健康信息推荐方法，其中，所述自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组之前，还包括：
    获取日常样本数据集以及根据知识图谱构建的预设的知识数据库；所述日常样本数据集中包含至少一个样本健康信息以及与所述样本健康信息一一对应关联的样本对象的样本对象特征；
    根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理；所述不合理信息是指相互关联且不匹配的所述样本对象特征以及所述样本健康信息；
    根据清洗处理之后的所述日常样本数据集中的样本对象特征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联；
    根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库。
  3. 如权利要求2所述的基于知识图谱的健康信息推荐方法，其中，所述根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库，包括：
    根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系；
    获取每一所述样本实体关系在所述预设的健康推荐数据库的支持频率；
    在所述样本实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据所述样本实体关系的支持频率提升所述样本实体关系的推荐优先级；
    根据提升之后的所述推荐优先级以及预设的推荐算法确定所述样本实体关系中的所述样本健康信息的所述推荐值；
    将相互对应的所述推荐值、所述样本分类标签以及所述样本健康信息关联为一个样本三元组之后，根据所述样本三元组构建所述预设的健康推荐数据库。
  4. 如权利要求3所述的基于知识图谱的健康信息推荐方法，其中，所述根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系之后，还包括：
    获取每一所述样本实体关系在所述预设的健康推荐数据库的置信度；
    在所述样本实体关系的置信度大于或等于预设置信度阈值时，按照预设的置信优先级规则，根据所述样本实体关系的置信度提升所述样本实体关系的推荐优先级。
  5. 如权利要求2所述的基于知识图谱的健康信息推荐方法，其中，所述获取日常样本数据集以及根据知识图谱构建的预设的知识数据库之前，还包括：
    获取知识样本数据集，所述知识样本数据集包含至少一个知识样本数据；
    提取所述知识样本数据中的所有样本实体，并根据所述知识样本数据中各数据与每一所述样本实体之间的距离，获取与提取的各所述样本实体关联的位置编码向量；
    对所述知识样本数据进行特征识别，得到所述知识样本数据对应的样本特征向量；
    将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，所述样本分类结果表征了至少两个所述样本实体之间的亲密度。
  6. 如权利要求5所述的基于知识图谱的健康信息推荐方法，其中，所述将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，包括：
    在对所述样本特征向量以及所述位置编码向量进行第一拼接处理得到样本拼接向量之后，将所述样本拼接向量输入至预设的卷积神经网络中，通过所述预设的卷积神经网络对所述样本拼接向量进行特征提取，得到至少一个特征提取向量；
    通过所述预设的卷积神经网络的池化层对各所述特征提取向量进行池化处理后，将池化处理后的各所述特征提取向量进行第二拼接处理，得到特征拼接向量；
    通过所述预设的卷积神经网络的全连接层对所述特征拼接向量进行分类识别，得到所述样本分类结果。
  7. 如权利要求5所述的基于知识图谱的健康信息推荐方法，其中，所述根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理，包括：
    根据所述样本分类结果，获取所述日常样本数据集中每一样本健康信息以及与其对应的样本对象特征之间的匹配度；
    将所述匹配度低于预设匹配阈值的所述样本对象特征以及所述样本健康信息记录为不合理信息，并对所述不合理信息进行清洗处理。
  8. 一种基于知识图谱的健康信息推荐装置，其中，包括：
    特征信息获取模块，用于获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
    分类标签确定模块，用于将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
    样本三元组获取模块，用于自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
    健康信息推荐模块，用于自获取的所有所述三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象的移动终端推送与其关联的所述样本健康信息。
  9. 一种计算机设备，包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令，其中，所述处理器执行所述计算机可读指令时实现如下步骤：
    获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
    将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
    自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
    自获取的所有所述样本三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象推送与其关联的所述样本健康信息。
  10. 如权利要求9所述的计算机设备，其中，所述自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组之前，所述处理器执行所述计算机可读指令时还实现如下步骤：
    获取日常样本数据集以及根据知识图谱构建的预设的知识数据库；所述日常样本数据集中包含至少一个样本健康信息以及与所述样本健康信息一一对应关联的样本对象的样 本对象特征；
    根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理；所述不合理信息是指相互关联且不匹配的所述样本对象特征以及所述样本健康信息；
    根据清洗处理之后的所述日常样本数据集中的样本对象特征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联；
    根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库。
  11. 如权利要求10所述的计算机设备，其中，所述根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库，包括：
    根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系；
    获取每一所述样本实体关系在所述预设的健康推荐数据库的支持频率；
    在所述样本实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据所述样本实体关系的支持频率提升所述样本实体关系的推荐优先级；
    根据提升之后的所述推荐优先级以及预设的推荐算法确定所述样本实体关系中的所述样本健康信息的所述推荐值；
    将相互对应的所述推荐值、所述样本分类标签以及所述样本健康信息关联为一个样本三元组之后，根据所述样本三元组构建所述预设的健康推荐数据库。
  12. 如权利要求11所述的计算机设备，其中，所述根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系之后，所述处理器执行所述计算机可读指令时还实现如下步骤：
    获取每一所述样本实体关系在所述预设的健康推荐数据库的置信度；
    在所述样本实体关系的置信度大于或等于预设置信度阈值时，按照预设的置信优先级规则，根据所述样本实体关系的置信度提升所述样本实体关系的推荐优先级。
  13. 如权利要求10所述的计算机设备，其中，所述获取日常样本数据集以及根据知识图谱构建的预设的知识数据库之前，所述处理器执行所述计算机可读指令时还实现如下步骤：
    获取知识样本数据集，所述知识样本数据集包含至少一个知识样本数据；
    提取所述知识样本数据中的所有样本实体，并根据所述知识样本数据中各数据与每一所述样本实体之间的距离，获取与提取的各所述样本实体关联的位置编码向量；
    对所述知识样本数据进行特征识别，得到所述知识样本数据对应的样本特征向量；
    将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，所述样本分类结果表征了至少两个所述样本实体之间的亲密度。
  14. 如权利要求13所述的计算机设备，其中，所述将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，包括：
    在对所述样本特征向量以及所述位置编码向量进行第一拼接处理得到样本拼接向量之后，将所述样本拼接向量输入至预设的卷积神经网络中，通过所述预设的卷积神经网络对所述样本拼接向量进行特征提取，得到至少一个特征提取向量；
    通过所述预设的卷积神经网络的池化层对各所述特征提取向量进行池化处理后，将池化处理后的各所述特征提取向量进行第二拼接处理，得到特征拼接向量；
    通过所述预设的卷积神经网络的全连接层对所述特征拼接向量进行分类识别，得到所述样本分类结果。
  15. 如权利要求13所述的计算机设备，其中，所述根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理，包括：
    根据所述样本分类结果，获取所述日常样本数据集中每一样本健康信息以及与其对应的样本对象特征之间的匹配度；
    将所述匹配度低于预设匹配阈值的所述样本对象特征以及所述样本健康信息记录为不合理信息，并对所述不合理信息进行清洗处理。
  16. 一个或多个存储有计算机可读指令的可读存储介质，其中，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行如下步骤：
    获取目标对象的目标特征信息；所述目标对象为请求推送健康信息的用户；所述目标特征信息指的是目标对象的个体特征信息；
    将所述目标特征信息输入至预设的健康特征相似度模型中，得到与所述目标特征信息对应的健康分类标签；
    自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组；所述样本三元组是由样本分类标签、样本健康信息以及与样本健康信息和样本分类标签均关联的推荐值关联构成；
    自获取的所有所述样本三元组中提取样本健康信息以及与所述样本健康信息关联的推荐值，根据所述推荐值向所述目标对象推送与其关联的所述样本健康信息。
  17. 如权利要求16所述的可读存储介质，其中，所述自基于知识图谱构建的预设的健康推荐数据库中，获取具有与所述健康分类标签相同的样本分类标签的所有样本三元组之前，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器还执行如下步骤：
    获取日常样本数据集以及根据知识图谱构建的预设的知识数据库；所述日常样本数据集中包含至少一个样本健康信息以及与所述样本健康信息一一对应关联的样本对象的样本对象特征；
    根据所述预设的知识数据库，对所述日常样本数据集中的不合理信息进行清洗处理；所述不合理信息是指相互关联且不匹配的所述样本对象特征以及所述样本健康信息；
    根据清洗处理之后的所述日常样本数据集中的样本对象特征生成样本分类标签，并将所述样本分类标签和与其对应的样本健康信息关联；
    根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库。
  18. 如权利要求17所述的可读存储介质，其中，所述根据所述样本分类标签、所述样本健康信息以及预设的推荐算法构建所述预设的健康推荐数据库，包括：
    根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系；
    获取每一所述样本实体关系在所述预设的健康推荐数据库的支持频率；
    在所述样本实体关系的支持频率大于或等于预设支持度阈值时，按照预设的频率优先级规则，根据所述样本实体关系的支持频率提升所述样本实体关系的推荐优先级；
    根据提升之后的所述推荐优先级以及预设的推荐算法确定所述样本实体关系中的所述样本健康信息的所述推荐值；
    将相互对应的所述推荐值、所述样本分类标签以及所述样本健康信息关联为一个样本三元组之后，根据所述样本三元组构建所述预设的健康推荐数据库。
  19. 如权利要求18所述的可读存储介质，其中，所述根据所述样本分类标签以及与其对应的所述样本健康信息构建样本实体关系之后，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器还执行如下步骤：
    获取每一所述样本实体关系在所述预设的健康推荐数据库的置信度；
    在所述样本实体关系的置信度大于或等于预设置信度阈值时，按照预设的置信优先级规则，根据所述样本实体关系的置信度提升所述样本实体关系的推荐优先级。
  20. 如权利要求17所述的可读存储介质，其中，所述获取日常样本数据集以及根据知识图谱构建的预设的知识数据库之前，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器还执行如下步骤：
    获取知识样本数据集，所述知识样本数据集包含至少一个知识样本数据；
    提取所述知识样本数据中的所有样本实体，并根据所述知识样本数据中各数据与每一所述样本实体之间的距离，获取与提取的各所述样本实体关联的位置编码向量；
    对所述知识样本数据进行特征识别，得到所述知识样本数据对应的样本特征向量；
    将所述样本特征向量以及所有所述位置编码向量输入至预设的卷积神经网络中，得到样本分类结果，所述样本分类结果表征了至少两个所述样本实体之间的亲密度。
PCT/CN2020/117090 2020-07-28 2020-09-23 基于知识图谱的健康信息推荐方法、装置、设备及介质 WO2021135429A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010737217.6A CN111883228B (zh) 2020-07-28 2020-07-28 基于知识图谱的健康信息推荐方法、装置、设备及介质
CN202010737217.6 2020-07-28

Publications (1)

Publication Number Publication Date
WO2021135429A1 true WO2021135429A1 (zh) 2021-07-08

Family

ID=73200785

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/117090 WO2021135429A1 (zh) 2020-07-28 2020-09-23 基于知识图谱的健康信息推荐方法、装置、设备及介质

Country Status (2)

Country Link
CN (1) CN111883228B (zh)
WO (1) WO2021135429A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114021023A (zh) * 2021-12-08 2022-02-08 东南大学 一种基于图卷积神经网络的电量视角推荐方法及系统
CN114400071A (zh) * 2022-01-19 2022-04-26 平安国际智慧城市科技股份有限公司 饮食数据管理方法、相关设备及介质
CN117954036A (zh) * 2024-03-26 2024-04-30 青岛益生康健科技股份有限公司 基于大数据的健康管理方法及系统
CN118070130A (zh) * 2024-04-19 2024-05-24 北京东方通科技股份有限公司 一种消息中间件数据转换方法及系统

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112992367B (zh) * 2021-03-23 2021-09-28 微脉技术有限公司 基于大数据的智慧医疗交互方法及智慧医疗云计算系统
CN113241147B (zh) * 2021-04-28 2023-03-28 厦门艾地运动科技有限公司 健身计划生成方法、装置及电子设备
CN113239203A (zh) * 2021-06-02 2021-08-10 北京金山数字娱乐科技有限公司 一种基于知识图谱的筛选方法及装置
CN113359682B (zh) * 2021-06-30 2022-12-02 西安力传智能技术有限公司 设备故障预测方法、装置、设备故障预测平台及介质
CN113689929B (zh) * 2021-08-25 2023-05-26 深圳平安智慧医健科技有限公司 药物信息推送方法、装置、计算机设备及存储介质
CN113823410A (zh) * 2021-09-24 2021-12-21 联仁健康医疗大数据科技股份有限公司 一种信息推荐方法、装置、存储介质及电子设备
CN113988214A (zh) * 2021-11-15 2022-01-28 平安科技（深圳）有限公司 一种基于语音识别结果的相似用户推荐方法及装置
CN114021672B (zh) * 2022-01-06 2022-04-22 苏州百孝医疗科技有限公司 标签生成方法和装置、监控血糖水平的系统
CN114974508B (zh) * 2022-05-20 2024-05-24 西南大学 一种运动处方的生成装置及电子设备
CN114998816B (zh) * 2022-08-08 2022-11-01 深圳市指南针医疗科技有限公司 基于骨骼ai视频的病例改进方法、装置及存储介质
CN118352081A (zh) * 2024-04-17 2024-07-16 深圳市儿童医院 一种基于数据处理的肥胖干预效果评估方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110674312A (zh) * 2019-09-18 2020-01-10 泰康保险集团股份有限公司 构建知识图谱方法、装置、介质及电子设备
CN110866093A (zh) * 2018-08-10 2020-03-06 珠海格力电器股份有限公司 机器问答方法及装置
CN111177403A (zh) * 2019-12-16 2020-05-19 恩亿科(北京)数据科技有限公司 样本数据的处理方法和装置
CN111191020A (zh) * 2019-12-27 2020-05-22 江苏省人民医院（南京医科大学第一附属医院） 基于机器学习和知识图谱的处方推荐方法和系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110310721B (zh) * 2018-03-27 2023-05-02 九阳股份有限公司 一种基于知识图谱构建食谱推荐策略的方法及装置
CN108965416B (zh) * 2018-07-04 2022-04-15 平安科技（深圳）有限公司 医疗数据共享方法、装置、计算机设备和存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110866093A (zh) * 2018-08-10 2020-03-06 珠海格力电器股份有限公司 机器问答方法及装置
CN110674312A (zh) * 2019-09-18 2020-01-10 泰康保险集团股份有限公司 构建知识图谱方法、装置、介质及电子设备
CN111177403A (zh) * 2019-12-16 2020-05-19 恩亿科(北京)数据科技有限公司 样本数据的处理方法和装置
CN111191020A (zh) * 2019-12-27 2020-05-22 江苏省人民医院（南京医科大学第一附属医院） 基于机器学习和知识图谱的处方推荐方法和系统

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114021023A (zh) * 2021-12-08 2022-02-08 东南大学 一种基于图卷积神经网络的电量视角推荐方法及系统
CN114400071A (zh) * 2022-01-19 2022-04-26 平安国际智慧城市科技股份有限公司 饮食数据管理方法、相关设备及介质
CN114400071B (zh) * 2022-01-19 2024-05-14 平安国际智慧城市科技股份有限公司 饮食数据管理方法、相关设备及介质
CN117954036A (zh) * 2024-03-26 2024-04-30 青岛益生康健科技股份有限公司 基于大数据的健康管理方法及系统
CN117954036B (zh) * 2024-03-26 2024-06-07 青岛益生康健科技股份有限公司 基于大数据的健康管理方法及系统
CN118070130A (zh) * 2024-04-19 2024-05-24 北京东方通科技股份有限公司 一种消息中间件数据转换方法及系统

Also Published As

Publication number Publication date
CN111883228A (zh) 2020-11-03
CN111883228B (zh) 2023-07-07

Similar Documents

Publication Publication Date Title
WO2021135429A1 (zh) 基于知识图谱的健康信息推荐方法、装置、设备及介质
WO2021073277A1 (zh) 一种个性化精准用药推荐方法及装置
CN113707297B (zh) 医疗数据的处理方法、装置、设备及存储介质
Rubaiat et al. Important feature selection & accuracy comparisons of different machine learning models for early diabetes detection
US10430716B2 (en) Data driven featurization and modeling
WO2021139106A1 (zh) 分群决策模型生成、分群处理方法、装置、设备及介质
US20210035693A1 (en) Methods, systems, and apparatuses for predicting the risk of hospitalization
Rudovic et al. Personalized federated deep learning for pain estimation from face images
WO2021151358A1 (zh) 基于解释模型的分诊信息推荐方法、装置、设备及介质
Yan et al. Concurrent healthcare data processing and storage framework using deep-learning in distributed cloud computing environment
US20180151254A1 (en) High-speed similar case search method and device through reduction of large scale multi-dimensional time series health data to multiple dimensions
Gupta et al. Concurrent imputation and prediction on EHR data using bi-directional GANs: Bi-GANs for EHR imputation and prediction
US20220391760A1 (en) Combining model outputs into a combined model output
US11710572B2 (en) Experience engine-method and apparatus of learning from similar patients
WO2023240837A1 (zh) 基于病患数据的服务包生成方法、装置、设备及存储介质
Bashir et al. WebMAC: A web based clinical expert system
Saravagi et al. [Retracted] Diagnosis of Lumbar Spondylolisthesis Using a Pruned CNN Model
CN114743647A (zh) 医疗数据处理方法、装置、设备及存储介质
Alwan et al. Diabetes diagnosis system using modified Naive Bayes classifier
WO2020132918A1 (zh) 药品预测方法、装置、计算机设备及存储介质
US12087442B2 (en) Methods and systems for confirming an advisory interaction with an artificial intelligence platform
Zhang et al. Predicting treatment initiation from clinical time series data via graph-augmented time-sensitive model
CN114758743A (zh) 信息预测方法、装置、存储介质及计算机设备
CN114822849A (zh) 基于数字孪生的数据监测方法、装置、设备和存储介质
KR20180062353A (ko) 대용량 다차원 시계열 건강 데이터의 복합 차원으로의 축소를 통한 고속 유사 사례 검색 방법 및 그 장치

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: 20908817

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: 20908817

Country of ref document: EP

Kind code of ref document: A1