CN114417138B - Health information recommendation method and equipment - Google Patents
Health information recommendation method and equipment Download PDFInfo
- Publication number
- CN114417138B CN114417138B CN202111612594.8A CN202111612594A CN114417138B CN 114417138 B CN114417138 B CN 114417138B CN 202111612594 A CN202111612594 A CN 202111612594A CN 114417138 B CN114417138 B CN 114417138B
- Authority
- CN
- China
- Prior art keywords
- recommendation
- user
- matrix
- information
- round
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 230000036541 health Effects 0.000 title claims abstract description 236
- 238000000034 method Methods 0.000 title claims abstract description 120
- 239000011159 matrix material Substances 0.000 claims abstract description 328
- 230000008569 process Effects 0.000 claims abstract description 84
- 230000008859 change Effects 0.000 claims abstract description 64
- 239000013598 vector Substances 0.000 claims description 24
- 238000012216 screening Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 7
- 238000010586 diagram Methods 0.000 description 10
- 238000004590 computer program Methods 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 238000003860 storage Methods 0.000 description 6
- 238000000354 decomposition reaction Methods 0.000 description 5
- 239000008280 blood Substances 0.000 description 4
- 210000004369 blood Anatomy 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 239000000203 mixture Substances 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 235000005911 diet Nutrition 0.000 description 2
- 230000037213 diet Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000008103 glucose Substances 0.000 description 2
- 230000005802 health problem Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 210000000577 adipose tissue Anatomy 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000036772 blood pressure Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 235000005955 light diet Nutrition 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 229940050561 matrix product Drugs 0.000 description 1
- 235000013336 milk Nutrition 0.000 description 1
- 239000008267 milk Substances 0.000 description 1
- 210000004080 milk Anatomy 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
Classifications
-
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/906—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a health information recommendation method and equipment, wherein the method comprises the following steps: acquiring target data of a target object, wherein the target data comprises user characteristic data of each user of the target object; determining the category to which the target object belongs according to the target data; determining a target recommendation model corresponding to the category in the recommendation model set, and determining health information matched with the target object according to the target recommendation model; in the process of determining each recommendation model in the recommendation model set, taking a model constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting a preset stop condition as a recommendation model through multiple rounds of updating; in the updating process of each round, the health management terminal and the server interact with each other to form a change gradient of the recommended information matrix; and recommending the health information to each user in the target object. On the premise of not depending on the historical health information of the user, accurate health information is recommended to the user.
Description
Technical Field
The invention relates to the technical field of intelligent health management, in particular to a health information recommendation method and equipment.
Background
Along with the development of science and technology and the improvement of living standard of people, the health problem is more and more important. People hope to acquire health advice and the like at any time and any place in life so as to discover own problems and improve the problems in time. With the popularization of intelligent equipment, people can timely learn various body indexes such as sleeping conditions, blood sugar and the like.
In the related art, when health advice is provided for a user according to body indexes, in one case, the body indexes of the user need to be acquired to perform corresponding recommendation, so that the body indexes of the user are leaked; in another case, the history related health advice of the user needs to be obtained and then the targeted recommendation is performed, and the situation is that the history related health advice is excessively depended on and the history related health advice of the user is also revealed.
Disclosure of Invention
The invention provides a health information recommending method and equipment, which are used for recommending accurate health information for a user on the premise of not depending on historical health information of the user.
According to a first aspect in an exemplary embodiment, there is provided a health information recommendation method, the method comprising:
acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
Determining the category to which the target object belongs according to the target data;
determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model;
in the process of determining each SVD recommendation model in the SVD recommendation model set, using models constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting preset stop conditions as SVD recommendation models through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between recommendation information and an object; the recommendation information matrix characterizes a second association relationship between recommendation information and an object; the recommended information is health information;
and recommending the health information to each user in the target object.
According to the method, user characteristic data of each user in the target object are acquired first, then the category to which the target object belongs is determined based on the user characteristic data, in this way, in the SVD recommendation model set, a target SVD recommendation model corresponding to the category is determined, and health information matched with the target object is determined according to the target SVD recommendation model. Because the historical health information and the physical index information of the user are not needed in the process, the determined health information protects the privacy of the user to the greatest extent. In addition, in the process of determining each SVD recommendation model, the object factor matrix and the recommendation information matrix representing the relation between recommendation information and objects are involved in processing, and in the updating process of each round, the health management terminal and the server transmit the change gradient between the recommendation information matrices of two adjacent rounds instead of the data, so that the user privacy is also protected and the data safety is ensured on the aspect. In addition, the SVD recommendation model is applied to obtain the matching of the health information and the simple corresponding relation, so that the matching is more accurate.
In some exemplary embodiments, the set of SVD recommendation models is determined by:
determining object data of each object;
classifying each of the objects according to each of the object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
the SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of each object in the category to the nth round and the recommendation information matrix of the nth round; the object factor matrix is determined according to object data of a corresponding object and health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all the health management terminals determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
Receiving a second change gradient of the (N+1) th recommended information matrix sent by the health management terminal relative to the (N) th recommended information matrix; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (n+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
The above embodiment is a process of determining an SVD recommendation model, in which the health management terminal and the server transmit a gradient of change of the recommendation information matrix, and in the process of updating the object factor matrix and the recommendation information matrix, the two are interdependent, so that the finally determined SVD recommendation model covers health information most conforming to the object.
In some exemplary embodiments, the recommending the health information to each user in the target object includes:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
and recommending the selected health information to the user corresponding to the user identifier.
In the above embodiment, because the ages or sexes of different family members in the same family may be different, in order to implement accurate recommendation, after determining the health information corresponding to the family, the user identifier is used to recommend the health information as the responding member user. And when a plurality of recommended information exists, screening can be performed according to the weight, and the accuracy is higher.
In some exemplary embodiments, the acquiring the target data of the target object includes:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
If the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of a second preset object.
In the above embodiment, when the target data of the target object cannot be directly obtained, the user feature data of the newly added user or the object data of the newly added family may be determined in the corresponding preset object according to different situations. The situation that recommendation is not possible due to accidental factors is avoided.
In some exemplary embodiments, the first variation gradient matrix is formed by combining first variation gradient vectors of respective objects in the class, and the respective first variation vectors are determined by the following formula:
wherein f (j, m) is a first variation vector, m represents the mth recommendation information, j represents the jth object in the category, X j Representing the object factor matrix in the N-1 round of updating process, Y (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; wherein N is an integer greater than 1.
In the above embodiment, the information obtained by calculating the object factor matrix and the recommended information matrix of the previous round is covered in the change gradient, so that the related information of the object factor matrix and the recommended information matrix of the same round is covered in the new round of recommended information matrix which is reversely deduced after the change gradient is transmitted. In this way, the determined SVD recommendation model is more closely related to the actual needs of the subject.
In some exemplary embodiments, the object factor matrix for the nth round is determined by the following formula:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1.
In the above embodiment, the recommendation information matrix of the nth round is integrated in the object factor matrix of the nth round, so that the recommendation information matrix can be gradually optimized in the whole updating process, and the recommendation information matrix which best meets the object condition is obtained. And further obtaining health information to be recommended.
According to a second aspect in an exemplary embodiment, there is provided a health information recommendation device comprising a processor and a data transmission unit, wherein:
the data transmission unit is configured to perform:
transmitting target data of a target object to the processor;
the processor is configured to perform:
acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
determining the category to which the target object belongs according to the target data;
Determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model;
in the process of determining each SVD recommendation model in the SVD recommendation model set, using models constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting preset stop conditions as SVD recommendation models through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between recommendation information and an object; the recommendation information matrix characterizes a second association relationship between recommendation information and an object; the recommended information is health information;
and recommending the health information to each user in the target object.
In some exemplary embodiments, the processor is configured to perform determining the set of SVD recommendation models by:
Determining object data of each object;
classifying each of the objects according to each of the object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
the SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of each object in the category to the nth round and the recommendation information matrix of the nth round; the object factor matrix is determined according to object data of a corresponding object and health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all the health management terminals determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
receiving a second change gradient of the (N+1) th recommended information matrix sent by the health management terminal relative to the (N) th recommended information matrix; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (n+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
Stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
In some exemplary embodiments, the processor is configured to perform:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
and recommending the selected health information to the user corresponding to the user identifier.
In some exemplary embodiments, the processor is configured to perform:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of a second preset object.
In some exemplary embodiments, the first variation gradient matrix is a combination of first variation gradient vectors for each object in the class, and the processor is configured to determine each first variation vector by:
wherein f (j, m) is a first variation vector, m represents the mth recommendation information, j represents the jth object in the category, X j Representing the object factor matrix in the N-1 round of updating process, Y (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; wherein N is an integer greater than 1.
In some exemplary embodiments, the processor is configured to determine the object factor matrix for the nth round by:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1.
According to a third aspect in an exemplary embodiment, there is provided a health information recommendation apparatus, the apparatus comprising:
the data acquisition module is used for acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
The category determining module is used for determining the category to which the target object belongs according to the target data;
the matching module is used for determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set and determining health information matched with the target object according to the target SVD recommendation model;
in the process of determining each SVD recommendation model in the SVD recommendation model set, using models constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting preset stop conditions as SVD recommendation models through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between recommendation information and an object; the recommendation information matrix characterizes a second association relationship between recommendation information and an object; the recommended information is health information;
and the recommending module is used for recommending the health information to each user in the target object.
In some exemplary embodiments, the method further comprises a model determination module for determining the set of SVD recommendation models by:
determining object data of each object;
classifying each of the objects according to each of the object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
the SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of each object in the category to the nth round and the recommendation information matrix of the nth round; the object factor matrix is determined according to object data of a corresponding object and health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all the health management terminals determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
Receiving a second change gradient of the (N+1) th recommended information matrix sent by the health management terminal relative to the (N) th recommended information matrix; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (n+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
In some exemplary embodiments, the recommendation module is specifically configured to:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
and recommending the selected health information to the user corresponding to the user identifier.
In some exemplary embodiments, the data acquisition module is specifically configured to:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of a second preset object.
In some exemplary embodiments, the first gradient matrix is formed by combining first gradient vectors of the objects in the category, and the model determining module is specifically configured to determine each first gradient vector by the following formula:
wherein f (j, m) is a first variation vector, m represents the mth recommendation information, j represents the jth object in the category, X j Representing the object factor matrix in the N-1 round of updating process, Y (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; wherein N is an integer greater than 1.
In some exemplary embodiments, the object factor matrix for the nth round is determined based on the model determination module by:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1.
According to a fourth aspect in an exemplary embodiment, a computer storage medium is provided, in which computer program instructions are stored which, when run on a computer, cause the computer to perform the health information recommendation method according to the first aspect.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it will be apparent that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram schematically illustrating an application scenario of a health information recommendation method according to an embodiment of the present invention;
Fig. 2 schematically illustrates data collection of a health management terminal according to an embodiment of the present invention;
fig. 3 is a flowchart illustrating a health information recommendation method according to an embodiment of the present invention;
FIG. 4 is a flowchart illustrating another health information recommendation method according to an embodiment of the present invention;
fig. 5 schematically illustrates a structural diagram of a health information recommendation device according to an embodiment of the present invention;
fig. 6 schematically illustrates a structure of a health information recommendation device according to an embodiment of the present invention.
Detailed Description
The following description will be given in detail of the technical solutions in the embodiments of the present application with reference to the accompanying drawings. Wherein, in the description of the embodiments of the present application, "/" means or is meant unless otherwise indicated, for example, a/B may represent a or B; the text "and/or" is merely an association relation describing the associated object, and indicates that three relations may exist, for example, a and/or B may indicate: the three cases where a exists alone, a and B exist together, and B exists alone, and in addition, in the description of the embodiments of the present application, "plural" means two or more than two.
The terms "first," "second," and the like, are used below for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature, and in the description of embodiments of the present application, unless otherwise indicated, the meaning of "a plurality" is two or more.
(1) SVD (Singular Value Decomposition ) is an important matrix decomposition in linear algebra, and singular value decomposition is a generalization of feature decomposition on arbitrary matrices. The method is widely applied to recommendation systems and used for mining potential relation between users and recommended content.
Along with the development of science and technology and the improvement of living standard of people, the health problem is more and more important. People hope to acquire health advice and the like at any time and any place in life so as to discover own problems and improve the problems in time. With the popularization of intelligent equipment, people can timely learn various body indexes such as sleeping conditions, blood sugar and the like.
In the related art, when health advice is provided for a user according to body indexes, in one case, the body indexes of the user need to be acquired to perform corresponding recommendation, so that the body indexes of the user are leaked; in another case, the history related health advice of the user needs to be obtained and then the targeted recommendation is performed, and the situation is that the history related health advice is excessively depended on and the history related health advice of the user is also revealed.
For this purpose, the present application provides a health information recommendation method, in which target data of a target object is obtained, the target object is, for example, a family, one target object corresponds to at least one user, and the target data includes user characteristic data of each user. Determining a category to which the target object belongs according to the target data, determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model; and recommending the health information to each user in the target object. In the process, the user can be recommended with accurate health information only by the user characteristic data such as gender, age and the like without the need of historical health information of the user or body indexes of the user. In addition, in the determining process of each SVD recommendation model, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are all gradient changes between the recommendation information matrixes of two adjacent rounds, and the gradient changes are not the data, so that the safety of the data is further ensured.
After the design concept of the embodiment of the present application is introduced, some simple descriptions are made below for application scenarios applicable to the technical solution of the embodiment of the present application, and it should be noted that the application scenarios described below are only used to illustrate the embodiment of the present application and are not limiting. In specific implementation, the technical scheme provided by the embodiment of the application can be flexibly applied according to actual needs.
Referring to fig. 1, a schematic diagram of an application scenario of a health information recommendation method is shown, in the scenario, in a health information recommendation stage, when recommending a new user, user feature data of the new user is directly applied to match a corresponding SVD recommendation model, and a recommendation result of the SVD recommendation model is health information of the new user. In the determining stage of the SVD recommendation model, taking three target objects (families) as examples, each family includes 3 intelligent health devices, in addition, each family includes a health management terminal, the health management terminal can communicate with each intelligent health device in the family through the internet, for example, health index data from each intelligent health device is obtained, the health index data are summarized and then sent to a server, and accordingly the server determines each SVD recommendation model based on each health index data and user characteristic data of each user.
Taking one server 101 and 3 groups of intelligent health devices 102 and three home management terminals 103 (103-1, 103-2 and 103-3) as an example, wherein the intelligent health devices 102_1 of the home 1 comprise an intelligent mattress, an intelligent toilet and an intelligent dressing mirror; the intelligent health equipment of the family 2 comprises an intelligent sphygmomanometer, an intelligent blood glucose meter and an intelligent temperature measuring gun; the intelligent health equipment of the family 3 is an intelligent pedometer and an intelligent body fat scale. The intelligent health terminal 102 user obtains the data of each user in the home, and sends the data of the user in the home to the server through the corresponding home management terminal 103.
Referring to fig. 2, a schematic diagram of a process for collecting user data is shown; taking a family as an example, each intelligent health device sends the collected health index data of each member user to the health management terminal. Health index data such as: blood pressure, blood glucose, sleep conditions; health information is for example: diet information, exercise information, etc.
In order to further explain the technical solutions provided in the embodiments of the present application, the following details are described with reference to the accompanying drawings and the detailed description. Although the embodiments of the present application provide the method operational steps as shown in the following embodiments or figures, more or fewer operational steps may be included in the method based on routine or non-inventive labor. In steps where there is logically no necessary causal relationship, the execution order of the steps is not limited to the execution order provided by the embodiments of the present application.
The technical solution provided in the embodiments of the present application is described below with reference to an application scenario shown in fig. 1 and fig. 2, and a health information recommendation method shown in fig. 3.
S301, acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data includes user characteristic data of the respective users.
S302, determining the category to which the target object belongs according to the target data.
S303, determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model.
In the process of determining each SVD recommendation model in the SVD recommendation model set, the model constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting the preset stop condition is used as the SVD recommendation model through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between the recommendation information and the object; the recommendation information matrix characterizes a second association relationship between recommendation information and the object; the recommended information is health information;
S304, recommending the health information to each user in the target object.
According to the method and the device, user characteristic data of each user in the target object are acquired first, then the category to which the target object belongs is determined based on the user characteristic data, in this way, in the SVD recommendation model set, a target SVD recommendation model corresponding to the category is determined, and health information matched with the target object is determined according to the target SVD recommendation model. Because the historical health information and the physical index information of the user are not needed in the process, the determined health information protects the privacy of the user to the greatest extent. In addition, in the process of determining each SVD recommendation model, the object factor matrix and the recommendation information matrix representing the relation between recommendation information and objects are involved in processing, and in the updating process of each round, the health management terminal and the server transmit the change gradient between the recommendation information matrices of two adjacent rounds instead of the data, so that the user privacy is also protected and the data safety is ensured on the aspect. In addition, the SVD recommendation model is applied to obtain the matching of the health information and the simple corresponding relation, so that the matching is more accurate.
Referring to S301, a target object is, for example, a family, which corresponds to at least one user, and the target data includes user characteristic data of each user, such as age, sex, weight, and the like. In a specific example, the obtained target data of the target object are, for example, the ages, sexes, and weights of 3 members in family a. The acquisition process is manually entered, for example, by a user while using the intelligent health device, or actively entered when there is a need to recommend health information.
Specifically, the server can access the health management terminal regularly in a data acquisition mode, and the intelligent health terminal in each family acquires user data in real time through the intelligent health terminal and uploads the user data to the health management terminal. In the actual application process, an encryption module of an Embedding model is integrated in the intelligent health terminal, the module is pre-distributed to each intelligent health terminal by a server, and the encryption module is not updated in the use process, so that the possibility of cracking the encryption module can be reduced. The server obtains the object data of each object, which is encrypted by the intelligent health terminal, and decrypts the encrypted object data after receiving the encrypted object data, and then carries out the follow-up recommendation flow. It should be noted that, the interaction processes of the intelligent health terminal, the health management terminal and the server are all encrypted data, which is not described in detail later.
In the actual application process, the obtaining the target data of the target object may include the following two cases:
in the first case, if there is a new user in the target object and the user feature data of the new user is not acquired, determining the user feature data of the new user according to the user feature data of each user in the first preset object, and determining the target data of the target object.
Specifically, when a new user a is added in a family, but the user characteristic data of the user a is not acquired, for example, the user does not upload the user characteristic data of the user to the intelligent health terminal, or the intelligent health terminal cannot actively acquire the basic information of the user. At this time, the user characteristic data of the newly added user may be predicted from the user characteristic data of each user in the first preset object, where the first preset object is determined according to the common family composition. For example, the family composition of the first preset object includes dad, mom, child and grandma, the new member can be predicted according to the original member composition in the target object, and then the user characteristic data of the new user can be determined by combining with the first preset object, so that the user characteristic data of each member user of the target object can be obtained.
In the second case, the target object is a new object, and the user characteristic data of any user in the new object is not acquired, and the target data of the target object is determined according to the target data of the second preset object.
Specifically, a family is newly added, and at the same time, the user characteristic data of at least one user in the newly added family is not obtained, and the target data of the second preset object is directly used as the target data of the target object. The second preset object is for example a household in which a single male is living alone.
S302, since the target object is discussed in terms of family, the health information consistency of the same family is higher, for example, the health information can include specific advice of light diet for the old over 60 years old in the family. Therefore, in order to improve the accuracy of recommendation, the category to which the target object belongs may be determined from the target data.
In addition, the relation between different family groups and SVD recommendation models is pre-stored, namely, the SVD recommendation model set comprises a plurality of SVD recommendation models, each SVD recommendation model corresponds to one family group, and therefore a plurality of family groups exist. A specific implementation manner may be that a similarity between user feature data of each user in the target object and user feature data of each user in each family group is determined, and a category of a family group with the highest similarity is taken as a category to which the target object belongs. In a specific example, the determined category is, for example, a young child or an old person category.
Referring to S303, in the SVD recommendation model set, a target SVD recommendation model corresponding to the category is determined, and health information matched with the target object is determined according to the target SVD recommendation model.
Specifically, as multiple rounds of updating are performed in the process of obtaining the target SVD recommendation model, the model constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting the preset stop condition is used as the SVD recommendation model; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommended non-health information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between the recommendation information and the object; the recommendation information matrix characterizes a second association relationship between recommendation information and the object; the recommended information is health information. Therefore, the matched health information can be determined according to the user characteristic data of any target object.
The following description describes a process of determining the SVD recommendation model set:
specifically, determining object data of each object; classifying each object according to each object data; wherein the object data includes user characteristic data of a history user included in the corresponding object; and determining an SVD recommendation model corresponding to each category.
Exemplary, specific implementations of the classification process are as follows:
after the object data of each object is acquired, each object is classified based on each object data. For example, the server obtains object data of 100 objects (families), each family includes at least one user, and the classification process can enable families close to the family members to be classified into one class, and the family member close can refer to that the ages of the family members are relatively close.
In a specific example, the classification process may be implemented by a K-means clustering algorithm, or may be other clustering algorithms in the related art, which are not described herein. The classification result is, for example, a single family class, a young married child-free family class, a middle-aged married child family class, and an aged family class.
Thus, in this process, the classification process of the respective households is completed, and at least one household is included in each category. Thus, the embodiment of the invention takes the family as a unit, and the purpose of acquiring the clustering type is to reduce excessive interaction between the health management terminal and the server and reduce the possibility of data leakage. In addition, individual households of the same category typically have the same organization, such as the number of family members, etc., or dad, mom, grandpa, milk, and children in the household are similar to one another, i.e., may have similar behavioral characteristics.
It should be noted that, the object in the model determining process may be understood as a sample object, and the user refers to a user in the sample object, and there is no association between the user and the target object. The SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of the nth round by each object in the category and the recommendation information matrix of the nth round; the object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object; the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all health management terminals can determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient; receiving a second change gradient of the (N+1) th recommended information matrix relative to the (N) th recommended information matrix sent by the health management terminal; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (N+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round; stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions; and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
The object factor matrix characterizes a first association relationship between the recommendation information and the object; the recommendation information matrix characterizes a second association relationship between recommendation information and the object; the recommended information is health information. The first association and the second association can be found by the following formulas (1) to (3).
For example, description will be made of recommendation information, and in the foregoing recommendation scenario of family health information, the recommendation information may exist in a matrix form, which is called a recommendation information matrix. For example, when the elements in the recommendation information matrix are the contents to be recommended, such as the recommended contents are health articles, the recommendation information matrix represents the category of each article, such as sports, diet or treatment; but also information describing recommended content such as article length. In a specific example, the recommendation information matrix uses Y M And (3) representing the number of recommended contents in the recommended information required to be recommended, wherein the recommended information of the last round meeting the preset recommended condition is taken as target recommended information.
Because a category includes a plurality of families, each category corresponds to an SVD recommendation model, for the jth family in the category, in the specific recommendation process, for example, after the Nth round 1 is finished, the obtained object factor matrix is X (N-1)j The obtained recommendation information matrix is Y N-1 (m) The recommendation information matrix is only related to recommended information, and in the same category, the recommendation information matrix corresponding to each family is the same, so that the recommendation information matrix does not have a corner mark j. M is any recommended content, one piece of recommended information comprises M recommended contents, and the value of M is an integer from 1 to M. Since the processing procedure of each recommended content is the same, the recommended content m will be described as an example.
In this way, the first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round is determined firstly, specifically according to the object factor matrix in the updating process of the nth round, the interested parameters of each object in the category to the recommendation information of the nth round and the recommendation information matrix of the nth round.
The object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object; for example, X 0j The object factor matrix in the subsequent updating process also comprises the recommendation information of the round according to the object data of the corresponding object and the health data of each user corresponding to the object. Since the recommendation information is constituted by a plurality of recommendation contents, for the recommendation information matrix, each recommendation content is constituted by a recommendation vector, and thus, in determining the first variation gradient, it is also handled in units of vectors.
In a specific example, the first gradient matrix is formed by combining first gradient vectors of respective objects in the class, and the respective first gradient vectors are determined by the following formula:
wherein f (j, m) is a first variation vector, m represents the mth recommended content, j represents the jth object in the category, X (N-1)j Representing the object factor matrix in the N-1 round of updating process, Y N-1 (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; c jm Representing a confidence parameter for representing a degree of preference of the user j for uncertainty of the item m, which uncertainty can be measured based on the hyper-parameter α; p is p jm Indicating the preference degree of the jth family to the mth commodity; wherein N is an integer greater than 1.
According to the first change vector, the recommendation information matrix Y of the nth round can be obtained on the basis of the known recommendation information matrix of the nth round 1 N (m) The process is as follows:
wherein,for the change in the N-th round of updating, < >>Lambda is a constant in the training process
Based on the principle, the first change gradient is transmitted to the health management terminal, and the health management terminal is based on Y N-1 (m) On the basis of (a), Y can be obtained N (m) . In this way, in the data interaction process, the first change gradient is transmitted, which is an intermediate variable without practical meaning, so that the possibility of data leakage is avoided.
Next, the health management terminal determines a second change gradient of the n+1th recommendation information matrix relative to the nth recommendation information matrix in the same manner, and transmits the second change gradient to the server so that the server can continuously update the recommendation information matrix. The second gradient is determined according to the N-round object factor matrix, the interested parameter of the recommended information of each object in the category to the (N+1) -th round and the recommended information matrix of the (N) -th round. The determination process is the same as the process of determining the first gradient, and is not described in detail herein. After the second gradient of variation is determined, it is sent to the server so that the server continues to update the recommendation information matrix.
In this process, the update of the object factor matrix by the health management terminal is also involved. The object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the N-1 th round and the recommendation information matrix of the nth round.
Illustratively, the object factor matrix for the nth round is determined by the following formula:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1.
Thus, in the process of the model, the update process of the object factor matrix and the recommendation information matrix is included, and the object factor matrix and the recommendation information matrix are interdependent. In addition, in the transmission process, the change gradient of the transmitted recommended information matrix is combined with the change gradient and the recommended information matrix stored in the previous round to obtain the recommended information matrix of the current round.
And stopping updating until the obtained object factor matrixes and the recommendation information matrixes meet preset conditions. The preset condition is reflected in the training process to converge model parameters, and the convergence criterion is thatWherein (1)>Representing user data during any one recommendation, Y (m) And the recommendation information matrix in any recommendation process is represented, and only when the recommendation information matrix and the recommendation information matrix meet the convergence standard, the recommendation is ended, and the preset recommendation condition is determined to be met. X is X * ,Y * The symbol representations of the optimal object factor matrix and the recommendation information matrix after the recommendation is finished are respectively. This condition is also exemplified for any one of the recommended contents, when each of the recommended contents satisfies a preset condition, that is, the recommended information matrix satisfies the condition.
And finally, determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round. In a specific example, since a category includes a plurality of families, each family corresponds to an object factor matrix, and each recommendation information matrix is the same, the object factor matrices can be combined according to the family identifier, and then a matrix product operation is performed on the object factor matrices and the recommendation information matrix, so as to obtain a corresponding SVD recommendation model.
Referring to S304, since the health information is obtained in the unit of home, and from the perspective of intelligence, the health information needs to be recommended to each user in the home, and considering that the health information needed by different users in the same home is different, in this process, the user identification in the health information is identified; screening a preset number of health information according to the weight of each health information; the selected health information is recommended to the user corresponding to the user identification.
Specifically, the user identifier carried in the health information can be used for distinguishing each user in the family members, and when the health information is multiple, the weight of the health information can be determined according to the constitution of the family members, and the preset quantity of health information is screened out and recommended to the user corresponding to the user identifier.
In summary, in the interaction process of the server and the health management terminal, the gradient of f (j, m) is transmittedAnd the data security can be protected to the greatest extent by the method instead of the matrix, real data are not interacted, and only intermediate calculation results are interacted. In the matrix processing process, matrix singular value decomposition is used to obtain the relationship between the user and the recommendation information.
In order to make the technical solution of the present application more complete, fig. 4 shows a flowchart of a health information recommendation method, which at least includes the following steps:
s4011, target data of corresponding target objects are directly obtained from each health management terminal.
S4012, if a new user exists in the target object and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object.
S4013, if the target object is a new object and user characteristic data of any user in the new object is not acquired, determining target data of the target object according to target data of a second preset object.
Wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
s402, determining the category to which the target object belongs according to the target data.
S403, determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model.
S404, identifying the user identification in the health information.
S405, screening a preset number of health information according to the weight of each health information.
S406, recommending the selected health information to the user corresponding to the user identification.
The embodiment considers various situations of acquiring the target data of the target object, and avoids the situation that the target data cannot be acquired due to accidental factors. And when recommending the health information, the requirement difference of different users in the same family on the health information is considered, so that the user identification is used for distinguishing, and the accurate recommendation is realized.
As shown in fig. 5, based on the same inventive concept, an embodiment of the present invention provides a health information recommendation, including a data acquisition module 51, a category determination module 52, a matching module 53, and a recommendation module 54.
Wherein, the data acquisition module 51 is configured to acquire target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
a category determining module 52, configured to determine a category to which the target object belongs according to the target data;
the matching module 53 is configured to determine a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determine health information matched with the target object according to the target SVD recommendation model;
in the process of determining each SVD recommendation model in the SVD recommendation model set, the model constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting the preset stop condition is used as the SVD recommendation model through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between the recommendation information and the object; the recommendation information matrix characterizes a second association relationship between recommendation information and the object; the recommended information is health information;
And a recommending module 54, configured to recommend the health information to each user in the target object.
In some exemplary embodiments, the method further comprises a model determination module for determining a set of SVD recommendation models by:
determining object data of each object;
classifying each object according to each object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
the SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of the nth round by each object in the category and the recommendation information matrix of the nth round; the object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all health management terminals can determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
Receiving a second change gradient of the (N+1) th recommended information matrix relative to the (N) th recommended information matrix sent by the health management terminal; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (N+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
In some exemplary embodiments, recommendation module 54 is specifically configured to:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
the selected health information is recommended to the user corresponding to the user identification.
In some exemplary embodiments, the data acquisition module is specifically configured to:
If the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of the second preset object.
In some exemplary embodiments, the first gradient matrix is formed by combining first gradient vectors of respective objects in the category, and the model determining module is specifically configured to determine the respective first gradient vectors by the following formula:
wherein f (j, m) is a first variation vector, m represents mth recommendation information, j represents jth object in category, X j Representing the object factor matrix in the N-1 round of updating process, Y (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; wherein N is an integer greater than 1.
In some exemplary embodiments, the object factor matrix for the nth round is determined based on the model determination module by:
Wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1.
Since the device is the device in the method according to the embodiment of the present invention, and the principle of the device for solving the problem is similar to that of the method, the implementation of the device may refer to the implementation of the method, and the repetition is omitted.
As shown in fig. 6, based on the same inventive concept, an embodiment of the present invention provides a health information recommendation apparatus including a processor 61 and a data transmission unit 62, wherein:
the data transmission unit 62 is configured to perform:
transmitting target data of the target object to the processor;
the processor 61 is configured to perform:
acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
determining the category to which the target object belongs according to the target data;
determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model;
In the process of determining each SVD recommendation model in the SVD recommendation model set, the model constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting the preset stop condition is used as the SVD recommendation model through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between the recommendation information and the object; the recommendation information matrix characterizes a second association relationship between recommendation information and the object; the recommended information is health information;
and recommending the health information to each user in the target object.
In some exemplary embodiments, the processor 61 is configured to perform determining the set of SVD recommendation models by:
determining object data of each object;
classifying each object according to each object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
The SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of the nth round by each object in the category and the recommendation information matrix of the nth round; the object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all health management terminals can determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
receiving a second change gradient of the (N+1) th recommended information matrix relative to the (N) th recommended information matrix sent by the health management terminal; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (N+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
Stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
In some exemplary embodiments, the processor 61 is configured to perform:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
the selected health information is recommended to the user corresponding to the user identification.
In some exemplary embodiments, the processor 61 is configured to perform:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of the second preset object.
In some exemplary embodiments, the first variation gradient matrix is a combination of first variation gradient vectors for each object in the class, and the processor 61 is configured to determine each first variation vector by:
Wherein f (j, m) is a first variation vector, m represents mth recommendation information, j represents jth object in category, X j Representing the object factor matrix in the N-1 round of updating process, Y (m) Recommendation information matrix representing the N-1 th round, c jm And p jm The interested parameter of the recommendation information of the jth object in the nth round; wherein N is an integer greater than 1.
In some exemplary embodiments, the processor 61 is configured to determine the object factor matrix for the nth round by the following formula:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; wherein N is an integer greater than 1. />
The embodiment of the invention also provides a computer storage medium, wherein the computer storage medium stores computer program instructions which, when run on a computer, cause the computer to execute the steps of the health information recommendation method.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the spirit or scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims and the equivalents thereof, the present application is intended to cover such modifications and variations.
Claims (8)
1. A health information recommendation method, comprising:
acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
determining the category to which the target object belongs according to the target data;
determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model;
In the process of determining each SVD recommendation model in the SVD recommendation model set, using models constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting preset stop conditions as SVD recommendation models through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between recommendation information and an object; the recommendation information matrix characterizes a second association relationship between recommendation information and an object; the recommended information is health information;
recommending the health information to each user in the target object;
the set of SVD recommendation models is determined by:
determining object data of each object;
classifying each of the objects according to each of the object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
The SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of each object in the category to the nth round and the recommendation information matrix of the nth round; the initial value of the object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all the health management terminals determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
receiving a second change gradient of the (N+1) th recommended information matrix sent by the health management terminal relative to the (N) th recommended information matrix; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (n+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
Stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions; the preset conditions are model parameter convergence conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
2. The method of claim 1, wherein the recommending the health information to each user in the target object comprises:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
and recommending the selected health information to the user corresponding to the user identifier.
3. The method of claim 1, wherein the acquiring the target data of the target object comprises:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of a second preset object.
4. The method of claim 1, wherein the first gradient matrix is a combination of first gradient vectors for each object in the class, each first gradient vector being determined by the formula:
wherein f (j, m) is a first variation vector, m represents the mth recommendation information, j represents the jth object in the category, X (N-1)j Representing the object factor matrix in the N-1 round of updating process, Y N-1 (m) Recommendation information matrix representing the N-1 th round, c jm And p jm A parameter of interest representing recommendation information of the jth object at the nth round; wherein N is an integer greater than 1.
5. The method of any one of claims 1 to 4, wherein the object factor matrix for the nth round is determined by the following formula:
wherein X is Nj The object factor matrix is the N-th round; n-th round user content correlation matrix +.>Alpha is a constant in the training process; y is Y N (m) Recommending an information matrix for the nth round; where N is an integer greater than 1 and λ is a constant during training.
6. A health information recommendation device, comprising a processor and a data transmission unit:
the data transmission unit is configured to perform:
transmitting target data of a target object to the processor;
The processor is configured to perform:
acquiring target data of a target object; wherein one target object corresponds to at least one user; the target data comprises user characteristic data of each user;
determining the category to which the target object belongs according to the target data;
determining a target SVD recommendation model corresponding to the category in the SVD recommendation model set, and determining health information matched with the target object according to the target SVD recommendation model;
in the process of determining each SVD recommendation model in the SVD recommendation model set, using models constructed by each object factor matrix and each recommendation information matrix determined in the last round of updating meeting preset stop conditions as SVD recommendation models through multiple rounds of updating; in addition, in the updating process of each round, the first updating information sent to the health management terminal and the second updating information received from the health management terminal are both the change gradients between the recommendation information matrixes of two adjacent rounds; the object factor matrix characterizes a first association relationship between recommendation information and an object; the recommendation information matrix characterizes a second association relationship between recommendation information and an object; the recommended information is health information;
Recommending the health information to each user in the target object;
the processor is configured to perform determining the set of SVD recommendation models by:
determining object data of each object;
classifying each of the objects according to each of the object data; wherein the object data includes user characteristic data of a history user included in the corresponding object;
determining SVD recommendation models corresponding to each category;
the SVD recommendation model corresponding to each category is determined as follows:
determining a first change gradient of the recommendation information matrix of the nth round relative to the recommendation information matrix of the nth round according to the object factor matrix in the N-1 round updating process, the interested parameters of the recommendation information of each object in the category to the nth round and the recommendation information matrix of the nth round; the initial value of the object factor matrix is determined according to the object data of the corresponding object and the health data of each user corresponding to the object;
the first change gradient is sent to health management terminals corresponding to all objects in the category, so that all the health management terminals determine the recommendation information matrix of the N-1 th round according to the recommendation information matrix of the N-1 th round and the first change gradient;
Receiving a second change gradient of the (N+1) th recommended information matrix sent by the health management terminal relative to the (N) th recommended information matrix; the second change gradient is determined according to the N rounds of object factor matrix, the interested parameter of the recommended information of each object in the category to the (n+1) th round and the recommended information matrix of the (N) th round; the object factor matrix of the nth round is determined according to the recommendation information matrix of the nth round and the user content correlation matrix of the nth round; the user content correlation matrix of the nth round is determined according to the object factor matrix of the (N-1) th round and the recommendation information matrix of the (N) th round;
stopping updating until each obtained object factor matrix and each obtained recommendation information matrix meet preset conditions;
and determining the SVD recommendation model corresponding to the category according to each object factor matrix and each recommendation information matrix of the last round.
7. The recommendation device of claim 6, wherein the processor is configured to perform:
identifying a user identification in the health information;
screening a preset number of health information according to the weight of each health information;
and recommending the selected health information to the user corresponding to the user identifier.
8. The recommendation device of claim 6 or 7, wherein the processor is configured to perform:
if the target object has a new user and the user characteristic data of the new user is not acquired, determining the user characteristic data of the new user according to the user characteristic data of each user in the first preset object, and determining the target data of the target object;
if the target object is a new object and the user characteristic data of any user in the new object is not acquired, determining the target data of the target object according to the target data of a second preset object.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111612594.8A CN114417138B (en) | 2021-12-27 | 2021-12-27 | Health information recommendation method and equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111612594.8A CN114417138B (en) | 2021-12-27 | 2021-12-27 | Health information recommendation method and equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114417138A CN114417138A (en) | 2022-04-29 |
CN114417138B true CN114417138B (en) | 2024-04-02 |
Family
ID=81269373
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111612594.8A Active CN114417138B (en) | 2021-12-27 | 2021-12-27 | Health information recommendation method and equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114417138B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106326367A (en) * | 2016-08-11 | 2017-01-11 | 华南师范大学 | Mixed collaborative recommendation algorithm based on WNBI and RSVD |
CN108510402A (en) * | 2018-06-06 | 2018-09-07 | 中国平安人寿保险股份有限公司 | Insurance kind information recommendation method, device, computer equipment and storage medium |
CN109657189A (en) * | 2018-12-20 | 2019-04-19 | 南阳理工学院 | A kind of Mathematical Modeling Methods using computer hyperspace |
CN110189192A (en) * | 2019-05-10 | 2019-08-30 | 深圳前海微众银行股份有限公司 | A kind of generation method and device of information recommendation model |
WO2020133398A1 (en) * | 2018-12-29 | 2020-07-02 | 深圳市欢太科技有限公司 | Application recommendation method and apparatus, server and computer-readable storage medium |
CN112380439A (en) * | 2020-11-19 | 2021-02-19 | 平安科技(深圳)有限公司 | Target object recommendation method and device, electronic equipment and computer-readable storage medium |
CN113642707A (en) * | 2021-08-12 | 2021-11-12 | 深圳平安智汇企业信息管理有限公司 | Model training method, device, equipment and storage medium based on federal learning |
-
2021
- 2021-12-27 CN CN202111612594.8A patent/CN114417138B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106326367A (en) * | 2016-08-11 | 2017-01-11 | 华南师范大学 | Mixed collaborative recommendation algorithm based on WNBI and RSVD |
CN108510402A (en) * | 2018-06-06 | 2018-09-07 | 中国平安人寿保险股份有限公司 | Insurance kind information recommendation method, device, computer equipment and storage medium |
CN109657189A (en) * | 2018-12-20 | 2019-04-19 | 南阳理工学院 | A kind of Mathematical Modeling Methods using computer hyperspace |
WO2020133398A1 (en) * | 2018-12-29 | 2020-07-02 | 深圳市欢太科技有限公司 | Application recommendation method and apparatus, server and computer-readable storage medium |
CN110189192A (en) * | 2019-05-10 | 2019-08-30 | 深圳前海微众银行股份有限公司 | A kind of generation method and device of information recommendation model |
CN112380439A (en) * | 2020-11-19 | 2021-02-19 | 平安科技(深圳)有限公司 | Target object recommendation method and device, electronic equipment and computer-readable storage medium |
CN113642707A (en) * | 2021-08-12 | 2021-11-12 | 深圳平安智汇企业信息管理有限公司 | Model training method, device, equipment and storage medium based on federal learning |
Non-Patent Citations (2)
Title |
---|
Modeling and Correlation Analysis between Complex Networks;Liao Xiyang;《 PROCEEDINGS OF 2016 9TH INTERNATIONAL SYMPOSIUM ON COMPUTATIONAL INTELLIGENCE AND DESIGN (ISCID)》;20170616;全文 * |
融合社交网络信息的协同过滤方法;贺超波;《暨南大学学报(自然科学与医学版)》;20130615;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN114417138A (en) | 2022-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8706731B2 (en) | System and method for providing healthcare program service based on vital signals and condition information | |
Esteban et al. | Predicting the co-evolution of event and knowledge graphs | |
US20110295711A1 (en) | Apparel Fit Advisory Service | |
CN108776684A (en) | Optimization method, device, medium, equipment and the system of side right weight in knowledge mapping | |
CN111966904A (en) | Information recommendation method based on multi-user portrait model and related device | |
WO2015112202A1 (en) | Dynamic document matching and merging | |
Tekin et al. | RELEAF: An algorithm for learning and exploiting relevance | |
CN110580278A (en) | personalized search method, system, equipment and storage medium according to user portrait | |
CN112259245A (en) | Method, device and equipment for determining items to be checked and computer readable storage medium | |
Pan et al. | Collaborative recommendation with multiclass preference context | |
CN111523971A (en) | Purchasing operation sharing method and device | |
JP6753833B2 (en) | Grant device, grant method, grant program, and program | |
Hompes et al. | Detecting changes in process behavior using comparative case clustering | |
Carvalho et al. | Extremal dichotomy for uniformly hyperbolic systems | |
CN112289423A (en) | Method and system for double-diagnosis and rehabilitation based on intelligent community patients | |
Ahn et al. | The use of ordered weighted averaging method for decision making under uncertainty | |
Wu | A SD-IITFOWA operator and TOPSIS based approach for MAGDM problems with intuitionistic trapezoidal fuzzy numbers | |
Abd El-Raheem et al. | Log-rank tests for censored clustered data under generalized randomized block design: Saddlepoint approximation | |
CN114417138B (en) | Health information recommendation method and equipment | |
JP2018206232A (en) | Generation device, generation method, and generation program | |
CN113569151A (en) | Data recommendation method, device, equipment and medium based on artificial intelligence | |
CN116340643B (en) | Object recommendation adjustment method and device, storage medium and electronic equipment | |
CN113657970A (en) | Artificial intelligence based medicine recommendation method, device, equipment and storage medium | |
CN113436738A (en) | Method, device, equipment and storage medium for managing risk users | |
Ye et al. | Personalized on-device e-health analytics with decentralized block coordinate descent |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |