CN106682385B - Health information interaction system - Google Patents
Health information interaction system Download PDFInfo
- Publication number
- CN106682385B CN106682385B CN201610881071.6A CN201610881071A CN106682385B CN 106682385 B CN106682385 B CN 106682385B CN 201610881071 A CN201610881071 A CN 201610881071A CN 106682385 B CN106682385 B CN 106682385B
- Authority
- CN
- China
- Prior art keywords
- video
- data
- processor
- user
- rehabilitation
- 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
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/20—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for electronic clinical trials or questionnaires
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Abstract
The application discloses a health information interaction platform, which comprises a cloud server and a plurality of clients connected with the cloud server through a network; the client comprises a first processor, and an input assembly, a display screen and a first memory which are respectively connected with the first processor; the cloud server comprises a second processor, a second memory, a database and a verification assembly, wherein the second memory is connected with the second processor and stores a plurality of exercise rehabilitation videos, the database is used for updating a user questionnaire and answers of the user questionnaire in real time, and the verification assembly is used for verifying identity information input by a user and answers of the questionnaire; the first processor is internally provided with a judging module for judging input information of the input assembly; an AHP model is arranged in the judging module; and the judging module extracts evaluation factors of questionnaire answers input by the user aiming at the questionnaire.
Description
Technical Field
The invention relates to the field of health information interaction, in particular to a health information interaction platform.
Background
In recent years, the population aging of China is accelerated, the pace of life of people is accelerated along with the rapid development of the times, the pressure of life is increased, the incidence of various chronic diseases is increased, and the serious negative effects are brought to the physical and mental health of people and the stability and development of the society.
In order to conveniently and quickly acquire the health information condition of the user and condition the health information condition in daily life, health information interaction platforms for health management are newly developed. Because the health conditions of different people need to be intelligently judged and new diseases which continuously appear need to be identified and judged, machine learning systems used for enabling the information interaction platforms to be more intelligent are arranged in the information interaction platforms. As is well known, machine learning systems require the use of a certain amount of data as a basis to continually optimize learning. The existing health information interaction platform is excessively dependent on a machine learning system, so that the existing health information interaction platform can accurately run only after certain data to be recovered is input. In order to enable the existing health information interaction platform to have a relatively accurate feedback effect, a certain amount of data to be recovered must be input into the platform for model training before the platform is put into use. The whole debugging time of the health information interaction platform is wasted, and the operation is troublesome. What is more disadvantageous is that the accuracy of such a platform depends greatly on the overall quality of the data to be rehabilitated, and if the data to be rehabilitated used for training in the early stage is not appropriate, the accuracy of the health information interaction platform can still be low when the platform is used specifically, in other words, the use of the platform is greatly limited by the sample of the data to be rehabilitated used for training.
Therefore, there is an urgent need to develop a health information interaction platform which can be used immediately and does not depend on the data sample to be rehabilitated.
Disclosure of Invention
The invention aims to provide a health information interaction platform which can be directly used and does not depend on a data sample to be recovered.
The health information interaction platform comprises a cloud server and a plurality of clients connected with the cloud server through a network; the client comprises a first processor, and an input assembly, a display screen and a first memory which are respectively connected with the first processor; the cloud server comprises a second processor, a second memory, a database and a verification assembly, wherein the second memory is connected with the second processor and stores a plurality of rehabilitation video combinations, the database is used for updating a user questionnaire and answers of the user questionnaire in real time, and the verification assembly is used for verifying identity information input by a user and answers of the questionnaire; a grade reference table is arranged in the first memory; the relation between the grade reference table convention and the video number; the video numbers correspond to the rehabilitation video combinations stored in the database one to one.
Name interpretation:
rehabilitation video: the video file is composed of various combinations of video, audio and pictures and mainly comprises video for rehabilitation guidance. And combining the rehabilitation videos, namely a plurality of rehabilitation videos.
User questionnaires and answers: a questionnaire for data acquisition for the user, and answers to the questionnaire to be filled out by the user. The answers of the questionnaire filled by the user are the data to be rehabilitated, which can be understood as all input data of the user before the pushed video is obtained.
Rating reference table: various factors are extracted according to the data to be rehabilitated input by the user, the factors are graded according to different degrees, and a one-to-one corresponding relation table of each grade of factors and the video numbers is formed.
Video numbering: each rehabilitation video is numbered for unique identification.
The working principle and the beneficial effects are as follows:
the user answers the questionnaire transmitted into the client through the client, and answers of the questionnaire answered by the user are input into the first processor as data to be recovered through the input component. The first processor directly corresponds the data to be recovered with each factor of the grade reference table according to the grade reference table arranged in the first memory, and respectively obtains the video number corresponding to each factor. The client side sends the video numbers acquired from the grade reference table to the cloud server, the second processor sends the rehabilitation compound video combinations which correspond to the video numbers in the second storage one by one to the first processor according to the video numbers, and the first processor transmits the rehabilitation video combinations to the display screen to be displayed.
The health information interaction platform in the scheme can directly generate the result of the rehabilitation video combination only by the grade reference table in the first memory when a certain amount of data to be rehabilitated is not collected, namely when the accumulated data to be rehabilitated in the database is zero, and can be used immediately without depending on the health information interaction platform of the data to be rehabilitated.
Furthermore, a judging module used for judging input information of the input assembly is arranged in the first processor; an AHP model is arranged in the judging module; the AHP model adopts a two-stage method to solve the relative weight of each evaluation factor of the questionnaire; the first processor calculates a first association constant of the user according to the weight value of each evaluation factor, and the first processor transmits description data corresponding to the first association constant, which is set in the first processor, to the display screen.
A judgment module in a first processor carries out weight calculation on input data to be recovered by using an AHP model to obtain a first association constant, and the first processor transmits description data such as health or unhealthy data to a display screen to display after comparing the description data which is preset in the first processor and corresponds to the first association constant. The user selects to quit or carry out the next operation through the description data displayed by the display screen, and whether the user is healthy or not is conveniently and directly judged by inputting the data to be recovered.
Furthermore, a verification module used for connecting the video number with the specific video is arranged in the verification component; a second correlation parameter which is set by the expert for the rehabilitation video combination correspondingly according to each option in each factor in the questionnaire is arranged in the verification module; and (3) taking the total number of the rehabilitation video combinations as a random variable number, and constraining the priority relationship or the incidence relationship among the rehabilitation video combinations by taking the standard combination number of the rehabilitation video combination scheme given by an expert as a model constraint condition to establish a planning model.
When receiving a request transmitted by a rehabilitation video combination transmitted from a client, a verification component in the cloud server calculates questionnaire survey data obtained by a questionnaire through an AHP (advanced health care) model and a second associated parameter for each rehabilitation video combination in a setting verification module to obtain a plurality of rehabilitation video combinations suitable for pushing, namely rehabilitation training scheme combinations. And the second processor transmits the data of the rehabilitation video combination scheme combination obtained by the verification module to the second memory, extracts the corresponding rehabilitation video combination from the second memory and transmits the rehabilitation video combination to the client.
Further, an optimization module is also arranged in the first processor; doctor guidance suggestion data compiled according to various common phenomena are arranged in the optimization module; the optimization module calculates corresponding video numbers through doctor guidance opinion data, calculates absolute values of differences between video combinations and the maximum treatment degrees of experts by taking the adjustment quantities of the video numbers as variables, obtains multiple groups of comprehensive parameter adjustment quantities by adopting an improved intelligent algorithm to perform successive iteration, performs factor option distribution and entropy processing on the adjustment quantities according to weights, and performs adjustment optimization on video number tables corresponding to the rehabilitation video combinations.
After the client receives the rehabilitation video combination transmitted from the cloud server, the optimization module in the first processor automatically optimizes various parameter values related to the rehabilitation video combination by adopting an improved intelligent algorithm through doctor guidance suggestion data preset in the optimization module.
Furthermore, a screening module is arranged in the second processor; the screening module takes the data to be recovered updated in real time in the database as input data and category data of a decision tree model arranged in the screening module, classifies problems in questionnaires corresponding to the data to be recovered, and adopts a CART algorithm to construct the decision tree model; the screening module transmits a judgment result of whether the rehabilitation video combination needs to be pushed or not to the first processor according to the user data input from the input assembly.
And the screening module in the second processor performs regression tree classification on the questionnaire survey answers input from the input assembly according to the user input data updated in real time in the database, so that the screening module obtains a judgment result of the rehabilitation video combination or the non-rehabilitation video combination according to the input of the user symptom data.
Further, a learning module arranged in the verification component trains and learns all data transmitted to the database through the BP neural network to obtain corresponding BP neural network parameters, and the parameters are stored in the database; when the verification component receives the judgment result of the rehabilitation video combination, the learning module outputs a corresponding video number of the pushed rehabilitation video combination to the second processor according to the data to be rehabilitated input by the input component and the BP neural network parameter; meanwhile, the background program updates the previous BP neural network parameters with the data.
And the learning module in the verification component receives the result of the rehabilitation video combination transmitted by the screening module, and calculates the user data which is stored in the database and is obtained after real-time updating through the BP neural network to obtain BP neural network parameters. The BP neural network parameters refer to all parameters that can determine the topology of the BP neural network. When a user customizes the rehabilitation video combination scheme through the input component, the learning module adds the BP neural network parameters into the verification component in the rehabilitation video combination process, changes the video number of the pushed rehabilitation video combination, optimizes the pushed video combination, and simultaneously stores the model parameters of the BP neural network including the input to the database. And the second processor carries out video pushing on the client through the obtained video combination.
Furthermore, a correction module is also arranged in the second processor; the correction module is internally provided with a support vector machine model, judges whether to carry out video push again according to feedback data of a user, and meanwhile, the second processor updates the relation between the rehabilitation video combination and the video number in real time according to the feedback data of the user; and transmitting the corrected associated data to a database for storage.
After the user operates through the pushed rehabilitation video combination, the user inputs a feedback result aiming at the pushed rehabilitation video combination through the input component, wherein the user feeds back by filling in a questionnaire after the rehabilitation video combination operation. And a correction module in the second processor corrects the relevant parameters of the rehabilitation video combination through the feedback data motion support vector machine model, and updates the corrected relevant parameters to a database for storage.
Further, the operation process of the health information interaction platform comprises at least two stages, wherein the initial operation stage enters the middle operation stage after 1000-2000 samples of data to be recovered are collected.
The invention passes through two video pushing periods, the first video pushing is mainly carried out according to the grade reference table arranged in the first memory, the data to be recovered (such as user input data and user health data) are effectively accumulated through the process, and the invention provides the premise for machine learning later and continuously perfecting the whole information interaction platform.
According to the invention, various inputs of the user are collected in a questionnaire survey mode, and the input of the user is guided, so that the user can provide type data required by a health information interaction platform, and the effective utilization rate of the input information of the user is improved.
After the first-time video pushing is completed, various parameter values related to the rehabilitation video combination can be automatically optimized through an improved intelligent algorithm through an optimization module arranged in a first processor, so that the health information interaction platform has preliminary learning capacity, and the accuracy of the rehabilitation video combination is effectively improved.
And when the video is pushed for the second time, the database collects certain data to be recovered. Parameters related to rehabilitation video combination in the health information interaction platform are adjusted and perfected through a regression tree model, a BP neural network and a support vector machine model sequentially on the basis of collected user data, and the whole platform has a machine learning function in gradual operation. The video pushing is more humanized and intelligent according to the accumulation of user data, and the accuracy of the combination of the recovered videos is improved.
Drawings
Fig. 1 is a schematic structural diagram of an embodiment of the present invention.
FIG. 2 is a schematic diagram of the initial operation of an embodiment of the present invention.
Fig. 3 is a schematic diagram of the middle stage operation of the embodiment of the present invention.
Detailed Description
The present invention will be described in further detail below by way of specific embodiments:
reference numerals in the drawings of the specification include: the device comprises a verification component 1, a second memory 2, a database 3, an external port 4, a wireless communication component 5, a second processor 6, an input component 7, a video analysis component 8, a display screen 9, a first memory 10, a network transceiving component 11 and a first processor 12.
As shown in fig. 1, the health information interaction platform includes a cloud server and a plurality of clients connected to the cloud server via a network; the client comprises a first processor 12, an input component 7, a video analysis component 8, a display screen 9, a first memory 10 and a network transceiving component 11, wherein the input component 7, the video analysis component 8, the display screen 9, the first memory 10 and the network transceiving component 11 are connected with the first processor 12 through a data bus, and the first processor 12 controls the operation of each component of the client; the network transceiving component 11 establishes a data communication link between the client and the cloud server through at least one network protocol; the input component 7 is used for inputting information such as identity authentication information and questionnaire answers by a user; the first memory 10 stores the sports rehabilitation video data and the survey to-be-rehabilitated data from the cloud server received via the network transceiving component 11; the video analysis component 8 is used for analyzing the sports rehabilitation video data from the cloud server received by the network transceiving component 11; the display screen 9 is used for playing the analyzed exercise rehabilitation video data.
The cloud server comprises a second memory 2 for storing a plurality of exercise rehabilitation videos, a database for updating answers of questionnaires of the user in real time, a verification component 1 for verifying identity information input by the user and the answers of the questionnaires, a wireless communication component 5, an external port 4 and a second processor 6; the external port 4, the second memory 2, the database, the verification component 1 and the wireless communication component 5 are all connected with the second processor 6; the external port 4 is adapted to couple the second memory 2 of the cloud server and the database to other devices directly or indirectly through a network; the authentication component 1 is used for authenticating the user identity information received via the wireless communication component 5 and sending an authentication result signal to the second processor 6; the second processor 6 controls the questionnaire in the database to be transmitted to the aforementioned client via the wireless communication component 5 in response to the authentication pass signal transmitted by the authentication component 1. The user inputs answers to questions in the questionnaire into the first processor 12 through the input component 7, and the first processor 12 sends the user data to the cloud server through the network transceiving component 11. The wireless communication component 5 of the cloud server receives the user data and passes these to the second processor 6. The second processor 6 transfers the user data to the user database 3 and transfers the data to the verification component 1, the verification component 1 sends a signal to the second processor 6 after passing the verification answer, so that the second processor 6 retrieves the corresponding video data from the second memory 2 and transfers the video data to the client through the wireless communication component 5.
The second processor is provided with a later learning module for later learning, and the number of nodes of the BP neural network model arranged in the later learning module is more than that of the nodes in the BP neural network model in the learning module.
Through calculation of a multi-node BP neural network in the later learning module, a video parameter table related to rehabilitation video combination can be optimized, and video push of the health information interaction platform is more intelligent.
The invention passes through two video pushing periods, the first video pushing is mainly carried out according to the grade reference table arranged in the first memory, the data to be recovered (such as user input data and user health data) are effectively accumulated through the process, and the invention provides the premise for machine learning later and continuously perfecting the whole information interaction platform.
According to the invention, various inputs of the user are collected in a questionnaire survey mode, and the input of the user is guided, so that the user can provide type data required by a health information interaction platform, and the effective utilization rate of the input information of the user is improved.
After the first-time video pushing is completed, various parameter values related to the rehabilitation video combination can be automatically optimized through an improved intelligent algorithm through an optimization module arranged in a first processor, so that the health information interaction platform has preliminary learning capacity, and the accuracy of the rehabilitation video combination is effectively improved.
And when the video is pushed for the second time, the database collects certain data to be recovered. Parameters related to rehabilitation video combination in the health information interaction platform are adjusted and perfected through a regression tree model, a BP neural network and a support vector machine model sequentially on the basis of collected user data, and the whole platform has a machine learning function in gradual operation. The video pushing is more humanized and intelligent according to the accumulation of user data, and the accuracy of the combination of the recovered videos is improved.
As shown in fig. 2, at the initial stage of the operation of the health information interaction platform, the health information interaction platform does not have any user data, and the user symptom data and the corresponding rehabilitation training data are continuously collected by the user in the process of using the health information interaction platform.
First, the cloud server transmits a questionnaire to the client, and the user reads the questionnaire through the display screen 9 and inputs answers to the questionnaire, that is, user symptom data into the first processor 12 through the input component 7.
The first processor 12 performs a preliminary judgment on the symptom description through the judgment module, and then presents the judgment result to the user through the display screen 9, so that the user can perform subsequent operation selection.
AHP weighting is performed in advance for conventional user phenomenological input using the AHP model in the first processor 12. The relative weight of each factor is solved by adopting a two-stage method: and (3) comparing every two reference factors by adopting a (0, 1, 2) three-scale method to establish a comparison matrix shown in the table 1.
TABLE 1
Calculating the importance ranking index of each element; the second stage employs the following transformations: b is
iWith B (factor 1, factor 2, factor 3, factor 4, factor 5, factor 6, factor 7, factor 8, factor 9)
jThe comparison matrix composed of (factor 1, factor 2, factor 3, factor 4, factor 5, factor 6, factor 7, factor 8, and factor 9) is converted into a determination matrix, and a final weight vector is calculated by a geometric averaging method. Among the factors to be considered may be factors related to symptom examination such as the cause of pain, the length of pain, and the level of pain, which are buried in the question of the questionnaire. The user enters data containing these factors in the form of data to be rehabilitated by answering a questionnaire.
Based on known elements B
iAnd B
jOf importanceRanking index r
iAnd r
jTo construct a decision matrix. C ═ C
ij. Judging matrix coefficient c
ijRepresents element B
iAnd B
jIf c is expressed, the degree of importance between
ijIf > 1, then B is indicated
iRatio B
jImportantly, B
i≥B
j。c
ijThe greater the degree of importance. If c is
ijIf < 1, then it means B
iRatio B
jImportantly, B
i≤B
j。c
ijWhen 1, B
i＝B
j. In order to improve the reliability and consistency of the judgment matrix, the relative importance degree of all element pairs is given based on the same transformation criterion, and the transformation criterion is set as f (r)
i,r
j)＝c
ij。
Through the determination matrix, the first processor 12 calculates the input diagnosis data and obtains the weight, and transmits the determination result set in the first memory 10 to the display 9 for the weight, so that the user can know the health condition at that time.
Three health phenomenon grade reference tables are established for main body parts after the medical expert analyzes each question factor option by using the SPA model in the first memory 10 in advance. The specific factors of the recommended scheme are nine in total, and are respectively corresponding to various considered factors.
TABLE 2-1
| Pain part (waist) | Severe phenomenon | General phenomena | Health care | Description of the invention |
| Consideration 1 | 1,3 | 2 | 4 | Number represents option code number |
| Consideration 2 | 4,5,6 | 2,3 | 1 | Number represents option code number |
| Consideration 3 | 3 | 2 | 1 | Number represents option code number |
| Consideration 4 | 3,4 | 2 | 1 | Number represents option code number |
| Consideration factor 5 | 4 | 2,3 | 1 | Number represents option code number |
| Consideration of factor 6 | 1 | 2 | 2 | Number represents option code number |
| Consideration 7 | 1 | 2 | 2 | Number represents option code number |
| Consideration of factor 8 | 1 | 2 | 2 | Number represents option code number |
| Consideration 9 | 4,5,6 | 2,3,4 | 1 | Number of digital representative options |
Tables 2 to 2
| Pain part (neck) | Severe phenomenon | General phenomena | Health care | Description of the invention |
| Consideration 1 | 1,3 | 2 | 4 | Number represents option code number |
| Consideration 2 | 4,5,6 | 2,3 | 1 | Number represents option code number |
| Consideration 3 | 3 | 2 | 1 | Number represents option code number |
| Consideration 4 | 3,4 | 2 | 1 | Number represents option code number |
| Consideration factor 5 | 4 | 2,3 | 1 | Number represents option code number |
| Consideration of factor 6 | 1 | 2 | 2 | Number represents option code number |
| Consideration 7 | 1 | 2 | 2 | Number represents option code number |
| Consideration of factor 8 | 1 | 2 | 2 | Number represents option code number |
| Consideration 9 | 4,5 | 2,3 | 1 | Number of digital representative options |
Tables 2 to 3
| Pain part (shoulder) | Severe phenomenon | General phenomena | Health care | Description of the invention |
| Consideration 1 | 1,3 | 2 | 4 | Number represents option code number |
| Consideration 2 | 4,5,6 | 2,3 | 1 | Number represents option code number |
| Consideration 3 | 3 | 2 | 1 | Number represents option code number |
| Consideration 4 | 3,4 | 2 | 1 | Number represents option code number |
| Consideration factor 5 | 4 | 2,3 | 1 | Number represents option code number |
| Consideration of factor 6 | 1 | 2 | 2 | Number represents option code number |
| Consideration 7 | 1 | 2 | 2 | Number represents option code number |
| Consideration of factor 8 | 1 | 2 | 2 | Number represents option code number |
| Consideration 9 | 4,5,6 | 2,3,4 | 1 | Number of digital representative options |
Tables 2 to 4
| Number of | 1 | 2 | 3 | 4 | 5 | 6 |
| Options for | A | B | C | D | E | F |
In tables 2-1 to 2-3, the number represents the option code number, which means that the selected number is the code number of the option, i.e. the number of the rehabilitation video combination. The number represents the number of the options, and means that the number of the options is limited by the number, for example, when AB/AC/BD is selected, the number is 2, and ABC/BCD is 3. The scaling relationships for each answer choice in the evaluation table are shown in tables 2-4.
The health condition of the user is divided into three levels according to the health symptom grade reference table, so the contact element in the SPA is ternary. And performing weighted calculation on the identity, the difference and the inverse in the evaluation system according to the symptom data and the health symptom grade reference table submitted by the user to finally obtain a comprehensive joint coefficient, and determining the corresponding health symptom according to the magnitude relation of the joint coefficient. The homologies correspond to the symptom severity level, the heterocoefficiences correspond to the symptom general level, the inverse coefficients correspond to the symptom well type (health type), and the maximum coefficient in the calculated coefficients is the health level of the user. The first processor 12 retrieves the health rating data from the first memory 10 and passes it on to the display screen 9, enabling the user to visually see his health status.
And then, the client transmits the symptom data input by the user to the cloud server. The second processor 6 receives the user symptom data communicated from the wireless communication assembly 5 and communicates the data to the verification assembly 1 for verification and storage in the user database 3 simultaneously.
The verification component 1 is provided with a planning model. The relevance between the video and the factor options influencing the scheme recommendation is established by setting corresponding treatment parameters for the video by a medical expert according to possible options in model factors, namely directly determining the treatment effect of each video on a specific possible symptom (such as 0-3 points of pain degree). The method comprises the steps of establishing a planning model by taking videos as random variables (the total number of the videos is the number of the random variables, the random variables can only take 0 or 1), taking the priority relation or the incidence relation among the videos and the standard combination number of video schemes given by experts as model constraint conditions, obtaining the random variables with the final value of 1 by maximizing the treatment degree, and taking the videos corresponding to the variables as the final rehabilitation training scheme combination.
The verification component 1 obtains the rehabilitation training scheme combination data through the planning model and transmits the data to the second processor 6, and the second processor 6 calls the rehabilitation video combination pre-stored in the second memory 2 according to the data and sends the data of the rehabilitation video combination to the client. The client can see the specific rehabilitation video through the display screen 9 by accessing the rehabilitation video combination of the cloud server.
The specific algorithm of the planning model may be as follows:
in the embodiment, the description information of the current physical condition of the user based on the questionnaire response of the user is received through the client;
the data to be recovered is extracted symptom description information, namely questions and options selected by a user;
the pushing scheme is a rehabilitation training video;
a specific questionnaire form comprising a plurality of questions, each comprising a plurality of options for the question, each option being specific for a symptom description, I being the number of questions, J
iThe total number of options for the ith question;
for each option, a weight parameter is used for representing the importance degree of the option to the problem to which the option belongs;
for each question, another weight parameter is used to characterize the importance degree of the question to the selection of the combination with the rehabilitation video;
the comprehensive parameters of the rehabilitation training video comprise a plurality of efficacy parameters with the values of 0-100%, priority parameters with the values of 0,1, 2 and 3 … … M (M represents the maximum priority number), and phase parameters representing the phase of the video, wherein the phase parameters with the values of 1,2 and 3 … … K (K represents the maximum priority number);
the limited parameter of the rehabilitation video combination is preferably the number of rehabilitation training videos contained in a single rehabilitation video combination;
these parameters can be searched through a parameter correspondence table preset in the storage device.
Thus, the parameter settings in the present embodiment are specifically shown in table 3:
TABLE 3
The selection problem of the push result is then converted into an O-1 integer programming model; in particular, i.e. maximum calculation under certain conditions
max g(x)＝Kernel(Kernel(R,Handamard(λ,U)),x)
Wherein a Hadamard symbol represents a Hadamard product of the matrix/vector; kernel denotes the vector number product
The maximization calculation needs to satisfy the following constraints:
the number of videos is N;
in the rehabilitation stage where the user is positioned, the video which does not conform to the training of the stage cannot be output
Only videos belonging to the same stage are included, and the selection is carried out according to the rehabilitation stage corresponding to the user and the stage parameters of each video;
in case the treatment is satisfied as well, selecting the video with high priority (determined by the priority parameter of the video);
the specific maximization is calculated as solving:
the purpose of the maximization calculation is to solve X meeting the constraint conditions, select the corresponding rehabilitation video combination according to the result of the X, and push the rehabilitation video combination to the user, wherein the maximized numerical value is the comprehensive efficacy parameter of the video combination and is recorded as the comprehensive efficacy parameter below
g。
Finally, the model in the whole health information interaction platform is optimized through the improved intelligent algorithm model arranged in the first processor 12. The first processor 12 is provided with doctor's instruction data, which is data for several tens of symptoms commonly prescribed by medical experts, and gives an objective and accurate treatment plan combination for each symptom in a medical expert workshop. Calculating comprehensive parameters corresponding to the videos according to the dozens of types of symptom data, taking the adjustment quantity of the comprehensive parameters as a variable, taking the absolute value of the difference between the video combination calculated by the model and the maximum treatment degree of the expert as a target function, adopting an improved intelligent algorithm to carry out successive iteration to obtain a plurality of groups of comprehensive parameter adjustment quantities, carrying out factor option distribution and entropy processing on the adjustment quantities according to the weight, and finally carrying out adjustment optimization on the whole video parameter table corresponding to the videos.
The preset push scheme is preferably one or more of a rehabilitation training video, a rehabilitation guidance picture and text and a rehabilitation training comment audio; each individual push scheme is artificially assigned a synthesis parameter by a professional.
The composite parameters include, but are not limited to: the single push scheme is used for respectively corresponding to the efficacy parameters of the plurality of items of data to be recovered, the priority parameter of the single push scheme and the like;
for example: the effective therapeutic capacity of the push regimen A for the disease B is r
AB(r
AB∈[0,1]) The value can be used for measuring the strength of the effective treatment capacity of the push scheme.
The priority parameter is that the push schemes are subjected to priority ordering, the push schemes associated with the scheme to be pushed and the unrelated push schemes are firstly screened out, the unrelated push scheme priority sequence is set to be 0, the associated push schemes are set to be 1,2, 3 and … … according to the priority sequence level, the number of the associated push schemes is more than 1 and less than 2 and less than … and less than n (a plurality of videos are all in the same priority level), and the information of the same priority relation is subjected to number limitation output according to the comprehensive effective treatment capacity.
Meanwhile, the push scheme combination is also preset with a limiting parameter, and the limiting parameter is preferably the number of single push schemes contained in a single push scheme combination;
and calculating the optimized pushing scheme combination according to the data to be recovered, the comprehensive parameters of each pushing scheme and the limited parameters of the pushing scheme combination, and sending the optimized pushing scheme combination to the user.
And if the user participates in a new questionnaire diagnosis after the ith round of rehabilitation training, and when the diagnosis result shows that the user still needs to accept the next round of training, the system recommends the rehabilitation training and the corresponding training days carried out by the round according to the next round of diagnosis data, and pushes the (i + 1) th round of rehabilitation training according to the feedback condition in the training process until the user can completely reach the medical health standard. The large-amount and non-repeated targeted pushing scheme can realize real-time monitoring on the health condition of the user, and is beneficial to gradual rehabilitation of the body of the user.
And solving the optimal push scheme combination by adopting a mode of establishing a planning model in the first stage of the health information interaction platform.
And taking the push scheme as a random variable according to the information, setting a constraint condition according to the limited parameters of the push scheme combination to establish a planning model by taking the effectiveness of the maximum push scheme combination on the user as a target, and finally obtaining the result of the push scheme combination.
More preferably, the calculated pushing scheme combination can be compared with a pushing scheme combination formulated by a professional doctor according to the same user data; and updating the comprehensive parameters of each pushing scheme according to the comparison result.
In advance, medical experts compile dozens of common diseases, and call medical expert seminars to provide objective and accurate push scheme combinations for each symptom. After the pushing scheme combination for pushing is calculated each time, the updating quantity of the comprehensive parameters is used as a variable, the absolute value of the difference between the maximum treatment degrees of the pushing scheme combination calculated by the model and the combination given by the expert is used as a target function for the same symptom, successive iteration is carried out by adopting an intelligent algorithm to obtain a plurality of groups of comprehensive parameter updating quantities, the updating quantities are subjected to multi-factor distribution by an entropy weight method, and finally the whole video parameter table corresponding to the video is updated and optimized.
The traditional intelligent algorithm solves some complex engineering problems by simulating a certain natural process; for example, genetic algorithms simulate natural competitive selection, take all individuals in a group as objects, and guide to efficiently search an encoded parameter space by using a randomization technology; the basis of the simulated annealing algorithm is the similarity between the solid matter annealing process and the combinatorial optimization problem. Applying the thermodynamic theory to statistics, and imagining each point in the search space into molecules in the air; the energy of a molecule, that is, its own kinetic energy; each point in the search space, like air molecules, also carries "energy" to indicate how appropriate the point is for proposing a proposition. The algorithm starts with searching for an arbitrary point in space: each step first selects a "neighbor" and then calculates the probability of reaching the "neighbor" from the existing location.
Nowadays, more improved intelligent algorithms appear, and have higher computational efficiency and search accuracy; these algorithms, including but not limited to genetic annealing algorithms/modified particle algorithms/adaptively adjusted differential evolution algorithms, may be used in embodiments of the present invention.
In order to further optimize updating, doctor guidance suggestion is further adopted in the implementation to cooperate with the parameter table to calculate the updating amount.
The doctor suggests that the data is from the source that, in advance, the expert compiles dozens of common symptoms, each type of symptoms comprises a plurality of symptoms, which is equal to that a plurality of symptom options are selected in the questionnaire; if there are P symptoms, the symptom option of each symptom can be represented by an answer vector U
pTo represent;
then, an expert workshop is called to give objective and accurate pushing scheme combination for each disease.
Further parameter settings are shown in table 4:
TABLE 4
Thus, the update to the synthesis parameters translates into the following minimization problem:
the present implementation uses an improved intelligent algorithm to solve the minimization problem; the method can selectively adopt three algorithms of genetic annealing algorithm/improved particle algorithm/self-adaptive adjustment differential evolution algorithm.
In the genetic annealing algorithm, binary coding is specifically adopted, and after an initial generation population is generated, the best x is generated by generation evolution according to the principle of survival and excellence and disadvantage of a suitable person
pAn approximate solution of; and in order to accelerate the searching speed and enhance the local searching capability of the algorithm, the simulated annealing algorithm is used for searching by defining the energy evaluation function and the initial temperature of a single individual.
In the improved particle algorithm, a new group extremum and an individual extremum calculation mode are used.
For each individual extreme Pb
iRandomly selecting an individual extreme value from the rest other individual extreme values of the same generation as Pbj, and recording the newly generated individual extreme value as Pb
iThen:
Pb
i(t)＝r*Pb
i(t)+(1-r)*Pbj(t)
where t denotes run to the tth generation, i denotes the ith particle, j denotes the jth particle, and r is a randomly generated number between [0,1 ].
The improvement to the individual population extremum Gb is as follows:
all the individual extreme values are sorted, and the best K individual Pb is selected from the sorted individual extreme values
1′,Pb
2′,,,Pb
k'Gb' is represented by the weighted average of the K individuals, then
Wherein t represents the run to the tth generation, i represents the selected ith particle, a
iSatisfy the requirement of
In the self-adaptive adjustment differential evolution algorithm, a self-adaptive operator lambda is added in the traditional differential evolution algorithm, a mutation operator is changed in the iterative process, the population keeps the individual diversity at the initial stage in a step-by-step reduction mode, the precocity is avoided, the excellent information is kept at the later stage, the optimal solution is prevented from being damaged, and the probability of searching the global optimal solution is increased:
in the formula: f
0Is a mutation operator; g
mFor maximum evolutionary algebra: g is the current evolution algebra.
By one of the above methods, x is finally solved
pAnd obtaining a final fine tuning matrix Y according to the weight
pNamely:
Y
p＝Kernel(x
p,π
p)；
y is calculated once separately for all disorders
p；
And finally, updating the incidence matrix R of the rehabilitation video and the symptom description information by using the fine adjustment matrix, namely:
r' is the updated incidence matrix.
And the initial operation stage of the health information interaction platform is completed.
As shown in fig. 3, after a period of initial operation, after the user database 3 collects enough user data, the health information interaction platform starts to enter the middle-stage operation stage.
Firstly, the user inputs answers of questionnaires, namely symptom data, into the first processor 12 through the input component 7 through the client, and the first processor 12 transmits the symptom data to the cloud server through the network transceiving component 11. The wireless communication component 5 in the cloud server receives the symptom data and passes the symptom data to the second processor 6.
The second processor 6 uses a decision tree model, and takes user symptom data and user health condition data collected by the user database 3 in the initial operation as input data and category data of the decision tree model respectively, classifies questions in questionnaires, and processes the data of the questionnaires into decimal numbers, namely, all options in each question are regarded as binary values of 0 or 1, when the option is selected, 1 is selected, otherwise 0 is selected; then, the values are connected into a group of binary strings, and the binary strings are converted into decimal numbers by adopting a binary conversion method (the characteristic value aiming at the single selection problem is a discrete type {2 ]
1,2
2,…,2
n-1N is the number of options corresponding to the question }; for the multi-selection problem, the characteristic value is continuous {12
n-
1And-1, n is the number of options corresponding to the problem }), and a CART algorithm is adopted to construct a decision tree model.
Preprocessing question data in a questionnaire survey, regarding all options in each question as binary values of 0 or 1, and taking 1 when the option is selected, or taking 0 otherwise; then, connecting the values into a binary string, and converting the binary into a decimal number by adopting a binary conversion method; forming a feature set, wherein the feature set can be specifically expressed as: { Feature1, Feature2, Feature3, Feature4, Feature5, Feature6, Feature7, Feature8, Feature9}
Through conversion calculation, the specific value conditions of each feature set can be known as shown in the following table:
TABLE 5
According to the decision tree model setting, the class set is { Level1, Level2 and Level3}, and the corresponding values can be set to be {1, 2 and 3 }.
And classifying the information input by the user from the client through the decision tree model, and judging whether the scheme pushing is required to be carried out on the user or not according to the classification result. If the Level1 or Level2 is the case, rehabilitation therapy is proved to be needed, otherwise, no recommendation is needed (please refer to an evaluation table, specifically, the first Level and the second Level indicate obvious symptoms, and the third Level indicates that the body is healthy and no rehabilitation training is needed).
If the judgment result is that pushing is not needed, the cloud server sends information which does not meet the conditions to the client: you are very healthy and do not need a push regimen treatment. If the judgment result is that the proposal needs to be pushed, the output result of the decision tree model is directly sent to a BP neural network model arranged in the verification component 1.
The user accesses the cloud server through the client, and can see the health condition of the user judged by the decision tree model on the display screen 9. And further the following operation is performed according to the prompt on the display screen 9.
And the BP neural network in the verification component 1 trains and learns the data collected in the initial operation stage in the background to obtain corresponding BP neural network parameters and stores the parameters in the cloud. When a user customizes a scheme, the model stored in the cloud outputs a corresponding rehabilitation video combination according to the data to be rehabilitated submitted by the user; meanwhile, the background program updates the previous BP neural network parameters with the data. In the whole model training process, the network input is user symptom data (binary coding character strings), and the expected output is video output data obtained by correcting treatment video data obtained by performing 0-1 planning model calculation on a video parameter table by a doctor. And (3) modifying the video scheme obtained in the initial stage to obtain data (in a binary decision vector form), setting the network into a double-hidden-layer structure, and learning by adopting a gradient descent algorithm.
The specific calculation of the BP neural network model is as follows:
the method comprises the following steps: initializing network, composing input of neural network from user data transmitted by decision tree model and support vector machine model, processing the input data into binary numberAccordingly, the input and output of the neural network form a sequence, which is expressed as (x, y), and according to the sequence, the number of nodes of the input layer, the hidden layer and the output layer of the network can be determined, which is n nodes, l nodes and m nodes respectively. Then, initializing connection weight and threshold, and setting the connection weight between neuron in input layer and hidden layer as w
ijThe connection weight between the hidden layer and the output layer neuron is w
jkThe thresholds for the hidden and output layers are a and b, respectively, finally, the neuron excitation function f (x) and the learning rate η of the neural network are given.
Step two: computing the output of the hidden layer: according to the input variable X of the network, the connection weight w between the input layer and the hidden layer
ijAnd the threshold value a of the hidden layer, the output H of the hidden layer can be obtained:
in the above formula, f is the excitation function of the hidden layer; l is the number of nodes in the hidden layer. The excitation function is:
step three: calculating the output of the output layer: calculating the output of the hidden layer according to the step two, and connecting the connection weight w of the hidden layer and the output layer
jkAnd an output layer threshold b, from which a BP neural network prediction output 0:
step four: and calculating the error, namely calculating the prediction error e of the network according to the expected output Y of the network and the network prediction output O obtained in the step three:
e
k＝Y
k-O
kk＝1,2,…,m
step five: updating the weight value, namely updating the connection weight value w of the network according to the prediction error e of the neural network
ijAnd w
jkUpdating:
w
jk＝w
jk+ηH
je
kj＝1,2,…,l；k＝1,2,…,m
wherein η is the learning rate of the neural network.
Step six: updating the threshold, namely updating the thresholds a and b of the hidden layer and the output layer of the network according to the prediction error e of the network:
b
k＝b
k+e
kk＝1,2,…,m
step seven: and judging whether the iteration end conditions of the algorithm are met, if not, returning to the step two.
The BP calculation result obtained by the BP neural network model is the second push data, the second push data corresponds to the video scheme in the second memory 2, and the second processor 6 sends the video scheme to the client or sends the position data storing the video scheme to the client, and the client obtains the position data.
In order to improve the convergence rate of the BP neural network, a second-order gradient method is adopted after the step five:
w(t+1)＝w(t)-η[▽
2E(t)]
-1▽E(t)
wherein:
although the second-order gradient method has good convergence, the second derivative of E to w needs to be calculated, and the calculation amount is large. In general, the second-order gradient method is not directly adopted, but a variable-scale method or a conjugate gradient method is adopted, which has the advantage of fast convergence of the second-order gradient method without directly calculating the second-order gradient.
Wherein, the variable scale algorithm:
w(t+1)＝w(t)+μH(t)D(t)
Δw(t)＝w(t)-w(t-1)
ΔD(t)＝D(t)-D(t-1)
the support vector machine model provided in the second processor 6 adopts a common two-classification support vector machine model, and the feedback data set after the client uses the video is as follows: { pain sensation after treatment, degree of functional improvement } as a known training set.
The method comprises the following steps: let the known training set:
T＝{(x
1,y
1),…,(x
l,y
l)}∈(X×Y)
l
wherein x is
i∈X∈R
n，y
i∈Y＝{1，-1}(i＝1，2，...，l)；x
iIs a feature vector.
Step two: selecting a proper kernel function YK (x, x') and a proper parameter C, and constructing and solving an optimization problem:
obtaining an optimal solution:
selecting α
*A positive component of
And calculates therefrom a threshold value:
step four: constructing a decision function:
determining the category assignment according to whether f (x) is 1 or-1.
(x) when the value is 1, the functions of all parts of the user are recovered to be normal, the video pushed to the user by the model is accurate and useful through the neural network, and (x) when the value is-1, the user still needs to participate in the next stage of rehabilitation training, and at the moment, { pain feeling: mean, variance, degree of functional improvement: and inputting the mean value and the variance into a support vector machine model, and outputting video data to the client by the BP neural network model again, namely optimizing the calculation of the BP neural network model.
In the continuous feedback of the support vector machine model, the symptoms input by the user and the pushed video are continuously corrected according to the feedback data of the user, so that the following neural network model is more accurate when the rehabilitation video is combined for the user. Meanwhile, the correction data are tracked when the user carries out video rehabilitation training, so that the user can realize a quicker rehabilitation effect (the later-stage pushing scheme can be adjusted according to the symptom response of the user after the user receives the rehabilitation training, and humanized customization is achieved).
In the later operation of the health information interaction platform, a deep learning model is mainly established, the number of hidden nodes of the neural network in the middle-stage machine learning stage is increased, user feedback data is used as the basis of staged treatment, and a larger neural network is directly established to replace all the models in the past.
In the scheme, the health information interaction platform comprises an operation process of at least two stages, wherein the initial operation stage enters a middle operation stage after 1000-2000 samples of data to be recovered are collected. And the middle-stage operation stage enters a later-stage operation stage after the data to be recovered with the sample size of 5000-7000 are operated. A complete sample size comprises all data generated in the complete process from inputting the health information interaction platform to the user obtaining the satisfied rehabilitation video combination, namely stopping using the health information interaction platform naturally.
The initial operation stage generally comprises an AHP model, an SPV model and a planning model; the mid-stage operational phase generally includes a decision tree model, a BP neural network model, and a support vector machine model. And the later operation stage introduces a convolutional neural network to establish a deep learning system.
The improved intelligent algorithm used in this embodiment is a differential evolution algorithm. The method comprises the following specific steps:
the method comprises the following steps: and (5) initializing. Adopting a differential evolution algorithm to use NP real-value parameter vectors with dimension D as a population of each generation, wherein each individual is expressed as: x
i,G(i＝1,2,...,NP)
In the formula: i represents the sequence of an individual in a population; g represents evolution algebra; NP indicates population size and NP remains unchanged during minimization.
It is assumed that a uniform probability distribution is met for all randomly initialized populations. Setting the limit of the parameter variable as
Then:
where rand [0,1] is a uniform random number generated between [0,1 ].
Step two: and (5) carrying out mutation. For each target vector X
i,G(i 1, 2.., NP), the variant vector of the basic differential evolution algorithm is generated as follows:
wherein the randomly selected sequence number r
1r
2And r
3Are different from each other, and r
1r
2And r
3Since the target vector index i should be different, NP.gtoreq.4 must be satisfied. Mutation operator F ∈ [0,2 ]]Is a real constant factor that controls the amplification of the variation variable. Because the mutation operator takes a real constant, the mutation operator is difficult to determine in the implementation, the mutation rate is too large, the algorithm searching efficiency is low, and the problem of solving the problem is solvedThe obtained global optimal solution has low precision; the mutation rate is too small, the population diversity is reduced, and the phenomenon of early ripening is easy to occur. Therefore, the following adaptive mutation operator λ is added, and according to the algorithm search progress, the adaptive mutation operator is designed as follows:
in the formula: f
0Is a mutation operator; g
mFor maximum evolutionary algebra: g is the current evolution algebra.
The adaptive mutation operator at the beginning of the algorithm is F
0～2F
0The method has a large value, keeps the individual diversity in the initial stage and avoids precocity; the mutation operator is gradually reduced along with the algorithm progress until the mutation rate at the later stage approaches F
0And good information is kept, the optimal solution is prevented from being damaged, and the probability of searching the global optimal solution is increased.
Step three: and (4) crossing. In order to increase the diversity of the interference parameter vectors, interleaving is introduced. The trial vector becomes:
u
i,G+1＝(u
1i,G+1,u
2i,G+1,...,u
Di,G+1)
(i＝1,2,...NP；j＝1,3,...,D)
in the formula: randb (j) generating [0,1]J-th estimated value of random number generator; rnb (i) e1, 2, D is a randomly selected sequence used to ensure u
i,G+1At least from V
i,G+1Obtaining a parameter; CR is a crossover operator and has a value range of [0,1]]。
In order to keep the diversity of the population, the following random range crossover operators are designed:
p
0＝rand(0,1)；
step four: and (4) selecting. To determine the test vector u
i,G+1Whether the test vector can become a member in the next generation or not is judged, and the test vector and the target vector X in the current population are combined according to a greedy criterion
i,GA comparison is made. If the objective function is to be minimized, the vector with the smaller objective function value will win a position in the next generation population. All individuals in the next generation are better than or at least as good as the corresponding individuals of the current population.
Step five: and processing boundary conditions. In the case of a problem with boundary constraints, it is necessary to ensure that the parameter values of the new individuals that are generated are in the feasible domain of the problem, and a simple approach is to replace the new individuals that do not meet the boundary constraints with a parameter vector that is randomly generated in the feasible domain.
Namely: if it is
Or
Then:
test example:
for comparison, the health information interaction platform of the embodiment (the present example) and the existing health information interaction platform (the comparative example) are selected for testing, and the freshness and the manufacturing materials of the two health information interaction platforms are the same.
First, two groups of panelists of the same age, sex, and condition were selected and used at the same time every other day in the same period using the health information exchange platform of the present example and the existing health information exchange platform of the present comparative example. For comparison, users with the same age group and disease condition respectively adopt people of both male and female sexes for comparison. The response time from the completion of inputting the data to be rehabilitated to the receipt of the pushed rehabilitation video combination by each team member is recorded in table 6.
TABLE 6
The unit in table 5 is second, where the response time of the first test is the health information interaction platform when there is no data to be recovered at all. As can be seen from the above table, the response time of a person of each age group is the shortest between a person of 20 years and a person of 30 years in the same time period, which is related to the skilled operation of the persons of both ages. And the response times of men and women of the same age group and the same condition in the same group do not differ much.
In two groups of members with the same disease condition, the same age group and the same sex in each row, the member using the health information interaction platform in the embodiment has shorter response time when being used for the first time (without self rehabilitation data storage), and the health information interaction platform in the embodiment can shorten the response time along with the increase of the number of use times, while the response time of the member using the existing health information interaction platform is gradually shortened along with the increase of the number of use times, but the response time of the corresponding time period is longer than that of the other member. It is fully explained that the health information interaction platform in this embodiment can still be normally used without the data to be recovered, that is, the health information interaction platform in this embodiment does not depend on the data to be recovered. Meanwhile, the machine learning system can continuously perform learning optimization along with the data to be recovered continuously input by the user in the using process, so that the calculation speed is increased, and the whole health information interaction platform can quickly push the scheme suitable for the user to the user.
Then, each group member performs rehabilitation training for six months according to the push scheme obtained in the fourth time in table 6, and the rehabilitation effect of each group member is evaluated in the third month, the fourth month, the fifth month and the sixth month. The evaluation criteria are the same hospital visit to complete the same standard physical examination, and table 7 is the statistics of the results based on the results of the physical examination, wherein 100 means complete recovery, 0 means complete non-recovery, i.e. the push protocol is not functional at all.
TABLE 7
As can be seen from the above table, the recovery rate of people of each age group is different in the same time period, wherein the recovery rate of people of 10 years, 20 years and 30 years is faster, which is related to the physical quality of the people of the three ages. While men and women of the same age group and the same condition have a slightly different rate of recovery.
For two groups of members with the same disease conditions, the same age group and the same sex in each row, the physical examination result of the health information interaction platform in the embodiment is better than that of the other group of members in each physical examination month, so that the scheme of pushing by using the health information interaction platform in the embodiment is more suitable for users, the accuracy of pushing the rehabilitation video combination is higher, the generated effect is better, and the rehabilitation effect obtained by the health information interaction platform through the same data to be rehabilitated in the embodiment is better.
The foregoing is merely an example of the present invention, and common general knowledge in the field of known specific structures and characteristics is not described herein in any greater extent than that known in the art at the filing date or prior to the priority date of the application, so that those skilled in the art can now appreciate that all of the above-described techniques in this field and have the ability to apply routine experimentation before this date can be combined with one or more of the present teachings to complete and implement the present invention, and that certain typical known structures or known methods do not pose any impediments to the implementation of the present invention by those skilled in the art. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.
Claims (7)
1. The health information interaction system comprises a cloud server and a plurality of clients connected with the cloud server through a network; the method is characterized in that: the client comprises a first processor, and an input assembly, a display screen and a first memory which are respectively connected with the first processor; the cloud server comprises a second processor, a second memory, a database and a verification assembly, wherein the second memory is connected with the second processor and stores a plurality of rehabilitation video combinations, the database is used for updating a user questionnaire and answers of the user questionnaire in real time, and the verification assembly is used for verifying identity information input by a user and answers of the questionnaire; a grade reference table is arranged in the first memory; the relation between the grade reference table convention and the video number; the video numbers correspond to the rehabilitation video combinations stored in the database one by one;
the verification module is used for connecting the video number with the specific video; a second correlation parameter which is set by the expert for the rehabilitation video combination correspondingly according to each option in each factor in the questionnaire is arranged in the verification module; and (3) taking the total number of the rehabilitation video combinations as a random variable number, and constraining the priority relationship or the incidence relationship among the rehabilitation video combinations by taking the standard combination number of the rehabilitation video combination scheme given by an expert as a model constraint condition to establish a planning model.
2. The health information interaction system according to claim 1, wherein: the first processor is internally provided with a judging module for judging input information of the input assembly; an AHP model is arranged in the judging module; the AHP model adopts a two-stage method to solve the relative weight of each evaluation factor of the questionnaire; the first processor calculates a first association constant of the user according to the weight value of each evaluation factor, and the first processor transmits description data corresponding to the first association constant, which is set in the first processor, to the display screen.
3. The health information interaction system according to claim 1, wherein: an optimization module is also arranged in the first processor; doctor guidance suggestion data compiled according to various common phenomena are arranged in the optimization module; the optimization module calculates corresponding video numbers through doctor guidance opinion data, calculates absolute values of differences between video combinations and the maximum treatment degrees of experts by taking the adjustment quantities of the video numbers as variables, obtains multiple groups of comprehensive parameter adjustment quantities by adopting an improved intelligent algorithm to perform successive iteration, performs factor option distribution and entropy processing on the adjustment quantities according to weights, and performs adjustment optimization on video number tables corresponding to the rehabilitation video combinations.
4. The health information interaction system according to claim 1, wherein: a screening module is arranged in the second processor; the screening module takes the data to be recovered updated in real time in the database as input data and category data of a decision tree model arranged in the screening module, classifies problems in questionnaires corresponding to the data to be recovered, and adopts a CART algorithm to construct the decision tree model; the screening module transmits a judgment result of whether the rehabilitation video combination needs to be pushed or not to the first processor according to the user data input from the input assembly.
5. The health information interaction system according to claim 1, wherein: a learning module arranged in the verification component trains and learns all data transmitted to the database through the BP neural network to obtain corresponding BP neural network parameters and stores the parameters in the database; when the verification component receives the judgment result of the rehabilitation video combination, the learning module outputs a corresponding video number of the pushed rehabilitation video combination to the second processor according to the data to be rehabilitated input by the input component and the BP neural network parameter; meanwhile, the background program updates the BP neural network parameters before the data on line.
6. The health information interaction system according to claim 1, wherein: the second processor is also provided with a correction module; the correction module is internally provided with a support vector machine model, judges whether to carry out video push again according to feedback data of a user, and meanwhile, the second processor updates the relation between the rehabilitation video combination and the video number in real time according to the feedback data of the user; and transmitting the corrected associated data to a database for storage.
7. The health information interaction system of claim 1, wherein the operation process of the health information interaction system comprises at least two phases, wherein the initial operation phase enters the intermediate operation phase after collecting 1000-2000 samples of data to be rehabilitated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610881071.6A CN106682385B (en) | 2016-09-30 | 2016-09-30 | Health information interaction system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610881071.6A CN106682385B (en) | 2016-09-30 | 2016-09-30 | Health information interaction system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106682385A CN106682385A (en) | 2017-05-17 |
| CN106682385B true CN106682385B (en) | 2020-02-11 |
Family
ID=58840055
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610881071.6A Active CN106682385B (en) | 2016-09-30 | 2016-09-30 | Health information interaction system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106682385B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI704518B (en) * | 2018-04-18 | 2020-09-11 | 智信生醫科技股份有限公司 | Hypoxic factors expert system and method and computer program product thereof |
| CN108766320A (en) * | 2018-06-01 | 2018-11-06 | 深圳安麦思科技有限公司 | A kind of liquid crystal displaying windows system |
| CN111695090B (en) * | 2020-06-15 | 2023-05-05 | 甘肃中医药大学 | Health education investigation evaluation system based on Android platform |
| CN111834005B (en) * | 2020-07-02 | 2022-02-22 | 医渡云（北京）技术有限公司 | Method, device, medium and equipment for screening medical data based on infectious diseases |
| CN113485983B (en) * | 2021-06-07 | 2023-12-26 | 上海众言网络科技有限公司 | Questionnaire data processing method and device |
| CN113593697B (en) * | 2021-07-31 | 2023-10-13 | 吾征智能技术(北京)有限公司 | System, equipment and medium for judging diseases based on breath odor |
| TWI790795B (en) * | 2021-10-29 | 2023-01-21 | 財團法人資訊工業策進會 | Model adjustment method, model adjustment system and non-transitory computer readable medium |
| CN114999652A (en) * | 2022-06-15 | 2022-09-02 | 北京京东拓先科技有限公司 | Health state evaluation method and device and storage medium |
| CN115831371B (en) * | 2023-02-16 | 2023-04-28 | 成都体育学院 | Child physical health assessment system and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001025970A1 (en) * | 1999-10-06 | 2001-04-12 | Accenture Llp | Method and estimator for providing operations maturity model assessment |
| CN103632328A (en) * | 2013-12-05 | 2014-03-12 | 青岛海尔软件有限公司 | Household chronic disease management system with health guidance video playing function |
| CN104794355A (en) * | 2015-04-29 | 2015-07-22 | 孙卫唯 | Method for executing online rehabilitation guidance in computer system and rehabilitation guidance system |
| CN105321135A (en) * | 2014-07-04 | 2016-02-10 | 北京大学第三医院 | Personalized exercise prescription design method and system |
-
2016
- 2016-09-30 CN CN201610881071.6A patent/CN106682385B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2001025970A1 (en) * | 1999-10-06 | 2001-04-12 | Accenture Llp | Method and estimator for providing operations maturity model assessment |
| CN103632328A (en) * | 2013-12-05 | 2014-03-12 | 青岛海尔软件有限公司 | Household chronic disease management system with health guidance video playing function |
| CN105321135A (en) * | 2014-07-04 | 2016-02-10 | 北京大学第三医院 | Personalized exercise prescription design method and system |
| CN104794355A (en) * | 2015-04-29 | 2015-07-22 | 孙卫唯 | Method for executing online rehabilitation guidance in computer system and rehabilitation guidance system |
Non-Patent Citations (2)
| Title |
|---|
| 五自由度上肢康复机器人康复评价系统的研究和实现;张维秋;《中国优秀硕士学位论文全文数据库 信息科技辑》;20150515;第I140-1069页 * |
| 基于ＢＰ神经网络的学生运动处方管理系统;刘小菊 等;《浙江体育科学》;20150930;第37卷(第5期);第79-84页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106682385A (en) | 2017-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106682385B (en) | Health information interaction system | |
| CN106779084B (en) | Machine learning system and method | |
| Shukla | Neuro-genetic prediction of software development effort | |
| Papić et al. | Identification of sport talents using a web-oriented expert system with a fuzzy module | |
| CN106778014A (en) | A kind of risk Forecasting Methodology based on Recognition with Recurrent Neural Network | |
| CN107230113A (en) | A kind of house property appraisal procedure of multi-model fusion | |
| CN106295186A (en) | A kind of method and system of aided disease diagnosis based on intelligent inference | |
| CN108573303A (en) | It is a kind of that recovery policy is improved based on the complex network local failure for improving intensified learning certainly | |
| CN106485069A (en) | The method and system of rehabilitation information pushing | |
| CN109447167B (en) | Intelligent cigarette formula maintenance method based on non-negative matrix factorization | |
| CN109754317A (en) | Merge interpretation clothes recommended method, system, equipment and the medium of comment | |
| CN115644823B (en) | Dynamic prediction and individualized intervention system for rehabilitation effect | |
| TW202004771A (en) | Control method and medical system | |
| CN111640483A (en) | Fitness scheme recommendation method based on AKC model | |
| CN107491992A (en) | A kind of intelligent Service proposed algorithm based on cloud computing | |
| CN106651001A (en) | Needle mushroom yield prediction method based on improved neural network and implementation system | |
| CN107578822B (en) | Pretreatment and feature extraction method for medical multi-modal big data | |
| CN107103419A (en) | One kind of groups software development process analogue system and method | |
| CN106447117A (en) | Pet feeding method and system based on pet daily data analysis | |
| Sarra et al. | Enhanced accuracy for heart disease prediction using artificial neural network | |
| CN109583272B (en) | Footprint system capable of acquiring living state of human body | |
| CN113449182A (en) | Knowledge information personalized recommendation method and system | |
| CN116759100A (en) | Method for constructing chronic cardiovascular disease large model based on federal learning | |
| CN116304289A (en) | Information chain recommendation method and device for supply chain based on graphic neural network | |
| CN112329921B (en) | Diuretic dose reasoning equipment based on deep characterization learning and reinforcement learning |
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 |