CN108182967A - A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method - Google Patents
A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method Download PDFInfo
- Publication number
- CN108182967A CN108182967A CN201711338966.6A CN201711338966A CN108182967A CN 108182967 A CN108182967 A CN 108182967A CN 201711338966 A CN201711338966 A CN 201711338966A CN 108182967 A CN108182967 A CN 108182967A
- Authority
- CN
- China
- Prior art keywords
- vector
- traditional chinese
- medical science
- chinese medical
- medicinal material
- 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.)
- Pending
Links
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a kind of traditional Chinese medical science medicinal materials based on deep neural network to recommend method, including:The input of the feature vector of computer acquisition human body tongue fur picture and the digital code of corresponding traditional Chinese medical science medicinal material as data set;Dimension-reduction treatment is carried out to the feature vector of tongue fur picture;Embedded characterization processing is carried out to the digital code of traditional Chinese medicine material;According to the tongue fur feature vector of input and corresponding traditional Chinese medical science medicinal material feature vector, the relationship between tongue fur and traditional Chinese medical science medicinal material is learnt using proposed algorithm;According to the relationship of the tongue fur learnt and traditional Chinese medical science medicinal material, the degree of association score between tongue fur picture and traditional Chinese medical science medicinal material is calculated using traditional Chinese medical science medicinal material proposed algorithm.The present invention reduces the repetitive operation in traditional Chinese medical science prescription, and can realize that accuracy is high, speed is fast, performance steadily carries out corresponding traditional Chinese medical science medicinal material according to tongue fur and recommends for assisting the traditional Chinese medical science rapidly and accurately prescription.
Description
Technical field
It is more particularly to a kind of based on depth nerve the present invention relates to the applied technical field that computer is issued in tcm prescription
The traditional Chinese medical science medicinal material of network recommends method.
Background technology
In China, the traditional Chinese medical science has had the history of thousands of years for prevention disease and the research for treating disease.Wherein, lingual diagnosis is
Unique and important content and the important evidence of tcm diagnosis in tcm inspection, are the form and tongue fur by observing tongue
Color carry out a kind of important method of adjuvant treatment of diseases.
Prescription, also referred to as prescription are the wisdom crystallization and component part of Traditional Chinese Medicine culture, be to treat certain disease and
The title of several drug that combines, dosage and administration.
Commending system is mainly that the characteristic information in interest demand information and item objects model in user model is matched,
Calculating sifting is carried out using corresponding proposed algorithm simultaneously, the possible interested recommended project of user is found, project recommendation is given
User.In the present invention, tongue fur is the user in commending system, and traditional Chinese medical science medicinal material is then project, i.e., is believed according to different tongue fur features
Breath recommends corresponding traditional Chinese medical science medicinal material.
Commending system can be realized as a kind of intelligent personalized information service system by certain intelligent recommendation strategy
Targetedly customized information customizes, and is ground although still facing problems, the proposed algorithms such as similar cold start-up, Sparse row
Study carefully be still artificial intelligence industry research hotspot.Armstrong R etc. [1] propose the proposed algorithm of Cempetency-based education;
Linden G etc. [2] propose project-based collaborative filtering.
As deep learning quickly becomes industry hot spot, more and more deep learnings are applied in recommendation field.
Salakhutdinov etc. [3] proposes limitation Boltzmann machine (Restricted Boltzmann Machine, RBM) model,
It realizes and recommends possible interested project to user.Wang H etc. [4] are proposed using the stack denoising own coding phase from item
Feature is extracted in mesh, and used in collaborative filtering model, so as to improve recommendation performance.
Deep learning can carry out original input data by the model structure of layering from bottom to high-rise data
Feature extraction, so as to set up bottom data to the mapping relations between high-level semantic.Using deep neural network, can build
The training data of neural network model and magnanimity with multilayer hidden layer, so as to learn to more useful character representation,
So as to improve the accuracy of recommendation results.
In achievement in research before, there are some system and method being the theme with prescription, medicinal material recommendation in the country.Its
In, Wang Benyu [5] proposes one kind and is fitted energy test result automation based on physical examination data, exercise risk assessment and body and generated fortune
The method of dynamic prescription;Zhang Guanjing etc. [6] has invented one kind and doctor is enabled efficiently to output prescription, improves the diagnosis efficiency of doctor,
The system of the medical treatment efficiency of patient and the medical service quality of hospital;Yuan Weiwei etc. [7] has invented a kind of to specific crowd
Property chemical drug object recommend method, belong to drug assessment technique field in pharmacoeconomics.
In above-mentioned various systems approaches, 1) these systems approaches are more to be applied to the generation of doctor trained in Western medicine prescription, Western medicine medicinal material
Recommend rather than for the traditional Chinese medical science;2) generation of these prescriptions, medicinal material recommend method to need various index number according to physical examination
According on data acquisition, process is more numerous and diverse, and the time of waiting is long.
[1]WebWatcher:A Learning Apprentice for the World Wide Web[J].In
Working Notes of the Aaai Spring Symposium Series on Information Gathering
from Distributed,1995,30(6):6-12.
[2]Amazon.com recommendations:item-to-item collaborative filtering
[J].IEEE Internet Computing,2003,7(1):76-80.
[3]Restricted Boltzmann machines for collaborative filtering[C]
.International Conference on Machine Learning.ACM,2007:791-798.
[4]Collaborative Deep Learning for Recommender Systems[J].2014:1235-
1244.
[5] the method Chinese patents .2016 of automation generation exercise prescription
[6] the system and method Chinese patents .2016 that auxiliary doctor prescribes
[7] a kind of personalised drug based on probability recommends method Chinese patents .2016
Invention content
The shortcomings that it is an object of the invention to overcome the prior art and deficiency, provide a kind of based in deep neural network
Medical material recommends method, learns the relationship between tongue fur and traditional Chinese medical science medicinal material using deep neural network, makes full use of TCM tongue diagnosis
Importance, so as to realize that accuracy is high, speed is fast, performance steadily carries out corresponding traditional Chinese medical science medicinal material according to tongue fur and pushes away
It recommends.
The purpose of the present invention is realized by following technical solution:
A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method, includes the following steps:
S1, the feature vector of computer acquisition human body tongue fur picture and corresponding traditional Chinese medical science medicinal material digital code are as data set
Input;
S2, dimension processing is carried out to the feature vector of tongue fur picture;
S3, embedded characterization processing is carried out to the digital code of traditional Chinese medicine material;
Relationship between S4, study tongue fur picture and traditional Chinese medical science medicinal material;
S5, according to the relationship between tongue fur picture and traditional Chinese medical science medicinal material, tongue fur figure is calculated using traditional Chinese medical science medicinal material proposed algorithm
Degree of association score between piece and traditional Chinese medical science medicinal material, and choose the traditional Chinese medical science medicinal material as recommendation of degree of association highest scoring.
Preferably, in the step S2, the feature vector of tongue fur picture is carried out at dimensionality reduction using Principal Component Analysis Algorithm
Reason.
Preferably, the step S4 is specially:
S41, the method being superimposed using vector element, by the tongue fur feature vector vector of step S2uIn step S3
Medical material feature vector vectoriMerge into a new vector v ectoroutput, form equation below:
vectoroutput=[vectoru,vectori];
S42, the output vector vector by step S41outputAs the input vector of neural network, using neural network
Algorithm learns the relationship between tongue fur and traditional Chinese medical science medicinal material.
Further, neural network algorithm includes collaborative filtering and content-based filtering algorithm in step S42.
Specifically, in step S3, embedded characterization processing refers to convert the input into the output vector of specific dimension.
Specifically, a kind of traditional Chinese medical science medicinal material based on deep neural network recommends method, including training stage and test phase,
Wherein training stage step includes:
[1] the tongue fur picture feature vector dimensionality reduction that will be originally inputted, is denoted as vectoru, by traditional Chinese medical science medicinal material digital code into
The embedded characterization of row handles to obtain traditional Chinese medical science medicinal material feature vector, is denoted as vectori;
[2] by feature vector vectoruAnd vectoriA new vector is merged into, is denoted as vectorui;
[3] by feature vector vectoruiNeural network first layer is inputted, is denoted as vectorout1;
[4] by vectorout1The neural network second layer is inputted, is denoted as vectorout2;
[5] by vectorout2Neural network third layer is inputted, is denoted as vectorout3;
[6] by vectorout3Neural network output layer is inputted, is denoted as vectorout4;
[7] by vectorout4The last output of the neural network, as tongue fur are calculated in input sigmoid functions
The degree of association score y of picture and traditional Chinese medical science medicinal materialp;
[8] by the degree of association score value y of predictionpCompared with target score value y, its loss function and gradient G are calculated;
[9] process of above-mentioned [1] to [8] is repeated, until the value Jing Guo enough iteration or loss function is no longer bright
It is aobvious to become smaller;
Test phase step includes:
[1] by original tongue fur picture feature vector dimensionality reduction, it is denoted as vectoru, traditional Chinese medical science medicinal material digital code is carried out embedding
Enter formula characterization to handle to obtain traditional Chinese medical science medicinal material feature vector, be denoted as vectori;
[2] by feature vector vectoruAnd vectoriA new vector is merged into, is denoted as vectorui;
[3] it is loaded into the deep neural network model after training;
[4] by vectoruiBe input in the deep neural network model, respectively by neural network first layer, the second layer,
Third layer, output layer obtain output result vector vectorout;
[5] by vectoroutThe last output of the neural network, as tongue fur figure are calculated in input sigmoid functions
The degree of association score y of piece and traditional Chinese medical science medicinal materialp;
[6] process of [1] to [5] is repeated, calculates the degree of association score of the tongue fur picture and other various traditional Chinese medical science medicinal materials;
[7] consequently recommended medicinal material of the high medicinal material of degree of association score as tongue fur picture is chosen.
Further, neural network calculates the degree of association score y of tongue fur and traditional Chinese medical science medicinal material using Sigmoid functionsp, form
Equation below:
Wherein, P, Q represent the hidden factor matrix of tongue fur and traditional Chinese medical science medicinal material, θ respectivelyfRepresent the model ginseng of interaction function f ()
Number;Since f () is defined as a neural network containing three layers of hidden layer, f () can use following formula subrepresentation:
Wherein, φoutAnd φ1、φ2、φ3Output layer and the 1st, 2,3 layer of neural collaborative filtering layer are represented respectively.
Compared with prior art, the present invention having the following advantages that and advantageous effect:
1st, the method for the present invention uses deep neural network, by deep neural network to the expression ability of non-linear relation,
Relationship that can be well between the feature vector of tongue fur picture feature vector sum traditional Chinese medical science medicinal material learns, so as to recommend
With the traditional Chinese medical science medicinal material of tongue fur symptom.
2nd, the method for the present invention calculates the degree of association score of tongue fur and traditional Chinese medical science medicinal material using Sigmoid functions, can be good at
The degree of association between each traditional Chinese medical science medicinal material and tongue fur is distinguished, obtains the outstanding recommendation effect of comparison.
3rd, the method for the present invention with it is traditional it is artificial issue tcm prescription compared with, not only reduce prescription and issue the time, and
Accuracy is high.
4th, the data set of tongue fur picture and tcm prescription medicinal material of the method for the present invention based on magnanimity, by deep learning and recommendation
Algorithmic technique is applied to traditional traditional Chinese medical science medicinal material and recommends field, can not only carry out traditional Chinese medical science medicinal material recommendation, Er Qieke by computer
Very convenient to carry out traditional Chinese medical science medicinal material recommendation by mobile terminal, accuracy is high, and saves the time.
5th, deep learning is combined by the present invention with traditional traditional Chinese medical science medicinal material recommendation, is pushed away on the basis of big data
Recommend, solve the problems, such as to need to preengage doctor of traditional Chinese medicine in tradition and recommend medicinal material, this method have certain market value and
Promotional value.
Description of the drawings
Fig. 1 is the step flow chart of embodiment method.
Fig. 2 is the schematic diagram of deep neural network algorithm model in embodiment method.
Specific embodiment
With reference to embodiment and attached drawing, the present invention is described in further detail, but embodiments of the present invention are unlimited
In this.
Embodiment
As shown in Figure 1, a kind of traditional Chinese medical science medicinal material based on deep neural network recommends method, step includes:
S1, the feature vector of computer acquisition human body tongue fur picture and corresponding traditional Chinese medical science medicinal material digital code are as data set
Input;
S2, dimension processing is carried out to the feature vector of tongue fur picture;
S3, embedded characterization processing is carried out to the digital code of traditional Chinese medicine material;
Relationship between S4, study tongue fur picture and traditional Chinese medical science medicinal material:
S41, the method being superimposed using vector element, by the tongue fur feature vector of step S2 and the traditional Chinese medical science medicinal material feature of S3 to
Amount merges into a new vector.
S42, using the output vector of step S41 as the input vector of neural network, tongue is learnt using neural network algorithm
Relationship between tongue and traditional Chinese medical science medicinal material, neural network algorithm include collaborative filtering and content-based filtering algorithm.
S5, according to the relationship between tongue fur picture and traditional Chinese medical science medicinal material, tongue fur figure is calculated using traditional Chinese medical science medicinal material proposed algorithm
Degree of association score between piece and traditional Chinese medical science medicinal material, and choose the traditional Chinese medical science medicinal material as recommendation of degree of association highest scoring, traditional Chinese medicine
Material proposed algorithm calculates the degree of association score of tongue fur and traditional Chinese medical science medicinal material using Sigmoid functions.
Wherein, the data set of tongue fur picture and traditional Chinese medical science medicinal material is manually to be acquired in provincial institute of traditional Chinese medicine and pass through associated specialist mark
The mass data collection of note, and needed according to neural network algorithm data set being randomly divided into training set and test set.According to this
The flow of method, needs to carry out dimension-reduction treatment to tongue fur picture feature vector, and embedded coding is carried out to traditional Chinese medicine material digital code
Processing obtains the feature vector of traditional Chinese medical science medicinal material.
In step s 2, Feature Dimension Reduction is carried out to tongue fur picture feature vector using Principal Component Analysis Algorithm.
Then using the element overlaid method of vector, by the tongue fur picture feature vector after dimension-reduction treatment and traditional Chinese medical science medicinal material
Feature vector is superimposed as new feature vector, the input as neural network first layer.
As shown in Fig. 2, deep neural network, using the neural network connected entirely, effect is by tongue fur feature vector
With traditional Chinese medical science medicinal material feature vector as inputting, by full Connection Neural Network model, obtain effectively representing tongue fur and the traditional Chinese medical science
Relationship characteristic between medicinal material.
Tongue fur original feature vector refers to tongue fur picture size being adjusted to 32*32, and the RGB of each pixel of picture is taken to put down
Mean value, and it is launched into the feature vector that dimension is 1*1024.
Traditional Chinese medical science medicinal material digital code refers to be identified various traditional Chinese medical science medicinal materials with a different dimension words.
PCA, i.e. principal component analysis carry out the tongue fur feature vector of 1024 dimensions dimension-reduction treatment using PCA, and extraction dimension is
64 tongue fur feature vector.
Embedded characterization processing refers to convert the input into the output vector of specific dimension.
Tongue fur picture feature vector refers to the tongue fur feature vector extracted by Principal Component Analysis Algorithm.
Traditional Chinese medical science medicinal material feature vector refers to obtain specific dimension after the processing of embedded characterization, for representing traditional Chinese medical science medicinal material
Feature vector.
Vector element merging refers to tail before tongue fur picture feature vector sum traditional Chinese medical science medicinal material feature vector being connected to form one
New vector, the dimension of the vector after merging be tongue fur picture feature vector sum traditional Chinese medical science medicinal material feature vector dimension and, it is new
Input vector of the vector as neural network.
As shown in Fig. 2, its neural network is full Connection Neural Network, detailed design such as following table:
Table 1
Layer name | Neuron number | Activation primitive | Output vector dimension |
Layer 1 | 64 | ReLU | 1*64 |
Layer 2 | 32 | ReLU | 1*32 |
Layer 3 | 16 | ReLU | 1*16 |
Out Layer | 8 | Sigmoid | 1*1 |
As shown in Fig. 2, Sigmoid is the function of a common S type in biology, in information science, due to it
Properties, the Sigmoid functions such as single increasing and the increasing of inverse function list are often used as the threshold function table of neural network, make in the present embodiment
With the function by the DUAL PROBLEMS OF VECTOR MAPPING that the output dimension of neural network is 1*8 a to floating number between [0,1].
As shown in Fig. 2, ypThe degree of association between tongue fur picture and traditional Chinese medical science medicinal material exactly after neural computing obtains
Point.
The deep neural network is divided into training stage and test phase, and wherein training stage step is:
[1] the tongue fur picture feature vector dimensionality reduction being originally inputted is adjusted to 1*64, is denoted as vectoru, by traditional Chinese medical science medicinal material
Digital code carries out embedded characterization and handles to obtain traditional Chinese medical science medicinal material feature vector, and dimension 1*64 is denoted as vectori。
[2] by feature vector vectoruAnd vectoriA new vector is merged into, dimension 1*128 is denoted as
vectorui。
[3] by feature vector vectoruiNeural network first layer is inputted, the dimension for adjusting this layer output result is 1*64,
It is denoted as vectorout1。
[4] by vectorout1The neural network second layer is inputted, the dimension for adjusting this layer output result is 1*32, is denoted as
vectorout2。
[5] by vectorout2Neural network third layer is inputted, the dimension for adjusting this layer output result is 1*16, is denoted as
vectorout3。
[6] by vectorout3Neural network output layer is inputted, the dimension for adjusting this layer output result is 1*8, is denoted as
vectorout4。
[7] by vectorout4The last output of the neural network, as tongue fur are calculated in input sigmoid functions
The degree of association score y of picture and traditional Chinese medical science medicinal materialp。
[8] by the degree of association score value y of predictionpCompared with target score value y, its loss function and gradient G are calculated.
[9] process of above-mentioned [1] to [8] is repeated, until the value Jing Guo enough iteration or loss function is no longer bright
It is aobvious to become smaller.
In the prediction degree of association score being calculated, preceding 10 kinds of medicinal materials of highest scoring is selected to be pushed away as final medicinal material
Recommend list, the part iteration result in training process is as shown in the table:
Table 2
Epoch | Medicinal material recommends HR | Medicinal material recommends NDCG | Prescription generates accuracy | Prescription generates recall rate |
11 | 0.6771 | 0.3815 | 0.6911 | 0.3779 |
12 | 0.6777 | 0.3802 | 0.6944 | 0.3618 |
13 | 0.6782 | 0.3811 | 0.6909 | 0.3772 |
14 | 0.6767 | 0.3810 | 0.6957 | 0.3673 |
15 | 0.6791 | 0.3818 | 0.6947 | 0.3723 |
The test phase step of above-mentioned deep neural network is:
[1] original tongue fur picture feature vector dimensionality reduction is adjusted to 1*64, is denoted as vectoru, by traditional Chinese medical science medicinal material number
Code name carries out embedded characterization and handles to obtain traditional Chinese medical science medicinal material feature vector, and dimension 1*64 is denoted as vectori。
[2] by feature vector vectoruAnd vectoriA new vector is merged into, dimension 1*128 is denoted as
vectorui。
[3] it is loaded into the deep neural network model after training.
[4] by vectoruiBe input in the deep neural network model, respectively by neural network first layer, the second layer,
Third layer, output layer obtain output result vector vectorout。
[5] by vectoroutThe last output of the neural network, as tongue fur figure are calculated in input sigmoid functions
The degree of association score y of piece and traditional Chinese medical science medicinal materialp。
[6] process of [1] to [5] is repeated, calculates the degree of association score of the tongue fur picture and other various traditional Chinese medical science medicinal materials.
[7] consequently recommended medicinal material of the high medicinal material of degree of association score as tongue fur picture is chosen.
This method is used to assist the traditional Chinese medical science rapidly and accurately prescription, reduces the repetitive operation in traditional Chinese medical science prescription,
And it can realize that accuracy is high, speed is fast, performance steadily carries out corresponding traditional Chinese medical science medicinal material according to tongue fur and recommends.
Above-described embodiment is the preferable embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any Spirit Essences without departing from the present invention with made under principle change, modification, replacement, combine, simplification,
Equivalent substitute mode is should be, is included within protection scope of the present invention.
Claims (7)
1. a kind of traditional Chinese medical science medicinal material based on deep neural network recommends method, which is characterized in that includes the following steps:
S1, the feature vector of computer acquisition human body tongue fur picture and corresponding traditional Chinese medical science medicinal material digital code are as the defeated of data set
Enter;
S2, dimension processing is carried out to the feature vector of tongue fur picture;
S3, embedded characterization processing is carried out to the digital code of traditional Chinese medicine material;
Relationship between S4, study tongue fur picture and traditional Chinese medical science medicinal material;
S5, according to the relationship between tongue fur picture and traditional Chinese medical science medicinal material, using traditional Chinese medical science medicinal material proposed algorithm calculate tongue fur picture with
Degree of association score between traditional Chinese medical science medicinal material, and choose the traditional Chinese medical science medicinal material as recommendation of degree of association highest scoring.
2. the traditional Chinese medical science medicinal material according to claim 1 based on deep neural network recommends method, which is characterized in that the step
In rapid S2, dimension-reduction treatment is carried out to the feature vector of tongue fur picture using Principal Component Analysis Algorithm.
3. the traditional Chinese medical science medicinal material according to claim 1 based on deep neural network recommends method, which is characterized in that the step
Suddenly S4 is specially:
S41, the method being superimposed using vector element, by the tongue fur feature vector vector of step S2uWith the traditional Chinese medical science medicinal material of step S3
Feature vector vectoriMerge into a new vector v ectoroutput, form equation below:
vectoroutput=[vectoru,vectori];
S42, the output vector vector by step S41outputAs the input vector of neural network, using neural network algorithm
Learn the relationship between tongue fur and traditional Chinese medical science medicinal material.
4. the traditional Chinese medical science medicinal material according to claim 3 based on deep neural network recommends method, which is characterized in that step
Neural network algorithm includes collaborative filtering and content-based filtering algorithm in S42.
5. the traditional Chinese medical science medicinal material according to claim 1 based on deep neural network recommends method, which is characterized in that step S3
In, embedded characterization processing refers to convert the input into the output vector of specific dimension.
6. the traditional Chinese medical science medicinal material according to claim 1 based on deep neural network recommends method, which is characterized in that including instruction
Practice the stage and test phase, wherein training stage step include:
[1] the tongue fur picture feature vector dimensionality reduction that will be originally inputted, is denoted as vectoru, traditional Chinese medical science medicinal material digital code is embedded in
Formula characterization handles to obtain traditional Chinese medical science medicinal material feature vector, is denoted as vectori;
[2] by feature vector vectoruAnd vectoriA new vector is merged into, is denoted as vectorui;
[3] by feature vector vectoruiNeural network first layer is inputted, is denoted as vectorout1;
[4] by vectorout1The neural network second layer is inputted, is denoted as vectorout2;
[5] by vectorout2Neural network third layer is inputted, is denoted as vectorout3;
[6] by vectorout3Neural network output layer is inputted, is denoted as vectorout4;
[7] by vectorout4Be calculated the last output of the neural network in input sigmoid functions, as tongue fur picture and
The degree of association score y of traditional Chinese medical science medicinal materialp;
[8] by the degree of association score value y of predictionpCompared with target score value y, its loss function and gradient G are calculated;
[9] process of above-mentioned [1] to [8] is repeated, until the value Jing Guo enough iteration or loss function no longer significantly becomes
It is small;
Test phase step includes:
[1] by original tongue fur picture feature vector dimensionality reduction, it is denoted as vectoru, traditional Chinese medical science medicinal material digital code is subjected to embedded table
Sign handles to obtain traditional Chinese medical science medicinal material feature vector, is denoted as vectori;
[2] by feature vector vectoruAnd vectoriA new vector is merged into, is denoted as vectorui;
[3] it is loaded into the deep neural network model after training;
[4] by vectoruiIt is input in the deep neural network model, respectively by neural network first layer, the second layer, third
Layer, output layer obtain output result vector vectorout;
[5] by vectoroutBe calculated the last output of the neural network in input sigmoid functions, as tongue fur picture and
The degree of association score y of traditional Chinese medical science medicinal materialp;
[6] process of [1] to [5] is repeated, calculates the degree of association score of the tongue fur picture and other various traditional Chinese medical science medicinal materials;
[7] consequently recommended medicinal material of the high medicinal material of degree of association score as tongue fur picture is chosen.
7. the traditional Chinese medical science medicinal material according to claim 6 based on deep neural network recommends method, which is characterized in that nerve net
Network calculates the degree of association score y of tongue fur and traditional Chinese medical science medicinal material using Sigmoid functionsp, form equation below:
Wherein, P, Q represent the hidden factor matrix of tongue fur and traditional Chinese medical science medicinal material, θ respectivelyfRepresent the model parameter of interaction function f ();By
It is defined as a neural network containing three layers of hidden layer, therefore f () can use following formula subrepresentation in f ():
Wherein, φoutAnd φ1、φ2、φ3Output layer and the 1st, 2,3 layer of neural collaborative filtering layer are represented respectively.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711338966.6A CN108182967A (en) | 2017-12-14 | 2017-12-14 | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711338966.6A CN108182967A (en) | 2017-12-14 | 2017-12-14 | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108182967A true CN108182967A (en) | 2018-06-19 |
Family
ID=62545940
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711338966.6A Pending CN108182967A (en) | 2017-12-14 | 2017-12-14 | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108182967A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109448817A (en) * | 2018-09-28 | 2019-03-08 | 小伍健康科技(上海)有限责任公司 | A kind of recipe recommendation method and apparatus based on deep neural network |
CN109961841A (en) * | 2019-03-28 | 2019-07-02 | 广州麦迪森在线医疗科技有限公司 | A kind of optimal doctor's matching system and method towards mobile diagnosis and treatment |
CN110021428A (en) * | 2019-04-03 | 2019-07-16 | 北京大学第三医院(北京大学第三临床医学院) | A method of laser freckle effect is improved using neural network |
CN110176291A (en) * | 2019-04-19 | 2019-08-27 | 周凡 | A kind of health and fitness information recommended method based on deep learning |
CN110490774A (en) * | 2019-08-07 | 2019-11-22 | 北京安龙脉德医学科技有限公司 | Data based on APP acquire first-aid system |
CN110491521A (en) * | 2019-08-07 | 2019-11-22 | 北京安龙脉德医学科技有限公司 | First aid detection device Internet-based |
CN110507294A (en) * | 2019-08-07 | 2019-11-29 | 北京安龙脉德医学科技有限公司 | First-aid system based on internet information transmitting |
CN110970109A (en) * | 2019-11-14 | 2020-04-07 | 厦门中翎易优创科技有限公司 | Deep learning-based orthopaedic postoperative rehabilitation scheme recommendation method |
CN111599444A (en) * | 2020-05-18 | 2020-08-28 | 深圳市悦动天下科技有限公司 | Intelligent tongue diagnosis detection method and device, intelligent terminal and storage medium |
CN112489769A (en) * | 2019-08-22 | 2021-03-12 | 浙江远图互联科技股份有限公司 | Intelligent traditional Chinese medicine diagnosis and medicine recommendation system for chronic diseases based on deep neural network |
CN112652373A (en) * | 2020-12-09 | 2021-04-13 | 华南理工大学 | Prescription recommendation method based on tongue image retrieval |
CN113130043A (en) * | 2019-12-30 | 2021-07-16 | 杭州泉脉科技有限公司 | Intelligent tea drink recommendation system based on traditional Chinese medicine big data |
CN117877686A (en) * | 2024-03-13 | 2024-04-12 | 自贡市第一人民医院 | Intelligent management method and system for traditional Chinese medicine nursing data |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090123574A1 (en) * | 2006-03-31 | 2009-05-14 | Reddy N B Baktha | Compositions for the regression of spongiform pustules |
WO2012014082A2 (en) * | 2010-07-28 | 2012-02-02 | Purapharm International (Hk) Limited | Artificial intelligence and methods for relating herbal ingredients with illnesses in traditional chinese medicine |
CN104951665A (en) * | 2015-07-22 | 2015-09-30 | 浙江大学 | Method and system of medicine recommendation |
CN107242857A (en) * | 2017-06-12 | 2017-10-13 | 南开大学 | The intelligent traditional Chinese medical science based on deep learning integrates diagnosis and therapy system |
-
2017
- 2017-12-14 CN CN201711338966.6A patent/CN108182967A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090123574A1 (en) * | 2006-03-31 | 2009-05-14 | Reddy N B Baktha | Compositions for the regression of spongiform pustules |
WO2012014082A2 (en) * | 2010-07-28 | 2012-02-02 | Purapharm International (Hk) Limited | Artificial intelligence and methods for relating herbal ingredients with illnesses in traditional chinese medicine |
CN104951665A (en) * | 2015-07-22 | 2015-09-30 | 浙江大学 | Method and system of medicine recommendation |
CN107242857A (en) * | 2017-06-12 | 2017-10-13 | 南开大学 | The intelligent traditional Chinese medical science based on deep learning integrates diagnosis and therapy system |
Non-Patent Citations (3)
Title |
---|
XIANGNAN HE ETAL: ""Neural Collaborative Filtering"", 《HTTPS://ARXIV.ORG/ABS/1708.05031V2》 * |
张颖等: ""基于隐语义模型的中医在线辅助诊疗系统"", 《计算机应用》 * |
章新友主编: "《医学成像及处理技术》", 31 July 2011, 中国铁道出版社 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109448817A (en) * | 2018-09-28 | 2019-03-08 | 小伍健康科技(上海)有限责任公司 | A kind of recipe recommendation method and apparatus based on deep neural network |
CN109961841A (en) * | 2019-03-28 | 2019-07-02 | 广州麦迪森在线医疗科技有限公司 | A kind of optimal doctor's matching system and method towards mobile diagnosis and treatment |
CN110021428A (en) * | 2019-04-03 | 2019-07-16 | 北京大学第三医院(北京大学第三临床医学院) | A method of laser freckle effect is improved using neural network |
CN110176291A (en) * | 2019-04-19 | 2019-08-27 | 周凡 | A kind of health and fitness information recommended method based on deep learning |
CN110507294B (en) * | 2019-08-07 | 2022-02-08 | 北京安龙脉德医学科技有限公司 | First-aid system based on internet information transmission |
CN110491521B (en) * | 2019-08-07 | 2022-03-08 | 北京安龙脉德医学科技有限公司 | Internet-based first-aid detection equipment |
CN110507294A (en) * | 2019-08-07 | 2019-11-29 | 北京安龙脉德医学科技有限公司 | First-aid system based on internet information transmitting |
CN110491521A (en) * | 2019-08-07 | 2019-11-22 | 北京安龙脉德医学科技有限公司 | First aid detection device Internet-based |
CN110490774A (en) * | 2019-08-07 | 2019-11-22 | 北京安龙脉德医学科技有限公司 | Data based on APP acquire first-aid system |
CN112489769A (en) * | 2019-08-22 | 2021-03-12 | 浙江远图互联科技股份有限公司 | Intelligent traditional Chinese medicine diagnosis and medicine recommendation system for chronic diseases based on deep neural network |
CN112489769B (en) * | 2019-08-22 | 2024-05-17 | 浙江远图技术股份有限公司 | Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network |
CN110970109A (en) * | 2019-11-14 | 2020-04-07 | 厦门中翎易优创科技有限公司 | Deep learning-based orthopaedic postoperative rehabilitation scheme recommendation method |
CN113130043A (en) * | 2019-12-30 | 2021-07-16 | 杭州泉脉科技有限公司 | Intelligent tea drink recommendation system based on traditional Chinese medicine big data |
CN111599444A (en) * | 2020-05-18 | 2020-08-28 | 深圳市悦动天下科技有限公司 | Intelligent tongue diagnosis detection method and device, intelligent terminal and storage medium |
CN112652373B (en) * | 2020-12-09 | 2022-10-25 | 华南理工大学 | Prescription recommendation method based on tongue image retrieval |
CN112652373A (en) * | 2020-12-09 | 2021-04-13 | 华南理工大学 | Prescription recommendation method based on tongue image retrieval |
CN117877686A (en) * | 2024-03-13 | 2024-04-12 | 自贡市第一人民医院 | Intelligent management method and system for traditional Chinese medicine nursing data |
CN117877686B (en) * | 2024-03-13 | 2024-05-07 | 自贡市第一人民医院 | Intelligent management method and system for traditional Chinese medicine nursing data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108182967A (en) | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method | |
Wu et al. | Vessel-Net: Retinal vessel segmentation under multi-path supervision | |
Elshennawy et al. | Deep-pneumonia framework using deep learning models based on chest X-ray images | |
Montaño-Rojas et al. | Resistance training in breast cancer survivors: a systematic review of exercise programs | |
Schmider et al. | Is it really robust? | |
CN108198604A (en) | A kind of nutrition dietary based on personal characteristics recommends method | |
Larios Delgado et al. | Fast and accurate medication identification | |
Martinez-Martin et al. | PHAROS 2.0—A PHysical assistant RObot system improved | |
Aslani et al. | The Articial Intelligence of Health Sector: How to Advance based on Math and Science? | |
Díaz-García et al. | How mentally fatiguing are consecutive world padel tour matches? | |
CN108877923A (en) | A method of the tongue fur based on deep learning generates prescriptions of traditional Chinese medicine | |
Biondi et al. | An approach for improving automatic mouth emotion recognition | |
Zhu et al. | Application of the Kano model and DEMATEL technique to explore sustainable promotion strategies for Thai-Chinese temples as tourist attractions | |
Khan et al. | Detecting breast arterial calcifications in mammograms with transfer learning | |
Tu et al. | Consumer needs for hand-touch product designs based on the experience economy | |
Zhang et al. | Explorations of charm factors and development of fishing in southern taiwan based on miryoku engineering and the analytic network process | |
Wang et al. | The proper motor control model revealed by wheelchair curling quantification of elite athletes | |
Singh et al. | Assessment of the heart disease using soft computing methodology | |
CN109637618A (en) | A kind of Chinese medicinal formulae diversity recommended method based on label | |
Turoń-Skrzypińska et al. | Assessment of the relationship between selected factors and stress-coping strategies in handcyclists—a preliminary study | |
Monteiro et al. | Bodyweight and combined training reduce chronic low-grade inflammation and improve functional fitness of postmenopausal women | |
Yang et al. | Emotion variation from controlling contrast of visual contents through EEG-Based deep emotion recognition | |
Nam et al. | Effects of SNS social capital on E-service quality and sustained referral intentions of E-fitness apparel: Comparative body image satisfaction analysis | |
Farina | Systems precision medicine: putting the pieces back together | |
Shi et al. | Sparse logistic regression-based EEG channel optimization algorithm for improved universality across participants |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180619 |