CN106651517A - Hidden semi-Markov model-based drug recommendation method - Google Patents
Hidden semi-Markov model-based drug recommendation method Download PDFInfo
- Publication number
- CN106651517A CN106651517A CN201611184351.8A CN201611184351A CN106651517A CN 106651517 A CN106651517 A CN 106651517A CN 201611184351 A CN201611184351 A CN 201611184351A CN 106651517 A CN106651517 A CN 106651517A
- Authority
- CN
- China
- Prior art keywords
- user
- state
- medicine
- markov model
- hidden semi
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Abstract
The invention relates to a hidden semi-Markov model-based drug recommendation method. The method is characterized by comprising the following steps of 1, preprocessing training data, and generating a training data set of a user behavior sequence; 2, estimating parameters of a drug recommendation model; 3, acquiring a network access behavior sequence of a user on a medical platform; 4, taking the network access behavior sequence of the user as an observation value, and deducing a state sequence of the user by using the trained drug recommendation model; 5, calculating expected durations of all states of the state sequence; 6, sorting the obtained expected durations of all the states according to a descending order to obtain first multiple states most concerned by the user; and 7, recommending corresponding drugs to the user according to first multiple disease conditions most concerned by the user. According to the method, the disease conditions concerned by the user are accurately predicted by online behaviors of the user on a cloud platform, and related drugs are recommended to the user according to the disease conditions most concerned by the user, so that the correlation of drug recommendation results is improved.
Description
Technical field
The present invention relates to a kind of medicine based on hidden semi-Markov model recommends method.
Background technology
It is the full industrial chain big data money of medicine based on the medicine polymerization supply chains platform of cloud computing and big data technology
Source and public service cloud platform, merge with big data and storage, platform big data are excavated and application, Drug Administration and industry letter
The function such as integrated is ceased, the resource of medicine supply chain enterprise is incorporated, is conducive to the specification medicine online transaction order of the market economy, promoted
Enter the full industrial chain of medicine to develop healthy, benignly.How to be user under the fast-developing situation of medical big data application service
There is provided accurate medical service becomes the key issue of each big medical platform urgent need to resolve, and existing solution generally there are recommendation
The defect such as correlation as a result is not accurate enough.
The content of the invention
The present invention is directed to the deficiencies in the prior art, there is provided a kind of medicine based on hidden semi-Markov model recommends method.
Network behavior sequence of the method first according to observable user in cloud platform, predicts the state of an illness of user's concern, further according to
The state of an illness that user most pays close attention to user recommends related medicine；It solves user's medicine essence on medicine polymerization supply chains platform
The problem that standard is recommended.
In order to achieve the above object, a kind of medicine based on hidden semi-Markov model of the present invention recommends method, main bag
Include following steps：
The first step, training data pretreatment, i.e., carry out data cleansing to user in the internet behavior data of medical platform, raw
Into the training dataset of user behavior sequence；
Second step, the parameter to medicine recommended models are estimated；
3rd step, collection user medical platform internet behavior sequence；
4th step, with the internet behavior sequence of user as observation, infer user using the medicine recommended models that train
Status switch；
The expected duration of the 5th step, each state of calculating status switch；
6th step, the expected duration of each state of gained is sorted in descending order, obtain the front plural number that user most pays close attention to
The front plural number that individual state, i.e. user are most paid close attention to plants the state of an illness；
7th step, the front plural number kind state of an illness most paid close attention to according to user, to user corresponding medicine is recommended.
Preferably, the medicine recommended models are based on the model of hidden semi-Markov model.
Preferably, the parameter model of the medicine recommended models is expressed as：θ={ π, A, B }；Wherein, π is initial model
Initial state probabilities, A is state transition probability, and B is observation probability.
Preferably, the method that the parameter to medicine recommended models is estimated is forward-backward algorithm algorithm.
Preferably, the 4th step infers the estimating method of the status switch of user using the medicine recommended models for training
It is based on Viterbi algorithm.
So-called observation space, is online behavior sequence of the user on medicine polymerization supply chains platform, is expressed as x
=x1,x2,...,xT, including the page that browse, visit of the user on APP, the medical system such as cloud platform and robot or platform
The resource asked or the problem of proposition etc..The valued space of so-called state, is the state of an illness of user's concern, is expressed as y=y1,y2,
...yn。
Online behavior of the present invention by user in cloud platform, the state of an illness of Accurate Prediction user concern, further according to user most
The state of an illness of concern to user recommends related medicine, so as to improve the correlation of medicine recommendation results.
Description of the drawings
Fig. 1 is the schematic flow sheet that medicine of the present invention based on hidden semi-Markov model recommends method.
Specific embodiment
Describe the present invention below in conjunction with the drawings and specific embodiments, but it is not as a limitation of the invention.
So-called observation space, is online behavior sequence of the user on medicine polymerization supply chains platform, is expressed as x
=x1,x2,...,xT, including the page that browse, visit of the user on APP, the medical system such as cloud platform and robot or platform
The resource asked or the problem of proposition etc..The valued space of so-called state, is the state of an illness of user's concern, is expressed as y=y1,y2,
...yn。
The parameter model of the medicine recommended models is expressed as：θ={ π, A, B }；Wherein, π is the initial shape of initial model
State probability, A is state transition probability, and B is observation probability.
For convenience descriptive model, of the invention to adopt following presentation symbol：
1)t:T+d is represented from t and is started the time series until t+d, i.e. t, t+1 ..., t+d.
2)S[t-d+1:t]=j represents that the state on [t-d+1, t] time interval is j, and the state of t+1 and t-1 is not j.
3)St]=j table times t and its state before are j, and the state of t+1 is not j.
4)S[t=j represents time t and its state afterwards is j, and the state of t-1 is not j.
The parameter Estimation task of medicine recommended models be estimated by the online behavior sequence of the user for collecting it is corresponding hidden
The parameter of semi-Markov model.The present invention solves the Parameter Estimation Problem of medical recommended models, tool using forward-backward algorithm algorithm
Body is as described below.
1) forward-backward algorithm variable is defined：
αt(j, d)=P [St-d+1:t=j, o1:t|θ]；
βt(j, d)=P [ot+1:T|St-d+1:t=j, θ].
2) initialization of forward-backward algorithm algorithm：
α1(j, d)=πj,
βT(j, d)=1.
3) iteration derivation：
4) intermediate variable is calculated：
ηt(j, d)=P [S[t-d+1:t]=j, o1:T| λ]=αt(j,d)βt(j,d)；
ξt(i,d'；J, d)=P [St]=i, S[t+1:t+d]=j, o1:T| λ]=αt(i,d')a(i,d')(j,d)bj,d(ot+1:t+d)
βt+d(j,d)；
5) parameter more new formula
Wherein, whenWhen,Otherwise
The behavior sequence of given user, is to extract preferable user interest sequences y=y based on Viterbi algorithm1,y2,
...yt。
The expected duration of calculating state i is： User can be shown in the total of certain state
Expected time, you can to reflect the degree that user pays close attention to certain state of an illness.Usually, user is high to the degree of concern of certain state of an illness
The expected duration length of the low state to corresponding to the state of an illness is directly proportional.
With reference to Fig. 1, a kind of medicine based on hidden semi-Markov model of the embodiment of the present invention recommends method, mainly include with
Lower step：
The first step, training data pretreatment, i.e., carry out data cleansing to user in the internet behavior data of medical platform, raw
Into the training dataset of user behavior sequence；
Second step, the parameter to medicine recommended models are estimated；
3rd step, collection user medical platform internet behavior sequence；
4th step, with the internet behavior sequence of user as observation, infer user using the medicine recommended models that train
Status switch；
The expected duration of the 5th step, each state of calculating status switch
6th step, the expected duration of each state of gained is sorted in descending order, obtain the top n shape that user most pays close attention to
The front N kinds state of an illness that state, i.e. user are most paid close attention to；
7th step, the front plural number kind state of an illness most paid close attention to according to user, to user corresponding medicine is recommended.
The medicine recommended models are based on the model of hidden semi-Markov model.The parameter to medicine recommended models
The method estimated is forward-backward algorithm algorithm.4th step infers the state of user using the medicine recommended models for training
The estimating method of sequence is based on Viterbi algorithm.
Online behavior of the embodiment of the present invention by user in cloud platform, the state of an illness of Accurate Prediction user concern, further according to
The state of an illness that user most pays close attention to user recommends related medicine, so as to improve the correlation of medicine recommendation results.
Below the present invention has been described in detail, but it will be apparent that those skilled in the art can carry out various changing
Become and improve, without departing from the scope of the present invention that appended claims are limited.
Claims (5)
1. a kind of medicine based on hidden semi-Markov model recommends method, it is characterised in that mainly include the following steps that：
The first step, training data pretreatment, i.e., carry out data cleansing to user in the internet behavior data of medical platform, generates and uses
The training dataset of family behavior sequence；
Second step, the parameter to medicine recommended models are estimated；
3rd step, collection user medical platform internet behavior sequence；
4th step, with the internet behavior sequence of user as observation, infer the shape of user using the medicine recommended models that train
State sequence；
The expected duration of the 5th step, each state of calculating status switch；
6th step, the expected duration of each state of gained is sorted in descending order, obtain front a plurality of shapes that user most pays close attention to
The front plural number that state, i.e. user are most paid close attention to plants the state of an illness；
7th step, the front plural number kind state of an illness most paid close attention to according to user, to user corresponding medicine is recommended.
2. a kind of medicine based on hidden semi-Markov model according to claim 1 recommends method, it is characterised in that institute
It is based on the model of hidden semi-Markov model to state medicine recommended models.
3. a kind of medicine based on hidden semi-Markov model according to claim 1 recommends method, it is characterised in that institute
The parameter model for stating medicine recommended models is expressed as：θ={ π, A, B }；Wherein, π is the initial state probabilities of initial model, and A is
State transition probability, B is observation probability.
4. a kind of medicine based on hidden semi-Markov model according to claim 1 recommends method, it is characterised in that institute
It is forward-backward algorithm algorithm to state the method estimated the parameter of medicine recommended models.
5. a kind of medicine based on hidden semi-Markov model according to claim 1 recommends method, it is characterised in that institute
State the 4th step and infer that the estimating method of the status switch of user is to calculate based on Viterbi using the medicine recommended models for training
Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611184351.8A CN106651517B (en) | 2016-12-20 | 2016-12-20 | Drug recommendation method based on hidden semi-Markov model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611184351.8A CN106651517B (en) | 2016-12-20 | 2016-12-20 | Drug recommendation method based on hidden semi-Markov model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106651517A true CN106651517A (en) | 2017-05-10 |
| CN106651517B CN106651517B (en) | 2021-11-30 |
Family
ID=58834723
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611184351.8A Active CN106651517B (en) | 2016-12-20 | 2016-12-20 | Drug recommendation method based on hidden semi-Markov model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106651517B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110085292A (en) * | 2019-04-28 | 2019-08-02 | 广东技术师范大学 | Drug recommended method, device and computer readable storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102270212A (en) * | 2011-04-07 | 2011-12-07 | 浙江工商大学 | User interest feature extraction method based on hidden semi-Markov model |
| CN103345503A (en) * | 2013-07-01 | 2013-10-09 | 杭州万事利丝绸科技有限公司 | Silk product individualized recommendation method based on wavelet network |
| JP2014066949A (en) * | 2012-09-27 | 2014-04-17 | Univ Of Tokyo | Optimum route searching device |
| CN104008203A (en) * | 2014-06-17 | 2014-08-27 | 浙江工商大学 | User interest discovering method with ontology situation blended in |
| CN104854654A (en) * | 2012-10-18 | 2015-08-19 | 谷歌公司 | Methods and systems for speech recognition processing using search query information |
| US20160063993A1 (en) * | 2014-09-02 | 2016-03-03 | Microsoft Corporation | Facet recommendations from sentiment-bearing content |
-
2016
- 2016-12-20 CN CN201611184351.8A patent/CN106651517B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102270212A (en) * | 2011-04-07 | 2011-12-07 | 浙江工商大学 | User interest feature extraction method based on hidden semi-Markov model |
| JP2014066949A (en) * | 2012-09-27 | 2014-04-17 | Univ Of Tokyo | Optimum route searching device |
| CN104854654A (en) * | 2012-10-18 | 2015-08-19 | 谷歌公司 | Methods and systems for speech recognition processing using search query information |
| CN103345503A (en) * | 2013-07-01 | 2013-10-09 | 杭州万事利丝绸科技有限公司 | Silk product individualized recommendation method based on wavelet network |
| CN104008203A (en) * | 2014-06-17 | 2014-08-27 | 浙江工商大学 | User interest discovering method with ontology situation blended in |
| US20160063993A1 (en) * | 2014-09-02 | 2016-03-03 | Microsoft Corporation | Facet recommendations from sentiment-bearing content |
Non-Patent Citations (2)
| Title |
|---|
| 蒋冬梅 等: "《考虑状态持续时间的改进Viterbi算法及语音识别》", 《西北工业大学学报》 * |
| 血影雪梦: "《隐马尔可夫(HMM)、前/后向算法、Viterbi算法再次总结》", 《HTTPS://BLOG.CSDN.NET/XUEYINGXUE001/ARTICLE/DETAILS/52396494》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110085292A (en) * | 2019-04-28 | 2019-08-02 | 广东技术师范大学 | Drug recommended method, device and computer readable storage medium |
| CN110085292B (en) * | 2019-04-28 | 2022-07-26 | 广东技术师范大学 | Medicine recommendation method and device and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106651517B (en) | 2021-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ambekar et al. | Disease risk prediction by using convolutional neural network | |
| CN105574098B (en) | The generation method and device of knowledge mapping, entity control methods and device | |
| Brodu et al. | Exploring two novel features for EEG-based brain–computer interfaces: Multifractal cumulants and predictive complexity | |
| Gharehchopogh et al. | Neural network application in diagnosis of patient: a case study | |
| AU2022204257A1 (en) | Improved docket search and analytics engine | |
| Lafta et al. | An intelligent recommender system based on short-term risk prediction for heart disease patients | |
| CN106529110A (en) | Classification method and equipment of user data | |
| CN110706822B (en) | Health management method based on logistic regression model and decision tree model | |
| CN108231146A (en) | A kind of medical records model building method, system and device based on deep learning | |
| Davin et al. | Separating homophily and peer influence with latent space | |
| Reddy et al. | Prediction of heart disease using decision tree approach | |
| Naeem et al. | Recognising activities of daily life using hierarchical plans | |
| CN114420279A (en) | Medical resource recommendation method, device, equipment and storage medium | |
| CN109979592A (en) | Mental health method for early warning, user terminal, server and system | |
| CN103823667B (en) | Method and system according to visible feedback adjust automatically value sequence analysis task | |
| CN106651517A (en) | Hidden semi-Markov model-based drug recommendation method | |
| Duflou et al. | Systematic innovation through patent based product aspect analysis | |
| Pentney et al. | Learning large scale common sense models of everyday life | |
| Saiyed et al. | A survey on naive bayes based prediction of heart disease using risk factors | |
| Liu et al. | A learning-based system for predicting sport injuries | |
| CN115982460A (en) | Personalized recommendation method, system and medium for health science popularization information | |
| Yasodha et al. | Comparative study of diabetic patient data’s using classification algorithm in WEKA tool | |
| Kant et al. | Review of heart disease prediction using data mining classifications | |
| Lee et al. | Bayesian Nonparametric Collaborative Topic Poisson Factorization for Electronic Health Records-Based Phenotyping. | |
| CN110473636B (en) | Intelligent medical advice recommendation method and system based on deep 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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 510665 No. 293 Shipai Zhongshan Avenue, Tianhe District, Guangzhou City, Guangdong Province Applicant after: GUANGDONG POLYTECHNIC NORMAL University Address before: 510665 No. 293 Shipai Zhongshan Avenue, Tianhe District, Guangzhou City, Guangdong Province Applicant before: GUANGDONG POLYTECHNIC NORMAL University |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |