CN106803012B - Prescription function prediction method based on probability topic model and Chinese medicine essential attribute - Google Patents
Prescription function prediction method based on probability topic model and Chinese medicine essential attribute Download PDFInfo
- Publication number
- CN106803012B CN106803012B CN201611244641.7A CN201611244641A CN106803012B CN 106803012 B CN106803012 B CN 106803012B CN 201611244641 A CN201611244641 A CN 201611244641A CN 106803012 B CN106803012 B CN 106803012B
- Authority
- CN
- China
- Prior art keywords
- prescription
- chinese medicine
- effect
- drug
- probability
- 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
- 239000003814 drug Substances 0.000 title claims abstract description 248
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000000694 effects Effects 0.000 claims abstract description 97
- 229940079593 drug Drugs 0.000 claims abstract description 82
- 239000013598 vector Substances 0.000 claims abstract description 47
- 241000721047 Danaus plexippus Species 0.000 claims abstract description 25
- 238000005192 partition Methods 0.000 claims abstract description 7
- 235000019640 taste Nutrition 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 13
- 239000000796 flavoring agent Substances 0.000 claims description 11
- 235000019634 flavors Nutrition 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 239000003795 chemical substances by application Substances 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 6
- 229940126678 chinese medicines Drugs 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 5
- 238000013139 quantization Methods 0.000 claims description 5
- 230000010415 tropism Effects 0.000 claims description 5
- 206010007247 Carbuncle Diseases 0.000 claims description 3
- 241000607479 Yersinia pestis Species 0.000 claims description 3
- 235000009508 confectionery Nutrition 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims description 3
- 210000000232 gallbladder Anatomy 0.000 claims description 3
- 210000002216 heart Anatomy 0.000 claims description 3
- 210000003734 kidney Anatomy 0.000 claims description 3
- 210000002429 large intestine Anatomy 0.000 claims description 3
- 239000008141 laxative Substances 0.000 claims description 3
- 210000004185 liver Anatomy 0.000 claims description 3
- 208000030208 low-grade fever Diseases 0.000 claims description 3
- 210000004072 lung Anatomy 0.000 claims description 3
- 210000003516 pericardium Anatomy 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000001543 purgative effect Effects 0.000 claims description 3
- 238000011002 quantification Methods 0.000 claims description 3
- 239000000932 sedative agent Substances 0.000 claims description 3
- 230000001624 sedative effect Effects 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 210000000813 small intestine Anatomy 0.000 claims description 3
- 210000000952 spleen Anatomy 0.000 claims description 3
- 210000002784 stomach Anatomy 0.000 claims description 3
- 210000003932 urinary bladder Anatomy 0.000 claims description 3
- 238000010792 warming Methods 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 claims description 2
- 206010010774 Constipation Diseases 0.000 claims 1
- 230000008901 benefit Effects 0.000 claims 1
- 230000008859 change Effects 0.000 claims 1
- 201000006549 dyspepsia Diseases 0.000 claims 1
- 239000003172 expectorant agent Substances 0.000 claims 1
- 230000003419 expectorant effect Effects 0.000 claims 1
- 238000003556 assay Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 206010062717 Increased upper airway secretion Diseases 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 229910052602 gypsum Inorganic materials 0.000 description 2
- 239000010440 gypsum Substances 0.000 description 2
- 241000411851 herbal medicine Species 0.000 description 2
- 208000026435 phlegm Diseases 0.000 description 2
- 231100000614 poison Toxicity 0.000 description 2
- 238000000547 structure data Methods 0.000 description 2
- 239000003440 toxic substance Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 241001465754 Metazoa Species 0.000 description 1
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000002552 dosage form Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
Classifications
-
- G06F19/34—
Abstract
The present invention discloses a kind of Prescription Effect prediction technique based on probability topic model and Chinese medicine essential attribute.Structural data is extracted from prescription voluminous dictionary database and TCM Databases by traditional Chinese medicine and pharmacy language Words partition system first.Then, according to prescription Chinese medicine corresponding with its, the probability attribute vector that relationship is treated between Prescription Effect and Chinese medicine is obtained using LDA model.According to the Chinese medicine and the dosage of drug of every width prescription, monarch drug in a prescription and ministerial drug are extracted, integrates the attribute vector of monarch drug in a prescription and ministerial drug, and the prescription weight vectors based on TFIDF model on its basis, constitutes the feature vector of prescription, input SVM multi-categorizer is trained.Finally, user inputs the new prescription information for needing to predict, the effect of obtaining prescription with trained SVM multi-categorizer is modeled using theme.The effect of present invention can effectively predict the prescription according to the drug matching of new prescription, it is significant for the pre-clinical assay of automatic prescription, prescription.
Description
Technical field
The present invention relates to text minings and traditional Chinese medicine field of information processing.Probability topic model is based on more particularly, to one kind
With the prescription function prediction method of Chinese medicine essential attribute.
Background technique
Chinese medicine is the medicine based on traditional medicine of Created in China.The accumulation of more than one thousand years so that Chinese medicine to have accumulated content rich
Rich medicine ancient books and records and record, these resources have contained a large amount of unknown knowledge.With the development of digital technology, traditional Chinese medicine number
Word resource is more and more huger, and data mining technology is also applied to traditional Chinese medicine to realize specific purpose by more and more scholars
Analysis and law discovery.A research branch of the pharmacology of traditional Chinese medical formulae as traditional Chinese medicine and pharmacy, needs to select appropriate medicine according to the principles of formulating prescriptions
Object reasonable compatibility decides suitable dosage, dosage form and usage.Wherein, the effect of prescription, generally requires by excessively very long multiple
Miscellaneous animal or clinical trial, but clinical or zoopery expends a large amount of human and material resources and time.If passing through trusted computer
Breath digging technology predicts the prescription newly formed, obtains the partial efficacy information of the prescription, so that it may extensive to carry out
Clinical or zoopery provides extremely valuable reference, greatly promotes clinical conventional efficient.In consideration of it, inventor
Focus is that the effect of how passing through computerized information digging technology, realize prescription using existing traditional Chinese medicine digital resource is pre-
It surveys, to provide valuable clinical evidence for prescription research.
Summary of the invention
The purpose of the present invention is overcome the deficiencies of the prior art and provide one kind to belong to substantially based on probability topic model and Chinese medicine
The prescription function prediction method of property.
The technical solution adopted by the present invention to solve the technical problems the following steps are included:
Step 1, data prediction
For prescription voluminous dictionary database, word segmentation processing is carried out to prescription information using traditional Chinese medicine and pharmacy language Words partition system, is mentioned
Prescription name, Prescription Effect, the corresponding Chinese medicine of prescription, the dosage of drug and dosage unit are taken out, is unitized to dosage unit, it is right
The dosage of drug in prescription is standardized；For TCM Databases, the effect of using traditional Chinese medicine and pharmacy Words partition system to Chinese medicine, property
Taste and three large attribute of channel tropism are segmented, and stop words is removed, and carry out 0-1 quantification treatment to the structured attributes data extracted,
The attribute vector of every taste Chinese medicine is obtained, is stored in database, the prescription voluminous dictionary database and TCM Databases are half hitch
Structure data.
Step 2, according to prescription name and the corresponding Chinese medicine of prescription, " prescription-Chinese medicine " matrix is obtained, using Prescription Effect as hidden
Containing theme, theme modeling is carried out to " prescription-Chinese medicine " matrix using LDA probability topic model, combines compatibility data in a model
Library obtains the probability attribute vector that relationship is treated between Prescription Effect and Chinese medicine, is stored in database.
Step 3, according to the Chinese medicine list of every width prescription and it is unitized after the dosage of drug, calculate frequency of the Chinese medicine in prescription
It is secondary, take the highest two tastes Chinese medicine of the frequency as monarch drug in a prescription and ministerial drug, and existed according to the frequency of Chinese medicine using TFIDF model extraction Chinese medicine
Weight coefficient in prescription constructs the weight vectors of prescription, is stored in database.
Step 4, according to the principles of formulating prescriptions in Chinese medicine traditional theory, the Chinese medicine in prescription is divided into monarch, and prescription
The effect of mainly determined by monarch drug in a prescription and ministerial drug.For every secondary prescription in prescription voluminous dictionary database, merge the attribute of monarch drug in a prescription to
The weight vectors of amount, the attribute vector of ministerial drug and prescription, constitute the feature vector of prescription, input SVM classifier to model into
Row training, passes through the multi-class classifier of one-versus-rest method construct.
The new prescription information that step 5, input need to predict, new prescription information includes Chinese medicine name and corresponding dosage；According to
The probability attribute vector of relationship between Prescription Effect and Chinese medicine, calculates the prescription using Bayesian formula and is under the jurisdiction of some effect
Probability value is sorted from large to small according to probability value, the effect being chosen within the scope of specific threshold, obtains effect collection unification U1(U1Have
It may be empty set).
Step 6 obtains the feature vector of prescription to be predicted according to step 4, and it is pre- that input SVM multi-categorizer carries out Prescription Effect
It surveys, obtains two U of effect set2。
Step 7 is unified effect collection and effect set two makees union operation, and final Prescription Effect set U=U is obtained1
∪U2。
The specific implementation content of the step 1 includes:
1. by the various dose unit of Chinese medicine according to following rules equivalent Cheng Ke:
One liang=31.25 grams
One money=3.125 gram
One point=0.3125 gram
2. continuing to standardize to the dosage of Chinese medicines different in prescription according to the following formula:
Wherein: diIndicate the dosage of certain Chinese medicine in prescription,The dosage of drug after representing standardization, dmaxIndicate certain Chinese medicine
Maximum value in common dose；dminIndicate the minimum value in certain Chinese medicine common dose；
3. the Prescription Effect include harmonizing prescription, astringent prescription, sedative drugs prescriptions, resuscitative prescription, dryness-moistening prescription, wind-curing prescription, purgative drugs prescriptions,
Clearing food stagnation prescriptions, exterior-interior relieving prescription, heat-clearing formula, warming interior formula, qi-regulating prescription, blood-regulating prescription, prescriptions for boil and carbuncle, summer-heat clearing prescription, dampclearing prescription, eliminating the phlegm
Agent, tonifying recipes, diaphoretic recipes, pest repellant.
4. the Chinese herbal medicine nature and flavor refer to the property and smell of drug, i.e. four natures and five flavors of drug, including cold, hot, warm, cool, pungent,
It is sweet, sour, bitter, salty.Channel tropism using 12 zang-fu differentiation methods state, return heart, liver, spleen, lung, kidney, stomach, large intestine, small intestine, bladder, gallbladder,
Pericardium, tri-jiao channel.Effect is described as unit of two-character word, such as " heat-clearing ", " removing toxic substances ", " cool blood ".Wherein, occur in nature and flavor
" slightly cold ", the statement such as " low-grade fever ", with 0.5 quantization, remaining is using 0-1 quantization construction vector space.
The step 2 the specific implementation process is as follows:
2-1. sets hyper parameter α=2.5, β=0.1；
2-2. carries out parameter Estimation, the number of iterations 100 to LDA probability topic model using the Gibbs method of sampling；
After 2-3. completes sampling, drug matching database is inquired:
If Chinese medicine hiDo not have relevant compatibility drug, calculate the probability matrix ψ of " Chinese medicine-effect " according to the following formula:
If Chinese medicine hiAnd hjFor compatibility drug pair, the probability matrix θ of " the p- effect of medicine " is calculated:
Wherein, W represents the sum of Chinese medicine,Indicate Chinese medicine hiThe number of effect k is distributed to, does not include currently distributing；The Chinese medicine total degree for distributing to effect k is represented, does not include Chinese medicine hiDistribute to the number of effect k；Representative is distributed to
The Chinese medicine total degree of effect k does not include Chinese medicine hi、hjDistribute to the number of effect k.
The step 3 the specific implementation process is as follows:
3-1. is directed to every width prescription p, it is assumed that taste of traditional Chinese medicine number is Np, calculate Chinese medicine hiFrequency F (h in prescriptioni), take frequency
Secondary highest two tastes Chinese medicine is as monarch drug in a prescription and ministerial drug.
3-2. calculates Chinese medicine h using TFIDF modeliTo the significance level of prescription, according to every taste Chinese medicine in prescription
The weight vectors of TFIDF value construction prescription pM represents unique Chinese medicine sum in prescription database.
Wherein, if prescription p includes Chinese medicine hi, then ti=TFIDF (hi), otherwise ti=0.
The step 5 the specific implementation process is as follows:
5-1. user inputs prescription H to be predicted(p)={ h1,h2,...hNp, hiChinese medicine is represented, is calculated using Bayesian formula
The prescription is under the jurisdiction of the probability of effect k:
Wherein, xi=1 indicates Chinese medicine hiThere are compatibility drug pair, otherwise xi=0.
5-2. is directed to prescription H(p), the effect of meeting following formula k is returned, threshold value T=1e-8 is taken, obtains the unification of effect collection:
p(k|H(p)) > T.
The present invention has the beneficial effect that:
The present invention on the basis of existing traditional Chinese medicine digital resource, by probability topic model excavate Prescription Effect and
Treatment relationship between Chinese medicine, the function of prescription is effectively calculated by Bayesian formula.In addition, existing in conjunction with monarch drug in a prescription and ministerial drug
The main function played in Prescription Effect calculates the monarch automatically derived in prescription, ministerial drug by the frequency of the Chinese medicine in prescription, and
Using the weight vectors of TFIDF model extraction prescription, merge the power of the attribute vector of monarch drug in a prescription, the attribute vector of ministerial drug and prescription
Feature vector of the weight vector as prescription, further improves the accuracy of Prescription Effect prediction using SVM multi-categorizer, thus
Valuable clinical evidence is provided for prescription research.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings.
As shown in Figure 1, the prescription function prediction method based on probability topic model and Chinese medicine essential attribute includes following step
It is rapid:
Step 1, data prediction
For prescription voluminous dictionary database, word segmentation processing is carried out to prescription information using traditional Chinese medicine and pharmacy language Words partition system, is mentioned
Prescription name, Prescription Effect, the corresponding Chinese medicine of prescription, the dosage of drug and dosage unit are taken out, is unitized to dosage unit, it is right
The dosage of drug in prescription is standardized；For TCM Databases, the effect of using traditional Chinese medicine and pharmacy Words partition system to Chinese medicine, property
Taste and three large attribute of channel tropism are segmented, and stop words is removed, and carry out 0-1 quantification treatment to the structured attributes data extracted,
The attribute vector of every taste Chinese medicine is obtained, is stored in database, the prescription voluminous dictionary database and TCM Databases are half hitch
Structure data.
Step 2, according to prescription name and the corresponding Chinese medicine of prescription, " prescription-Chinese medicine " matrix is obtained, using Prescription Effect as hidden
Containing theme, theme modeling is carried out to " prescription-Chinese medicine " matrix using LDA probability topic model, combines compatibility data in a model
Library obtains the probability attribute vector that relationship is treated between Prescription Effect and Chinese medicine, is stored in database.
Step 3, according to the Chinese medicine list of every width prescription and it is unitized after the dosage of drug, calculate frequency of the Chinese medicine in prescription
It is secondary, take the highest two tastes Chinese medicine of the frequency as monarch drug in a prescription and ministerial drug, and existed according to the frequency of Chinese medicine using TFIDF model extraction Chinese medicine
Weight coefficient in prescription constructs the weight vectors of prescription, is stored in database.
Step 4, according to the principles of formulating prescriptions in Chinese medicine traditional theory, the Chinese medicine in prescription is divided into monarch, and prescription
The effect of mainly determined by monarch drug in a prescription and ministerial drug.For every secondary prescription in prescription voluminous dictionary database, merge the attribute of monarch drug in a prescription to
The weight vectors of amount, the attribute vector of ministerial drug and prescription, constitute the feature vector of prescription, input SVM classifier to model into
Row training, passes through the multi-class classifier of one-versus-rest method construct.
The new prescription information that step 5, input need to predict, new prescription information includes Chinese medicine name and corresponding dosage；According to
The probability attribute vector of relationship between Prescription Effect and Chinese medicine, calculates the prescription using Bayesian formula and is under the jurisdiction of some effect
Probability value is sorted from large to small according to probability value, the effect being chosen within the scope of specific threshold, obtains effect collection unification U1(U1Have
It may be empty set).
Step 6 obtains the feature vector of prescription to be predicted according to step 4, and it is pre- that input SVM multi-categorizer carries out Prescription Effect
It surveys, obtains two U of effect set2。
Step 7 is unified effect collection and effect set two makees union operation, and final Prescription Effect set U=U is obtained1
∪U2。
The specific implementation content of the step 1 includes:
1. by the various dose unit of Chinese medicine according to following rules equivalent Cheng Ke:
One liang=31.25 grams
One money=3.125 gram
One point=0.3125 gram
2. continuing to standardize to the dosage of Chinese medicines different in prescription according to the following formula:
Wherein: diIndicate the dosage of certain Chinese medicine in prescription,The dosage of drug after representing standardization, dmaxIndicate certain Chinese medicine
Maximum value in common dose；dminIndicate the minimum value in certain Chinese medicine common dose.
3. the Prescription Effect include harmonizing prescription, astringent prescription, sedative drugs prescriptions, resuscitative prescription, dryness-moistening prescription, wind-curing prescription, purgative drugs prescriptions,
Clearing food stagnation prescriptions, exterior-interior relieving prescription, heat-clearing formula, warming interior formula, qi-regulating prescription, blood-regulating prescription, prescriptions for boil and carbuncle, summer-heat clearing prescription, dampclearing prescription, eliminating the phlegm
Agent, tonifying recipes, diaphoretic recipes, pest repellant.
4. the Chinese herbal medicine nature and flavor refer to the property and smell of drug, i.e. four natures and five flavors of drug, including cold, hot, warm, cool, pungent,
It is sweet, sour, bitter, salty.Channel tropism using 12 zang-fu differentiation methods state, return heart, liver, spleen, lung, kidney, stomach, large intestine, small intestine, bladder, gallbladder,
Pericardium, tri-jiao channel.Effect is described as unit of two-character word, such as " heat-clearing ", " removing toxic substances ", " cool blood ".Wherein, occur in nature and flavor
" slightly cold ", the statement such as " low-grade fever ", with 0.5 quantization, remaining is using 0-1 quantization construction vector space.
The step 2 the specific implementation process is as follows:
2-1. sets hyper parameter α=2.5, β=0.1；
2-2. carries out parameter Estimation, the number of iterations 100 to LDA probability topic model using the Gibbs method of sampling；
After 2-3. completes sampling, drug matching database is inquired:
If Chinese medicine hiDo not have relevant compatibility drug, calculate the probability matrix ψ of " Chinese medicine-effect " according to the following formula:
If Chinese medicine hiAnd hjFor compatibility drug pair, the probability matrix θ of " the p- effect of medicine " is calculated:
Wherein, W represents the sum of Chinese medicine,Indicate Chinese medicine hiThe number of effect k is distributed to, does not include currently distributing；The Chinese medicine total degree for distributing to effect k is represented, does not include Chinese medicine hiDistribute to the number of effect k.Representative is distributed to
The Chinese medicine total degree of effect k does not include Chinese medicine hi、hjDistribute to the number of effect k.
The step 3 the specific implementation process is as follows:
3-1. is directed to every width prescription p, it is assumed that taste of traditional Chinese medicine number is Np, calculate Chinese medicine hiFrequency F (h in prescriptioni), take frequency
Secondary highest two tastes Chinese medicine is as monarch drug in a prescription and ministerial drug.
3-2. calculates Chinese medicine h using TFIDF modeliTo the significance level of prescription, according to every taste Chinese medicine in prescription
The weight vectors of TFIDF value construction prescription pM represents unique Chinese medicine sum in prescription database.
Wherein, if prescription p includes Chinese medicine hi, then ti=TFIDF (hi), otherwise ti=0.
The step 5 the specific implementation process is as follows:
5-1. user inputs prescription H to be predicted(p)={ h1,h2,...hNp, hiChinese medicine is represented, is calculated using Bayesian formula
The prescription is under the jurisdiction of the probability of effect k:
Wherein, xi=1 indicates Chinese medicine hiThere are compatibility drug pair, otherwise xi=0.
5-2. is directed to prescription H(p), the effect of meeting following formula k is returned, threshold value T=1e-8 is taken, obtains the unification of effect collection:
p(k|H(p)) > T.
Embodiment
1, a new prescription is inputted, includes each Chinese medicine and corresponding dosage.
2, it according to the probability attribute vector data library of relationship between Prescription Effect and Chinese medicine, is calculated using Bayesian formula new
Prescription is under the jurisdiction of the probability value of some effect, is sorted from large to small according to probability value, the effect being chosen within the scope of specific threshold,
Obtain effect collection unification U1。
3, the dosage of drug for the Chinese medicine list in new prescription and after unitizing, calculates the frequency of the Chinese medicine in prescription,
It takes the highest two tastes Chinese medicine of the frequency as monarch drug in a prescription and ministerial drug, and utilizes TFIDF model extraction Chinese medicine in side according to the frequency of Chinese medicine
Weight coefficient in agent constructs the weight vectors of prescription.
4, Chinese medicine properties database is searched, the attribute vector of monarch drug in a prescription and ministerial drug is obtained, merges attribute vector, the ministerial drug of monarch drug in a prescription
Attribute vector and prescription weight vectors, constitute the feature vector of new prescription, input SVM multi-categorizer carries out Prescription Effect
Prediction, obtains two U of effect set2。
5, merge the unification of effect collection and effect set two, obtain the forecasting power of new prescription.
The step 2 is realized by following sub-step:
2.1) the new prescription for assuming input is H(p)={ h1,h2,...hNp, hiChinese medicine is represented, is calculated using Bayesian formula
The prescription is under the jurisdiction of the probability of effect k:
Wherein, xi=1 indicates Chinese medicine hiThere are compatibility drug pair, otherwise xi=0.
2.2) it is directed to new prescription H(p), the effect of meeting following formula k is returned, threshold value T=1e-8 is taken, obtains the unification of effect collection:
p(k|H(p)) > T
The step 3 is realized by following sub-step:
3.1) it is directed to new prescription, it is assumed that taste of traditional Chinese medicine number is Np, calculate Chinese medicine hiFrequency F (h in prescriptioni), take the frequency most
Two high taste Chinese medicines are as monarch drug in a prescription and ministerial drug.
3.2) Chinese medicine h is calculated using TFIDF modeliTo the significance level of new prescription, according to every taste Chinese medicine in new prescription
TFIDF value construct the weight vectors of new prescriptionUnique Chinese medicine that m is represented in prescription database is total
Number.
Wherein, if new prescription includes Chinese medicine hi, then ti=TFIDF (hi), otherwise ti=0.
The new prescription information of the input of this example is gypsum (50g), and rhizoma anemarrhenae (18g), Radix Glycyrrhizae (6g), polished rice (9g) calculates
To monarch drug in a prescription and ministerial drug be respectively gypsum and rhizoma anemarrhenae, effect collection unifies U1={ heat-clearing formula }, two U of effect set2=resuscitative prescription, clearly
Thermit powder }, finally obtained forecasting power is resuscitative prescription and heat-clearing formula.
It should be noted that this embodiment assumes that SVM multi-categorizer is trained finishes.
Claims (5)
1. the prescription function prediction method based on probability topic model and Chinese medicine essential attribute, it is characterised in that including following step
It is rapid:
Step 1, data prediction
For prescription voluminous dictionary database, word segmentation processing is carried out to prescription information using traditional Chinese medicine and pharmacy language Words partition system, is extracted
The corresponding Chinese medicine of prescription name, Prescription Effect, prescription, the dosage of drug and dosage unit, unitize to dosage unit, to prescription
In the dosage of drug be standardized；For TCM Databases, the effect of using traditional Chinese medicine and pharmacy Words partition system to Chinese medicine, nature and flavor and
Three large attribute of channel tropism is segmented, and stop words is removed, and is carried out 0-1 quantification treatment to the structured attributes data extracted, is obtained
The attribute vector of every taste Chinese medicine is stored in database, and the prescription voluminous dictionary database and TCM Databases are semi-structured
Data；
Step 2, according to prescription name and the corresponding Chinese medicine of prescription, " prescription-Chinese medicine " matrix is obtained, using Prescription Effect as implicit master
Topic carries out theme modeling to " prescription-Chinese medicine " matrix using LDA probability topic model, combines compatibility database to obtain in a model
To the probability matrix for treating relationship between Prescription Effect and Chinese medicine, it is stored in database；
Step 3, according to the Chinese medicine list of every width prescription and it is unitized after the dosage of drug, calculate the frequency of the Chinese medicine in prescription,
It takes the highest two tastes Chinese medicine of the frequency as monarch drug in a prescription and ministerial drug, and utilizes TFIDF model extraction Chinese medicine in side according to the frequency of Chinese medicine
Weight coefficient in agent constructs the weight vectors of prescription, is stored in database；
Step 4, according to the principles of formulating prescriptions in Chinese medicine traditional theory, the Chinese medicine in prescription is divided into monarch, and the function of prescription
Effect is mainly determined by monarch drug in a prescription and ministerial drug；For every secondary prescription in prescription voluminous dictionary database, merge attribute vector, the minister of monarch drug in a prescription
The attribute vector of medicine and the weight vectors of prescription, constitute the feature vector of prescription, and input SVM classifier instructs model
Practice, passes through the multi-class classifier of one-versus-rest method construct；
The new prescription information that step 5, input need to predict, new prescription information includes Chinese medicine name and corresponding dosage；According to prescription
The probability attribute vector of relationship between effect and Chinese medicine, calculates the probability that the prescription is under the jurisdiction of some effect using Bayesian formula
Value, sorts from large to small, the effect being chosen within the scope of specific threshold according to probability value, obtains effect collection unification U1；
Step 6 obtains the feature vector of prescription to be predicted according to step 4, and input SVM multi-categorizer carries out Prescription Effect prediction,
Obtain two U of effect set2；
Step 7 is unified effect collection and effect set two makees union operation, and final Prescription Effect set U=U is obtained1∪U2。
2. the prescription function prediction method according to claim 1 based on probability topic model and Chinese medicine essential attribute,
The specific implementation content for being characterized in that the step 1 includes:
1. by the various dose unit of Chinese medicine according to following rules equivalent Cheng Ke:
One liang=31.25 grams
One money=3.125 gram
One point=0.3125 gram
2. continuing to standardize to the dosage of Chinese medicines different in prescription according to the following formula:
Wherein: diIndicate the dosage of certain Chinese medicine in prescription,The dosage of drug after representing standardization, dmaxIndicate that certain Chinese medicine commonly uses agent
Maximum value in amount；dminIndicate the minimum value in certain Chinese medicine common dose；
3. the Prescription Effect includes harmonizing prescription, astringent prescription, sedative drugs prescriptions, resuscitative prescription, dryness-moistening prescription, wind-curing prescription, purgative drugs prescriptions, treatment for relieving indigestion and constipation
Change product agent, exterior-interior relieving prescription, heat-clearing formula, warming interior formula, qi-regulating prescription, blood-regulating prescription, prescriptions for boil and carbuncle, summer-heat clearing prescription, dampclearing prescription, expectorant, benefit
Beneficial agent, diaphoretic recipes, pest repellant；
4. Chinese medicine nature and flavor refer to the property and smell of drug, i.e. four natures and five flavors of drug, including cold, hot, warm, cool, pungent, sweet, sour, bitter, salty；Return
Through being stated using 12 zang-fu differentiation methods, return heart, liver, spleen, lung, kidney, stomach, large intestine, small intestine, bladder, gallbladder, pericardium, tri-jiao channel；Function
Effect is described as unit of two-character word, is quantified for " slightly cold " and " low-grade fever " occurred in nature and flavor with 0.5, remaining two-character word in nature and flavor
Description using 0-1 quantization construction vector space.
3. the prescription function prediction method according to claim 2 based on probability topic model and Chinese medicine essential attribute,
Be characterized in that the step 2 the specific implementation process is as follows:
2-1. sets hyper parameter α=2.5, β=0.1；
2-2. carries out parameter Estimation, the number of iterations 100 to LDA probability topic model using the Gibbs method of sampling；
After 2-3. completes sampling, drug matching database is inquired:
If Chinese medicine hiDo not have relevant compatibility drug, calculate the probability matrix ψ of " Chinese medicine-effect " according to the following formula:
If Chinese medicine hiAnd hjFor compatibility drug pair, the probability matrix θ of " the p- effect of medicine " is calculated:
Wherein, W represents the sum of Chinese medicine,Indicate Chinese medicine hiThe number of effect k is distributed to, does not include currently distributing；Generation
Table distributes to the Chinese medicine total degree of effect k, does not include Chinese medicine hiDistribute to the number of effect k；Effect k is distributed in representative
Chinese medicine total degree, do not include Chinese medicine hi, hjDistribute to the number of effect k.
4. the prescription function prediction method according to claim 3 based on probability topic model and Chinese medicine essential attribute,
Be characterized in that the step 3 the specific implementation process is as follows:
3-1. is directed to every width prescription p, it is assumed that taste of traditional Chinese medicine number is Np, calculate Chinese medicine hiFrequency F (h in prescriptioni), take the frequency most
Two high taste Chinese medicines are as monarch drug in a prescription and ministerial drug；
3-2. calculates Chinese medicine h using TFIDF modeliTo the significance level of prescription, according to TFIDF value of every taste Chinese medicine in prescription
Construct the weight vectors of prescription pM represents unique Chinese medicine sum in prescription database；
Wherein, if prescription p includes Chinese medicine hi, then ti=TFIDF (hi), otherwise ti=0.
5. the prescription function prediction method according to claim 4 based on probability topic model and Chinese medicine essential attribute,
Be characterized in that the step 5 the specific implementation process is as follows:
5-1. user inputs prescription H to be predicted(p)={ h1,h2,...hNp, hiChinese medicine is represented, calculates the party using Bayesian formula
Agent is under the jurisdiction of the probability of effect k:
Wherein, xi=1 indicates Chinese medicine hiThere are compatibility drug pair, otherwise xi=0；
5-2. is directed to prescription H(p), the effect of meeting following formula k is returned, threshold value T=1e-8 is taken, obtains the unification of effect collection:
p(k|H(p)) > T.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611244641.7A CN106803012B (en) | 2016-12-29 | 2016-12-29 | Prescription function prediction method based on probability topic model and Chinese medicine essential attribute |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611244641.7A CN106803012B (en) | 2016-12-29 | 2016-12-29 | Prescription function prediction method based on probability topic model and Chinese medicine essential attribute |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106803012A CN106803012A (en) | 2017-06-06 |
| CN106803012B true CN106803012B (en) | 2019-03-22 |
Family
ID=58985228
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611244641.7A Active CN106803012B (en) | 2016-12-29 | 2016-12-29 | Prescription function prediction method based on probability topic model and Chinese medicine essential attribute |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106803012B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109346180B (en) * | 2018-08-03 | 2022-04-01 | 电子科技大学 | Traditional Chinese medicine prescription monarch, minister, assistant and guide training identification method and system |
| CN109378080A (en) * | 2018-09-14 | 2019-02-22 | 浙江大学 | A kind of similar Chinese medicine search method based on feature bag of words |
| CN109801697A (en) * | 2019-01-16 | 2019-05-24 | 中国中医科学院中医药信息研究所 | A kind of evaluation method of medicine materical crude slice significance level in Chinese medicinal formulae |
| CN109947901B (en) * | 2019-02-20 | 2020-10-20 | 杭州师范大学 | Prescription efficacy prediction method based on multilayer perceptron and natural language processing technology |
| CN110335684A (en) * | 2019-06-14 | 2019-10-15 | 电子科技大学 | The intelligent dialectical aid decision-making method of Chinese medicine based on topic model technology |
| CN110289106B (en) * | 2019-06-28 | 2023-06-16 | 淮阴工学院 | Method for analyzing compatibility relation of traditional Chinese medicines with corresponding efficacy and drug properties in traditional Chinese medicine compound |
| CN110619960B (en) * | 2019-09-10 | 2022-04-22 | 电子科技大学 | Traditional Chinese medicine incompatibility prediction method based on supervised learning framework |
| CN111180045B (en) * | 2019-11-25 | 2023-05-12 | 浙江大学 | Method for mining relation between drug pairs and efficacy from prescription information |
| CN112151140B (en) * | 2020-11-25 | 2021-03-12 | 西藏自治区人民政府驻成都办事处医院 | Antibacterial drug clinical application system, electronic device and computer readable storage medium |
| CN115050481B (en) * | 2022-06-17 | 2023-10-31 | 湖南中医药大学 | Traditional Chinese medicine prescription efficacy prediction method based on graph convolution neural network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101118562A (en) * | 2006-08-21 | 2008-02-06 | 凌强 | Herbalist doctor clinical reference system |
| CN102122325A (en) * | 2011-04-20 | 2011-07-13 | 天津师范大学 | Method for automatically analyzing efficacy of Chinese medicine formula |
| CN103823848A (en) * | 2014-02-11 | 2014-05-28 | 浙江大学 | LDA (latent dirichlet allocation) and VSM (vector space model) based similar Chinese herb literature recommendation method |
-
2016
- 2016-12-29 CN CN201611244641.7A patent/CN106803012B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101118562A (en) * | 2006-08-21 | 2008-02-06 | 凌强 | Herbalist doctor clinical reference system |
| CN102122325A (en) * | 2011-04-20 | 2011-07-13 | 天津师范大学 | Method for automatically analyzing efficacy of Chinese medicine formula |
| CN103823848A (en) * | 2014-02-11 | 2014-05-28 | 浙江大学 | LDA (latent dirichlet allocation) and VSM (vector space model) based similar Chinese herb literature recommendation method |
Non-Patent Citations (2)
| Title |
|---|
| LDA-Based Model for Traditional Chinese Medicine Diagnosis;Qian Tan 等;《3rd International Conference on Computer Science and Network Technology》;20131231;426-429 |
| 李味味，章新友，仵倚，周小玲;基于BP神经网络中药复方功效的预测研究;《中医药导报》;20160831;第22卷(第16期);38-41 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106803012A (en) | 2017-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106803012B (en) | Prescription function prediction method based on probability topic model and Chinese medicine essential attribute | |
| WO2018205739A1 (en) | Traditional chinese medicine knowledge graph and establishment method therefor, and computer system | |
| Jin et al. | An improved association-mining research for exploring Chinese herbal property theory: based on data of the Shennong's Classic of Materia Medica | |
| CN110289106B (en) | Method for analyzing compatibility relation of traditional Chinese medicines with corresponding efficacy and drug properties in traditional Chinese medicine compound | |
| Bu et al. | FangNet: mining herb hidden knowledge from TCM clinical effective formulas using structure network algorithm | |
| Dou et al. | Syndrome differentiation and treatment regularity in traditional Chinese medicine for type 2 diabetes: a text mining analysis | |
| Hu et al. | An analysis model of diagnosis and treatment for COVID-19 pandemic based on medical information fusion | |
| Su et al. | Retrospective study on Professor Zhongying Zhou's experience in Traditional Chinese Medicine treatment on diabetic nephropathy | |
| Yin et al. | Question answering system based on knowledge graph in traditional Chinese medicine diagnosis and treatment of viral hepatitis B | |
| Liu et al. | Application of the data mining algorithm in the clinical guide medical records | |
| Yang et al. | Application of genetic algorithm for discovery of core effective formulae in TCM clinical data | |
| CN110619960B (en) | Traditional Chinese medicine incompatibility prediction method based on supervised learning framework | |
| Zhang et al. | Analysis of prescription medication rules of traditional Chinese medicine for diabetes treatment based on data mining | |
| Xin-Di et al. | Research on herb pairs of classical formulae of ZHANG Zhong-Jing using big data technology | |
| CN104268656B (en) | Method for assessing cooperativity and effect degree of various kinds of traditional Chinese medicine with same biological function on biological function and traditional Chinese medicine compound optimizing method | |
| Yang et al. | Understanding traditional Chinese medicine via statistical learning of expert‐specific Electronic Medical Records | |
| Yadong et al. | Mining effect of Famous Chinese Medicine Doctors on Lung-cancer based on Association rules | |
| Yan et al. | Design of knowledge graph of traditional Chinese medicine prescription and knowledge analysis of implicit relationship | |
| Zhu et al. | TCDO: A community-based ontology for integrative representation and analysis of traditional Chinese drugs and their properties | |
| Lu et al. | Research on herbal combinations of traditional Chinese medicine for chronic gastritis based on network biology | |
| Ko | The Sikchi and the recorded cases of Seungjeongwon-Ilgi | |
| KR102376173B1 (en) | Method and apparatus for managing food cure information | |
| Li et al. | Data Exploration and Mining on Traditional Chinese Medicine | |
| Yu et al. | A Meta-analysis of Influencing Mediator Athletics on the Metabolic Syndrome Risk Factors Utilized Big Data Analysis | |
| Sun et al. | An approach to co-medication mechanism mining of Chinese Materia Medica and western medicines based on complex networks with the multi-source heterogeneous information |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201231 Address after: No.1 xc1001-2, Nanmen Gongnong Road, Chongfu Town, Tongxiang City, Jiaxing City, Zhejiang Province Patentee after: JIAXING JINJING E-COMMERCE Co.,Ltd. Address before: Hangzhou City, Zhejiang province 310036 Xiasha Higher Education Park forest Street No. 16 Patentee before: HANGZHOU NORMAL UNIVERSITY QIANJIANG College |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20211117 Address after: 257091 room 419, building a, Dongying Economic and Technological Development Zone Management Committee, No. 59, Fuqian street, Dongying District, Dongying City, Shandong Province Patentee after: Dongying dongkai New Material Industrial Park Co.,Ltd. Address before: No.1 xc1001-2, Nanmen Gongnong Road, Chongfu Town, Tongxiang City, Jiaxing City, Zhejiang Province Patentee before: JIAXING JINJING E-COMMERCE Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221229 Address after: 257091 Room 101, Building 5, Science and Technology Enterprise Accelerator, No. 17, Xinghuahe Road, Dongying Development Zone, Shandong Province Patentee after: Dongying Dongkai New Material Technology Research and Development Co.,Ltd. Address before: 257091 room 419, building a, Dongying Economic and Technological Development Zone Management Committee, No. 59, Fuqian street, Dongying District, Dongying City, Shandong Province Patentee before: Dongying dongkai New Material Industrial Park Co.,Ltd. |
|
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240126 Address after: Room 405, Building A, Technology Enterprise Accelerator, No. 17 Xinghuahe Road, Development Zone, Dongying City, Shandong Province, 257091 Patentee after: Dongying Dongkai Industrial Park Operation Management Co.,Ltd. Country or region after: China Address before: 257091 Room 101, Building 5, Science and Technology Enterprise Accelerator, No. 17, Xinghuahe Road, Dongying Development Zone, Shandong Province Patentee before: Dongying Dongkai New Material Technology Research and Development Co.,Ltd. Country or region before: China |