CN108717862B - Intelligent trial and development method based on machine learning - Google Patents
Intelligent trial and development method based on machine learning Download PDFInfo
- Publication number
- CN108717862B CN108717862B CN201810317800.4A CN201810317800A CN108717862B CN 108717862 B CN108717862 B CN 108717862B CN 201810317800 A CN201810317800 A CN 201810317800A CN 108717862 B CN108717862 B CN 108717862B
- Authority
- CN
- China
- Prior art keywords
- drug
- medicine
- preliminary
- probability
- preliminary diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/10—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to drugs or medications, e.g. for ensuring correct administration to patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/40—ICT specially adapted for the handling or processing of medical references relating to drugs, e.g. their side effects or intended usage
Abstract
The invention discloses an intelligent party-checking evolution model based on machine learning, which belongs to the technical field of machine learning models, and is characterized in that a classification algorithm is used for solving the model, historical prescription data and party-checking data are learned to obtain a probability matrix of the model based on a Bayesian algorithm, finally, an intelligent party-checking model is established through a medicine and a probability matrix of preliminary diagnosis, a medicine relevance model is established through medicine relevance analysis under the same preliminary diagnosis, relevance analysis between the preliminary diagnosis and the medicine is solved, and whether a recommended medicine accords with the preliminary diagnosis or not is determined on the premise of analyzing and formulating the preliminary diagnosis; standardizing the data of the medicine components through the correlation analysis between the preliminary diagnosis and the medicine components of the medicine; after the user symptom data and the drug indication data are normalized, the relevance between the user symptom and the corresponding drug component is analyzed.
Description
Technical Field
The invention belongs to the technical field of machine learning models, and particularly relates to an intelligent trial and development method based on machine learning.
Background
With the wide application of the existing internet remote inquiry, the inquiry flow of the patient is roughly simplified as follows: the patient comes to a PC and is connected with a video screen to describe the illness state through voice, a doctor obtains the symptoms of the illness state and the severity of the illness state, the doctor determines the etiology according to medical relevant knowledge to obtain initial diagnosis so as to obtain the disease type, then the doctor gives medicines according to symptoms and disease types, and corresponding relevant medicines are selected from a medicine database to form a prescription. And then the prescription is sent to a pharmacist by the system for prescription verification, and the pharmacist completes the prescription verification according to the prescription collocation, the patient information description, the preliminary diagnosis and the medicine dosage.
In general, a user consults a physician to describe relevant information such as the condition (i.e., symptoms) and the history of allergies. The physician then makes a preliminary diagnosis and prescribes a prescription based on the preliminary diagnosis and the user's symptoms, where the prescription typically contains one or more drugs, which all have attributes: the ingredients, applicable symptoms and contraindications of the medicine. The medicine components among different medicines need to be made into separate conflict analysis, the applicable symptoms among different medicines can be analyzed whether the medicines are applicable to the user symptoms or not, namely whether the medicines are taken according to the symptoms of the user, and the medicine contraindication of the medicines, the allergic history and the genetic medical history of the user and the like can form the verification of medicine taking safety.
Due to the development needs of products, an intelligent trial prescription development system based on machine learning is required to be established and used for assisting doctors to diagnose and serve patients quickly, accurately and efficiently.
Disclosure of Invention
In order to solve the above problems in the prior art, the present invention aims to provide an intelligent trial prescription development method based on machine learning to achieve the purpose of assisting a doctor pharmacist to diagnose and service a patient quickly, accurately and efficiently.
The technical scheme adopted by the invention is as follows: the utility model provides an intelligent auditor evolution method based on machine learning, which comprises the following contents:
(one) all ingredients of the drug are listed as { a1，a2，a3....awW components in total, the set of w components being denoted by A and containing a list of components { x } for the selected drug i1,x2.....,xkExpressing a list of k components of the ith medicine by using a set X, wherein the X belongs to A, i is a variable and takes a value of 1-n and expresses the ith medicine in the n medicines listed in the electronic prescription;
(II) obtaining the overall relevance F (y) based on a relevance analysis model of a classification algorithmj) The following functional formula is adopted:
wherein, wiRepresents the cost coefficient of drug i and wi∈[0，1]，P1ijRepresents the probability magnitude that the drug i belongs to the preliminary diagnosis group j and Pij∈[0，1](ii) a Then P2kjRepresents a drug component xkThe probability of belonging to the preliminary diagnostic packet j is then given by the following functional formula:
wherein, P2qjRepresenting the probability magnitude that the drug component q belongs to the preliminary diagnostic packet j;
by combining (2-1) and (2-2), the following functional formula is obtained:
the objective function is:
the preliminary diagnosis group j belongs to m preliminary diagnosis groups of the total, j belongs to (1, 2, 3.. eta.. m), and the probability of each medicine i belonging to each preliminary diagnosis group j is P1ijThen P1ijAn array of dimensions n × m, i.e.:
and (III) simplifying the process of the party to be examined into the following steps: the weighted sum of the correlation size of each ith drug in the n drugs prescribed in the prescription with the preliminary diagnosis group j of the disease is the overall correlation size F (y)j) Setting the total relevance weighted sum to be epsilon and epsilon to be (0, 1)]When the total correlation size is F (y)j) If the relation is less than epsilon, the grouping j of the initial diagnosis relation of the n medicines and the diseases is a strong relation, and the prescription is used for checking; when the overall relevance size F (y)j) When the dosage is larger than epsilon, it indicates that the prescribed n medicines and the primary diagnosis of the disease are presentIf the break relationship group j is weak, the prescription will not be checked.
Further, the probability matrix PijThe calculation steps are as follows:
(1) data initialization, for a given set of preliminary diagnoses CjSplitting and preliminarily diagnosing;
(2) selection of prescription E in preliminary diagnosisjPrescription EjIn which the medicine X is contained1，X2......Xk；
(3) For drug XiAnd i is less than or equal to k, and extracting a drug component vector { x1,x2.....,xnThe assumption is that the drug components are independent of each other; if the medicine components cannot be extracted, discarding the medicine, and regarding the medicine as an invalid medicine;
(4) recording the drug composition to preliminary diagnosis CjIf the component has appeared, the number of appearance times is increased by 1 in the record of the component, and at the same time, the corresponding medicine is recorded in the preliminary diagnosis CjThe number of occurrences in (1);
(5) recording preliminary diagnosis CjThe total number of samples is added by 1;
(6) repeating the steps (2) to (5) until all medicines in the prescription are completely learned so as to obtain a preliminary diagnosis list;
(7) traversing the preliminary diagnosis list, and calculating the preliminary diagnosis C of the medicine components in the preliminary diagnosis listjThe posterior probability of (2):
wherein x isiIs a pharmaceutical ingredient and xi∈{x1,x2.....,xnD is the sample space,
representing the drug ingredients in the preliminary diagnosis list for the preliminary diagnosis CjA posterior probability of (2), wherein, count (x)i∧Cj| D) represents the preliminary diagnosis CjChinese medicinal component xiIn the sample space DNumber of occurrences, count (C)j| D) represents the preliminary diagnosis CjNumber of occurrences in sample space D;
(8) repeating step (7) until a preliminary diagnosis CjAll the medicine components are calculated;
(9) traversing all the medicines in the preliminary diagnosis list;
(10) calculating drug XiFor preliminary diagnosis CjProbability of (c):
as a pharmaceutical ingredient<x1,x2...xn>Indicating drug XiDue to the medicine<x1,x2...xn>Probability p (x)1,x2...xn) And (3) as a constant, carrying out normalization processing on the constant, and according to Bayesian hypothesis, developing by applying a Bayesian formula to obtain:
wherein the formula (2-7) means the drug XiChinese medicine component for preliminary diagnosis CjThe product of the posterior probabilities of, i.e. the pharmaceutical product XiFor preliminary diagnosis CjThe probability of (d);
(11) repeating steps (9) - (10) until a preliminary diagnosis CjThe probability of all the medicines is calculated;
(12) repeating the steps (7) - (11) until all the medicines finish calculating the probability of all the initial diagnoses to obtain a probability matrix P1ij。
Further, it is characterized in that the cost coefficient w in the formula (2-1)iThe calculation formula is as follows:
wherein the cost coefficient wiPreliminary applicability to drugs omegaiIn inverse proportion, the stronger the applicability of the medicine is, the lower the cost of the medicine taking by mistake is, and k represents a correction parameter, wherein C'iTo representHow many different preliminary diagnoses the drug i can appear in to measure the suitability of the drug i, ω isi＝C′i。
The invention has the beneficial effects that:
1. the intelligent examination prescription developing method based on machine learning can simplify the inquiry flow, extract a mathematical model, establish a corresponding model and solve the model through a classification algorithm; learning historical prescription data and auditor data based on Bayesian algorithm to obtain probability matrix of model, and finally obtaining probability matrix P of medicine and preliminary diagnosisijAn intelligent auditor model is established, and a medicine relevance model is established by establishing medication relevance analysis under the same preliminary diagnosis, so that the purpose of opening an intelligent auditor is achieved.
Detailed Description
The invention is further illustrated by the following specific examples.
The invention provides an intelligent party-checking and developing method based on machine learning, which comprises the following steps:
(one) all ingredients of the drug are listed as { a1，a2，a3....awW components in total, the set of w components being denoted by A and containing a list of components { x } for the selected drug i1,x2.....,xkExpressing a list of k components of the ith medicine by using a set X, wherein the X belongs to A, i is a variable and takes a value of 1-n and expresses the ith medicine in the n medicines listed in the electronic prescription;
(II) a relevance analysis model based on a classification algorithm, wherein the relevance size is F (y)j) The following functional formula is adopted:
n
(II) obtaining the overall relevance F (y) based on a relevance analysis model of a classification algorithmj) The following functional formula is adopted:
wherein, wiRepresents the cost coefficient of drug i and wi∈[0，1]Indicates that if the cost of the drug is wrong, the cost is higher if the cost coefficient is larger, wherein wip1ijIndicating the effectiveness of the ith drug on the jth preliminary diagnosis, if any>0, indicates that the drug i has a promoting effect on the preliminary diagnosis j. Otherwise, it has inhibitory effect, P1ijRepresents the probability magnitude that the drug i belongs to the preliminary diagnosis group j and Pij∈[0，1](ii) a Then P2kjRepresents a drug component xkThe probability of belonging to the preliminary diagnostic packet j is then given by the following functional formula:
wherein, P2qjRepresenting the probability magnitude that the drug component q belongs to the preliminary diagnostic packet j;
by combining (2-1) and (2-2), the following functional formula is obtained:
the objective function is:
the preliminary diagnosis group j belongs to m preliminary diagnosis groups of the total, j belongs to (1, 2, 3.. eta.. m), and the probability of each medicine i belonging to each preliminary diagnosis group j is P1ijThen P1ijAn array of dimensions n × m, i.e.:
(III) for each prescription process, each ith medicine in the n medicines prescribed in the prescription is respectively associated with the preliminary diagnosis of the diseaseThe weighted sum of the relevance sizes of the packets j is the overall relevance size F (y)j) Setting the total relevance weighted sum to be epsilon and epsilon to be (0, 1)]When the total correlation size is F (y)j) If the relation is less than epsilon, the grouping j of the initial diagnosis relation of the n medicines and the diseases is a strong relation, and the prescription is used for checking; when the overall relevance size F (y)j) If the relation is more than epsilon, the grouping j of the preliminary diagnosis relation between the n medicines and the disease is weak, and the prescription is not used for diagnosis.
Relevance size F (y)j) The geometrical meaning of (A) is shown in FIG. 1, the area A represents a preliminary diagnosis (or a prescription), and the area B represents the area B1Region B2Region B3Region Bi..nIndicating the medicine, the areas A are respectively arranged in the areas B1Region B2Region B3Region Bi..nThe size of the area in (1) represents the probability. If only one drug is recommended then the degradation is: the region A and the region B1Region B2Region B3Region Bi..nThe repeat area in between. Medicine BiThe larger the area appearing in the preliminary diagnosis a, the larger the probability. If all drugs are associated with the preliminary diagnosis A, then the prescription is considered valid, if drug B is presentiIf the medicine is not in the area where the preliminary diagnosis of A is located, the medicine B is representediIndependent of the preliminary diagnosis A, or the drug BiNot applicable to the preliminary diagnosis A.
Cost coefficient w in the formula (2-1)iIs equal to the applicability omega of the medicineiWherein i represents the ith drug product according to the formula:
wherein the cost coefficient wiPreliminary applicability to drugs omegaiIn inverse proportion, the stronger the applicability of the medicine is, the lower the cost of the medicine taking by mistake is, and k represents a correction parameter, wherein C'iIndicates how many different initiatives the drug i may appear inIn the diagnosis, to measure the applicability of drug i, ω isi＝C′i。
Based on naive Bayes algorithm, the probability matrix P isijThe calculation steps are as follows:
(1) data initialization, for a given set of preliminary diagnoses CjSplitting and preliminarily diagnosing;
(2) selection of prescription E in preliminary diagnosisjPrescription EjIn which the medicine X is contained1，X2......Xk；
(3) For drug XiAnd i is less than or equal to k, and extracting a drug component vector { x1,x2.....,xnThe assumption is that the drug components are independent of each other; if the medicine components cannot be extracted, discarding the medicine, and regarding the medicine as an invalid medicine;
(4) recording the drug composition to preliminary diagnosis CjIf the component has appeared, the number of appearance times is increased by 1 in the record of the component, and at the same time, the corresponding medicine is recorded in the preliminary diagnosis CjThe number of occurrences in (1);
(5) recording preliminary diagnosis CjThe total number of samples is added by 1;
(6) repeating the steps (2) to (5) until all medicines in the prescription are completely learned so as to obtain a preliminary diagnosis list;
(7) traversing the preliminary diagnosis list, and calculating the preliminary diagnosis C of the medicine components in the preliminary diagnosis listjThe posterior probability of (2):
wherein x isiIs a pharmaceutical ingredient and xi∈{x1,x2.....,xn}; d is a sample space;
representing the drug ingredients in the preliminary diagnosis list for the preliminary diagnosis CjA posterior probability of (2), wherein, count (x)i∧Cj| D) represents the preliminary diagnosis CjChinese medicinal component xiNumber of occurrences in sample space D, count (C)j| D) represents the preliminary diagnosis CjNumber of occurrences in sample space D;
(8) repeating step (7) until a preliminary diagnosis CjAll the medicine components are calculated;
(9) traversing all the medicines in the preliminary diagnosis list;
(10) calculating drug XiFor preliminary diagnosis CjProbability of (c):
as a pharmaceutical ingredient<x1,x2...xn>Indicating drug XiDue to the medicine<x1,x2...xn>Probability p (of (a))<x1,x2...xn>) And (3) as a constant, carrying out normalization processing on the constant, and according to Bayesian hypothesis, developing by applying a Bayesian formula to obtain:
(11) repeating steps (9) - (10) until a preliminary diagnosis CjThe probability of all the medicines is calculated;
(12) repeating the steps (7) - (11) until all the medicines finish calculating the probability of all the initial diagnoses to obtain a probability matrix P1ij。
Aiming at the Bayes assumption in the step (3), the following assumption is introduced into the Bayes classifier, and in the given class C, all the attributes are assumed to be independent from each other, namely:
each pharmaceutical ingredient A of the medicine1,A2....AnIndependent of each other, each medicine component A1,A2....AnThe relationship with the preliminary diagnosis C is shown in fig. 2 below, the constant we use the normalization factor a for representation. The posterior probability given a preliminary diagnostic category C given drug X is:
assigning class ciIt should satisfy:
if classification ciFor optimal classification, it is then also necessary to satisfy:
p(ci|<a1,L,an>)>p(cj|<a1,L,an>),i≠j
wherein a is1,L,anRepresenting n components of the drug.
The prior probability distribution of the class C can simply obtain the maximum likelihood estimation of the class C from the training set data, the maximum likelihood estimation is equal to the frequency of the different class attributes appearing in the data set, and the calculation complexity is O (| D |).
The profile for the bayesian classification model is as follows:
the classification includes a rule-based classification (query) and an irregular classification (directed learning).
Bayesian classification
Is irregular classification, which is trained by a training set (classified example set) to generalize classifiers (the predicted variables are discrete, called classification, and continuous, called regression), and classifies the unclassified data by using the classifiers. Typical classifiers in the bayesian classifier include a naive bayes classifier, a bayesian network classifier, a tree-enhanced naive bayes classification model TAN, and the like.
The bayesian classification has the following characteristics:
A. bayesian classification does not absolutely assign an instance to a class, but rather computes the probability of belonging to a class, the class with the highest probability being the class to which the instance belongs.
B. In general, all attributes in a bayesian classification function directly or indirectly, i.e. all attributes participate in the classification, rather than one or several attributes determining the classification.
C. The attributes of the bayesian classification instances may be discrete, continuous, or mixed.
It is based on the prior probability of h, the probability of observing different data given the assumption, and the observed data itself. The prior probability of h is denoted by P (h) without training data. P (h) is called the prior probability (priorprobability) of h, representing background knowledge about the probability that h is the correct hypothesis. P (D) denotes the prior probability of the training data D to be observed, i.e. the probability of D when no hypothesis is determined. P (D | h) represents the probability of data D assuming h holds. We need to find the probability that h holds given the training data D, i.e. the posterior probability P (D | h) of h, then we find the method of calculating the posterior probability P (D | h) by the bayesian formula:
the maximum a posteriori assumption: in many learning scenarios, the learner considers a set of candidate hypotheses H, and the hypothesis H ∈ H in which the likelihood of requiring generation of data D is greatest. The most probable hypothesis is called the maximum a posteriori hypothesis, noted as: h isMAP
Where, we remove P (D) and replace it with a constant a, since P (D) is a constant that is independent of h.
In the present embodiment, assume that A1,A2....AnRepresents the characteristic attributes of the medicine (the values are all medicine components and n components in total), and assuming that m classes are counted in the preliminary diagnosis, C is equal to{C1,C2,C3,L,Cm}. Given a particular drug X, the properties of its drug components are { X }1,x2.....,xnX hereiIs attribute AiThe specific value of (b) represents all the component information of the drug X. The medicine belongs to class CiHas a posterior probability of P (X | C)i) And C (X) represents the category to which the medicine is finally classified, namely the category with the highest probability.
I.e. to predict to which class the drug will ultimately belong, given the drug composition, i.e. the posterior probability under that class is the greatest.
The summary of the algorithm for bayes is as follows:
the naive Bayes algorithm is based on a Bayes probability formula, and firstly, a conditional probability formula is introduced. The probability of occurrence at time B, under the condition that time a has occurred, is referred to as the conditional probability (also called a posteriori probability) of event B at a given event a, which is denoted as P (B | a), and accordingly P (a) is referred to as the unconditional probability (also called a priori probability).
In the case of event A, the probability of event B occurring is
Wherein:
P(A·B)＝P(A)P(B|A)＝P(B)·P(A|B)
combining the above formula to obtain:
where S is assumed to be the sample space, A1,A2,A3,L AmIs a division of the sample space S. Wherein P (A)j)>0(j＝1,2,3L,m) according to the total probability formula:
the Bayesian formula of the combination is:
for P (A) in the intelligent party checking modeljI B) indicates that the drug B belongs to the preliminary diagnosis AjThe posterior probability of (d). P (B | A)j) Indicates the preliminary diagnosis AjThe probability of containing medicine B is determined according to the central limit theorem P (A)j) A constant value can be derived from historical statistics.
The invention is not limited to the above alternative embodiments, and any other various forms of products can be obtained by anyone in the light of the present invention, but any changes in shape or structure thereof, which fall within the scope of the present invention as defined in the claims, fall within the scope of the present invention.
Claims (2)
1. An intelligent party-seeking method based on machine learning is characterized by comprising the following steps:
(one) tabulating all the ingredients of the drug as
In total, w components, the set of w components being denoted by A, and for a selected drug i, the list of components contained therein is
Tabulating and aggregating the k components of the ith drug
It is shown that,
wherein, i is a variable with a value of 1-n and represents the ith medicine in the n medicines listed in the electronic prescription;
(II) obtaining the overall relevance based on a relevance analysis model of a classification algorithm
The following functional formula is adopted:
wherein the content of the first and second substances,
represents the cost coefficient of drug i and
，
represents the size of the probability that the drug i belongs to the preliminary diagnostic group j and
；
indicating the composition of a drug
The probability of belonging to the preliminary diagnostic packet j is then given by the following functional formula:
wherein the content of the first and second substances,
representing the probability magnitude that the drug component q belongs to the preliminary diagnostic packet j;
by combining (2-1) and (2-2), the following functional formula is obtained:
the objective function is:
wherein epsilon represents a threshold value of the overall relevance size, and epsilon is a constant of (0, 1);
said preliminary diagnostic packet j belongs to the total of m preliminary diagnostic packets, then
The probability of each drug i belonging to each preliminary diagnostic grouping j is of the magnitude
Then probability matrix
An array of dimensions n × m, i.e.:
and (III) simplifying the process of the party to be examined into the following steps: the weighted sum of the correlation size of each ith drug in the n drugs prescribed in the prescription and the preliminary diagnosis group j of the disease is the overall correlation size
When the overall relevance size is large
If the relation is less than epsilon, the grouping j of the initial diagnosis relation of the n medicines and the diseases is a strong relation, and the prescription is used for checking; when the overall relevance size
If the relation is not less than epsilon, the grouping j of the preliminary diagnosis relation between the n medicines and the disease is weak, and the prescription is not used for diagnosis.
2. The machine-learning-based intelligent trial development method according to claim 1, wherein the cost coefficients in the formula (2-1)
The calculation formula is as follows:
wherein the cost coefficient
Preliminary applicability to drugs
Inversely proportional, the stronger the drug's applicability, the less costly the drug is, k represents the correction parameter, where
Indicating a drug
Can be presented in how many different preliminary diagnoses to measure the drug
Applicability of
。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810317800.4A CN108717862B (en) | 2018-04-10 | 2018-04-10 | Intelligent trial and development method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810317800.4A CN108717862B (en) | 2018-04-10 | 2018-04-10 | Intelligent trial and development method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108717862A CN108717862A (en) | 2018-10-30 |
| CN108717862B true CN108717862B (en) | 2022-05-03 |
Family
ID=63898920
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810317800.4A Active CN108717862B (en) | 2018-04-10 | 2018-04-10 | Intelligent trial and development method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108717862B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110706775A (en) * | 2019-09-29 | 2020-01-17 | 北京一药良心信息科技有限公司 | Remote intelligent party checking method and system |
| CN111383732B (en) * | 2020-03-23 | 2023-04-18 | 平安医疗健康管理股份有限公司 | Medicine auditing method, device, computer system and readable storage medium based on mutual exclusion identification |
| CN111640481B (en) * | 2020-05-22 | 2023-06-20 | 南京大学 | Parkinson's disease drug recommendation model based on multi-source data fusion |
| CN116825364A (en) * | 2023-08-29 | 2023-09-29 | 江苏盛泰科技集团有限公司 | High-risk group health identification judgment system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106108846A (en) * | 2016-06-20 | 2016-11-16 | 中山大学 | A kind of intelligent drug risk monitoring method and system |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5845255A (en) * | 1994-10-28 | 1998-12-01 | Advanced Health Med-E-Systems Corporation | Prescription management system |
| CN101313221A (en) * | 2005-09-21 | 2008-11-26 | Ccc诊断有限责任公司 | Comprehensive diagnostic testing procedures for personalized anticancer chemotherapy(pac) |
| KR101419753B1 (en) * | 2012-10-16 | 2014-07-17 | 안형준 | Medicine retrieval system minimized individual side-effects based on private single nucleotide polymorphism and the method thereof |
| CN103365978B (en) * | 2013-07-01 | 2017-03-29 | 浙江大学 | TCM data method for digging based on LDA topic models |
| CN104978474B (en) * | 2014-04-11 | 2019-02-01 | 中国中医科学院中医临床基础医学研究所 | A kind of method of evaluating drug effect and system based on molecular network |
| CN107302958A (en) * | 2016-04-19 | 2017-10-31 | 郭思东 | The Functional Design of Chinese medicine health-care series beverage and the method for formulation optimization |
| CN107341345A (en) * | 2017-06-22 | 2017-11-10 | 厦门大学 | A kind of clinical rational drug use methods of risk assessment based on electronic health record big data |
| CN206932392U (en) * | 2017-07-24 | 2018-01-26 | 四川骏逸富顿科技有限公司 | A kind of micro- interrogation service platform audio frequency processing system |
| CN107578818B (en) * | 2017-08-25 | 2021-09-10 | 广州宝荣科技应用有限公司 | Auxiliary evolution method and device based on deep learning |
-
2018
- 2018-04-10 CN CN201810317800.4A patent/CN108717862B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106108846A (en) * | 2016-06-20 | 2016-11-16 | 中山大学 | A kind of intelligent drug risk monitoring method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108717862A (en) | 2018-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108717862B (en) | Intelligent trial and development method based on machine learning | |
| US20210193320A1 (en) | Machine-learning based query construction and pattern identification for hereditary angioedema | |
| Wollenstein-Betech et al. | Personalized predictive models for symptomatic COVID-19 patients using basic preconditions: hospitalizations, mortality, and the need for an ICU or ventilator | |
| Arias-Londono et al. | Artificial intelligence applied to chest X-ray images for the automatic detection of COVID-19. A thoughtful evaluation approach | |
| US20210233658A1 (en) | Identifying Relevant Medical Data for Facilitating Accurate Medical Diagnosis | |
| US11923056B1 (en) | Discovering context-specific complexity and utilization sequences | |
| Jovanovic et al. | Building interpretable predictive models for pediatric hospital readmission using Tree-Lasso logistic regression | |
| Zhang et al. | Comparing data mining methods with logistic regression in childhood obesity prediction | |
| Gürbüz et al. | A new adaptive support vector machine for diagnosis of diseases | |
| WO2009083833A1 (en) | Retrieval of similar patient cases based on disease probability vectors | |
| US10566093B2 (en) | Data integration using graph structures | |
| Zheng et al. | Capturing feature-level irregularity in disease progression modeling | |
| US20200265953A1 (en) | Identifying Valid Medical Data for Facilitating Accurate Medical Diagnosis | |
| CN112216402A (en) | Epidemic situation prediction method and device based on artificial intelligence, computer equipment and medium | |
| Duggal et al. | Impact of selected pre-processing techniques on prediction of risk of early readmission for diabetic patients in India | |
| Tiwari et al. | A knowledge infused context driven dialogue agent for disease diagnosis using hierarchical reinforcement learning | |
| Robinson et al. | Deep learning models for COVID-19 chest x-ray classification: Preventing shortcut learning using feature disentanglement | |
| Lakshman Narayana et al. | Optimized Nature-Inspired Computing Algorithms for Lung Disorder Detection | |
| WO2021126562A1 (en) | Systems and methods for processing electronic images for health monitoring and forecasting | |
| Kumaravel et al. | A simplified framework for the detection of intracranial hemorrhage in CT brain images using deep learning | |
| CN115050442A (en) | Disease category data reporting method and device based on mining clustering algorithm and storage medium | |
| Zhang et al. | Fairness in missing data imputation | |
| Gollapalli et al. | An Artificial Intelligence Approach for Data Modelling Patients Inheritance of Sickle Cell Disease (SCD) in the Eastern Regions of Saudi Arabia. | |
| CN112435745B (en) | Method and device for recommending treatment strategy, electronic equipment and storage medium | |
| Uddin et al. | A novel approach utilizing machine learning for the early diagnosis of Alzheimer's disease |
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 |