CN113569999A - Training method and device for medicine recommendation model, storage medium and computer equipment - Google Patents
Training method and device for medicine recommendation model, storage medium and computer equipment Download PDFInfo
- Publication number
- CN113569999A CN113569999A CN202111017415.6A CN202111017415A CN113569999A CN 113569999 A CN113569999 A CN 113569999A CN 202111017415 A CN202111017415 A CN 202111017415A CN 113569999 A CN113569999 A CN 113569999A
- Authority
- CN
- China
- Prior art keywords
- information
- drug
- loss function
- recommendation model
- preset
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Public Health (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Probability & Statistics with Applications (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a training method and a training device for a drug recommendation model, a storage medium and computer equipment, and mainly aims to train a lightweight drug recommendation model to perform online deployment by using a knowledge distillation method. The method comprises the following steps: extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model; and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model. The present invention relates to the fields of artificial intelligence and digital medicine.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a training method and device of a medicine recommendation model, a storage medium and computer equipment.
Background
With the rise of the on-line inquiry mode in the medical field, a patient can go home and perform on-line inquiry through the Internet, and in the on-line inquiry process, a doctor determines diseases obtained by the patient according to the state of illness described by the patient and prescribes corresponding prescription medicines for the patient.
Currently, in the process of diagnosing patients, machine learning techniques can be used to recommend corresponding drugs to patients. However, because the existing machine learning model has many parameters and a heavy model, the machine learning model for drug recommendation is only suitable for offline and cannot be deployed online, and thus cannot assist doctors in online inquiry.
Disclosure of Invention
The invention provides a training method and a training device for a medicine recommendation model, a storage medium and computer equipment, and mainly aims to train a lightweight medicine recommendation model to be deployed online by using a knowledge distillation method, so that medicine recommendation precision is guaranteed, and medicine recommendation efficiency is improved.
According to a first aspect of the present invention, there is provided a method for training a drug recommendation model, comprising:
acquiring sample disease information of a patient and corresponding actual medicine information;
extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model;
inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information;
respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information;
and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
According to a second aspect of the present invention, there is provided a training apparatus for a drug recommendation model, comprising:
the system comprises an acquisition unit, a processing unit and a display unit, wherein the acquisition unit is used for acquiring sample disease information of a patient and corresponding actual medicine information;
the extraction unit is used for extracting first distribution characteristics of the sample disease information aiming at different medicine information by using a preset natural language model;
the recommending unit is used for inputting the sample disease information into a preset initial drug recommending model for drug recommending to obtain second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information;
a building unit, configured to respectively build a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristic and the second distribution characteristic, and the actual drug information and the recommended drug information;
and the training unit is used for carrying out iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
According to a third aspect of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:
acquiring sample disease information of a patient and corresponding actual medicine information;
extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model;
inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information;
respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information;
and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
According to a fourth aspect of the present invention, there is provided a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the program:
acquiring sample disease information of a patient and corresponding actual medicine information;
extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model;
inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information;
respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information;
and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
According to the training method, the training device, the storage medium and the computer equipment of the medicine recommendation model, compared with the mode that a doctor recommends corresponding medicines for a patient by adopting an offline machine learning technology aiming at the diseases obtained by the patient at present, the method and the system obtain sample disease information of the patient and corresponding actual medicine information; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; meanwhile, the sample disease information is input into a preset initial drug recommendation model for drug recommendation, and second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information are obtained; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; finally, iterative training is carried out on the initial drug recommendation model based on the first loss function and the second loss function, a preset drug recommendation model is built, so that a first loss function and a second loss function corresponding to the initial drug recommendation model can be built by obtaining a first distribution characteristic and a second distribution characteristic, and the actual drug information and the recommended drug information, and the preset drug recommendation model is built based on the first loss function and the second loss function, so that the preset drug recommendation model has learning capability of a model under a line, meanwhile, the preset drug recommendation model is light in weight and can be deployed on the line, meanwhile, as the preset drug recommendation model is compared with the model under the line, model parameters are reduced, the preset drug recommendation model can reduce the amount of calculation and improve the drug recommendation efficiency while ensuring the drug recommendation precision, effectively assisting the doctor to perform on-line inquiry, thereby improving the diagnosis efficiency of the doctor.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a flowchart illustrating a method for training a drug recommendation model according to an embodiment of the present invention;
FIG. 2 is a flow chart of another method for training a drug recommendation model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a training apparatus for a drug recommendation model according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a training apparatus for a drug recommendation model according to another embodiment of the present invention;
fig. 5 shows a physical structure diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The invention will be described in detail hereinafter with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
At present, because the existing machine learning model has more parameters and heavier model, the machine learning model for recommending the medicines is only suitable for offline and cannot be deployed online, so that a doctor cannot be assisted to perform online inquiry.
In order to solve the above problem, an embodiment of the present invention provides a method for training a drug recommendation model, as shown in fig. 1, the method includes:
101. acquiring sample disease information of a patient and corresponding actual medicine information thereof.
The sample disease information is the name or code of various diseases diagnosed by doctors, such as coronary heart disease, diabetes, acute enteritis, etc., and the actual medicine information is the name or code of medicines for treating diseases, such as erythromycin, penicillin, etc.
For the embodiment of the invention, in order to overcome the defect that the online inquiry of a doctor cannot be assisted due to the fact that the machine learning model has more parameters and a heavier model and cannot be deployed online in the prior art, the embodiment of the invention adopts a knowledge distillation method to perform slimming for the offline model, trains a lightweight model with the learning capability of the offline model to enable the online deployment and assist the doctor in performing online inquiry. The embodiment of the invention is mainly applied to a scene of training a medicine recommendation model, and the execution main body of the embodiment of the invention is a device or equipment capable of training the medicine recommendation model, and can be specifically arranged on one side of a server.
Specifically, the actual drug information corresponding to different disease information is different, a large amount of disease information suffered by the patient and the actual drug information prescribed by the doctor for the patient in the inquiry process are stored in the sample database, and when the drug recommendation model needs to be trained, the sample disease information of a large amount of patients and the actual drug information corresponding to the sample disease information need to be extracted from the sample database, the sample disease information and the actual drug information are used as sample data, and the preset drug recommendation model is constructed based on the sample data.
102. And extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model.
The preset natural language model may be a preset BERT model, and the first distribution feature is a distribution feature after disease information is processed by a full link layer in a prediction process and before a softmax layer is input.
For the embodiment of the present invention, in order to obtain first distribution characteristics of sample disease information for different pieces of drug information, a semantic information vector corresponding to the sample disease information is first extracted by using an encoder in a preset BERT model, and a specific process of extracting the semantic information vector is shown in step 203, and then the semantic information vector corresponding to the disease information is input into a preset classifier for classification, where the preset classifier includes a full connection layer and a softmax layer, and after the sample disease information in the preset classifier is extracted in the classification process and processed by the full connection layer, and a first distribution characteristic before the softmax layer is input, the sample disease information has different first distribution characteristics for different pieces of drug information. It should be noted that the preset BERT model and the preset classifier are both models trained in advance on line, and since the preset BERT model includes a large number of parameters and is difficult to deploy into the online inquiry system, in this embodiment, the preset BERT model and the preset classifier trained on line are used as a teacher model, and the teacher model is used to train a lightweight student model (preset drug recommendation model) to perform online deployment, so that the drug recommendation accuracy is ensured, and meanwhile, the drug recommendation efficiency is improved by reducing the model parameters and the calculation amount, thereby assisting a doctor to perform online inquiry.
103. And inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information.
The initial drug recommendation model is a model to be deployed on line, the initial drug recommendation model may be a multilayer sensor, and the second distribution feature is a distribution feature of the disease information after being processed by a full connection layer of the initial drug recommendation model and before being input into a softmax layer.
For the embodiment of the invention, in order to train an initial drug recommendation model (student model) by using a teacher model, firstly, parameters of a plurality of layers of sensors are initialized to obtain an initial drug recommendation model, then, sample disease information is input into the initial drug recommendation model, after the parameters are processed by a full connection layer of the initial drug recommendation model and before a softmax layer is input, second distribution characteristics of the sample disease information aiming at different drug information are obtained, then, the second distribution characteristics are input into the softmax layer of the plurality of layers of sensors to obtain recommended drug information corresponding to the sample disease information, then, a first loss function and a second loss function of the initial drug recommendation model are respectively constructed according to the first distribution characteristics and the second distribution characteristics as well as actual drug information and recommended drug information, and finally, based on the first loss function and the second loss function, a preset medicine recommendation model is built, online deployment is carried out on the model, a doctor is assisted to carry out online inquiry, and inquiry efficiency of the doctor is improved. The specific structure of the multi-layered sensor and the specific process of acquiring the second distribution characteristics and the recommended medicine information are shown in step 205.
104. And respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristic and the second distribution characteristic, and the actual drug information and the recommended drug information.
For the embodiment of the present invention, in order to enable the constructed drug recommendation model to have the learning capability of the preset BERT model and meet the requirement of the drug recommendation accuracy in the actual recommendation process, a first loss function and a second loss function corresponding to the initial drug recommendation model need to be respectively constructed, specifically, after a first distribution feature of sample disease information extracted from the BERT model for different drug information and a second distribution feature of the sample disease information extracted from the initial drug recommendation model for different drug information are constructed, a first loss function corresponding to the initial drug recommendation model is constructed according to the first distribution feature and the second distribution feature, a second loss function of the initial drug recommendation model is constructed according to the actual drug information and the recommended drug information, and specific formulas for constructing the first loss function and the second loss function are shown in step 206, and finally, a preset medicine recommendation model is constructed based on the first loss function and the second loss function, so that the medicine recommendation precision is ensured, and the medicine recommendation efficiency is improved.
105. And performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
The preset medicine recommendation model is a medicine recommendation model for final online deployment.
For the embodiment of the invention, in order to construct the preset drug recommendation model, after the first loss function and the second loss function corresponding to the initial drug recommendation model are constructed, the weight coefficients corresponding to the first loss function and the second loss function are respectively determined, the total loss function of the initial drug recommendation model is constructed based on the weight coefficients, the initial drug recommendation model is iteratively trained by using the total loss function, the parameters of the initial drug recommendation model are updated, and the preset drug recommendation model is finally obtained.
According to the training method of the medicine recommendation model, compared with the mode that a doctor recommends a corresponding medicine for a patient by adopting an offline machine learning technology aiming at the disease of the patient at present, the method provided by the invention acquires sample disease information of the patient and corresponding actual medicine information; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; meanwhile, the sample disease information is input into a preset initial drug recommendation model for drug recommendation, and second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information are obtained; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; finally, iterative training is carried out on the initial drug recommendation model based on the first loss function and the second loss function, a preset drug recommendation model is built, so that a first loss function and a second loss function corresponding to the initial drug recommendation model can be built by obtaining a first distribution characteristic and a second distribution characteristic, and the actual drug information and the recommended drug information, and the preset drug recommendation model is built based on the first loss function and the second loss function, so that the preset drug recommendation model has learning capability of a model under a line, meanwhile, the preset drug recommendation model is light in weight and can be deployed on the line, meanwhile, as the preset drug recommendation model is compared with the model under the line, model parameters are reduced, the preset drug recommendation model can reduce the amount of calculation and improve the drug recommendation efficiency while ensuring the drug recommendation precision, effectively assisting the doctor to perform on-line inquiry, thereby improving the diagnosis efficiency of the doctor.
Further, in order to better describe the above training process of the drug recommendation model, as a refinement and extension of the above embodiment, an embodiment of the present invention provides another training method of a drug recommendation model, as shown in fig. 2, the method includes:
201. acquiring sample disease information of a patient and corresponding actual medicine information thereof.
For the embodiment of the invention, when a doctor asks for a patient, the doctor needs to prescribe corresponding actual drug information for the patient according to the disease information suffered by the patient, and the disease information and the actual drug information prescribed by the doctor according to the disease information are both stored in the sample database, so when a drug recommendation model needs to be constructed, the sample disease information of the patient and the actual drug information corresponding to the sample disease information need to be extracted from the sample database, and the sample disease information and the actual drug information are used as a training set, the drug recommendation model is constructed based on the training set, and online deployment is performed, so that the doctor is assisted to perform online inquiry, and the inquiry efficiency of the doctor can be improved.
202. Determining each character contained in the sample disease information and an embedded vector corresponding to each character.
For the embodiment of the present invention, in order to obtain the first distribution characteristics of the sample disease information for different pieces of drug information, it is first required to determine each character included in the sample disease information, for example, if the sample disease information is acute enteritis, each character corresponding to the sample disease information is acute/sexual/intestinal/tract/inflammation, then each character in the sample disease information is converted into an embedded vector by Word embedding methods such as Word2Vec, and the embedded vector corresponding to each character is input into a preset natural language model, so as to extract the first distribution characteristics of the sample disease information for different pieces of drug information.
203. And inputting the embedded vector into the preset natural language model to extract semantic information, so as to obtain a semantic information vector corresponding to the sample disease information.
The preset natural language model may be specifically a preset BERT model, the preset BERT model includes a plurality of encoders, for example, 8 encoders, each encoder is connected end to end, an output of a previous encoder may be used as an input of a next encoder, and the encoder specifically includes an attention layer and a feedforward neural network layer.
For the embodiment of the present invention, in order to extract the semantic information vector corresponding to the sample disease information, step 203 specifically includes: inputting the embedded vector to the attention layer for feature extraction to obtain a first feature vector corresponding to each character; adding the first feature vector and the embedded vector to obtain a second feature vector corresponding to each character; and inputting the second feature vector to the feedforward neural network layer for feature extraction to obtain a semantic information vector corresponding to the sample disease information. The first feature vector is an output vector of an attention layer, and the semantic information vector corresponding to the sample disease information is an output vector of a feedforward neural network layer of a last encoder. Specifically, in the process of extracting semantic information vectors corresponding to sample disease information by using a preset BERT model, firstly, an embedded vector corresponding to each character is input to an attention layer of a first encoder in the preset BERT model to perform feature extraction, so as to obtain an output vector of the attention layer, that is, a first feature vector corresponding to each character, wherein the specific process of performing feature extraction in the attention layer is as follows: determining a query vector, a key vector and a value vector corresponding to each character according to the embedded vector corresponding to each character; multiplying a query vector corresponding to a target character in each character by a key vector corresponding to each character to obtain the attention score of each character for the target character; and multiplying and summing the attention scores corresponding to the characters and the value vectors to obtain a first feature vector corresponding to the target character.
For the embodiment of the present invention, in the process of obtaining the first feature vector corresponding to each character, the embedded vector corresponding to each character in the sample disease information may be multiplied by the weight matrix corresponding to the attention layer in the preset BERT model to obtain the query vector, the key vector, and the value vector corresponding to each character, furthermore, the attention score corresponding to each character needs to be calculated, when the attention score corresponding to any one character (target character) in each character is calculated, each character in the sample disease information needs to be used to score the target character, specifically, the query vector corresponding to the target character is multiplied by the key vector corresponding to each character to obtain the score value of each character to the target character, that is, the attention score value, and then the attention score and the value vector corresponding to each character are multiplied and summed to finally obtain the attention layer output vector corresponding to the target character, namely, the first feature vector corresponding to the target character, so that the first feature vector corresponding to each character can be determined in the above manner, so as to obtain the semantic information vector corresponding to the sample disease information by using the first feature vector corresponding to each character.
Further, in order to obtain semantic information vectors corresponding to sample disease information, after extracting first feature vectors corresponding to respective characters, the first feature vector is added to the embedded vector corresponding to each character to obtain a second feature vector corresponding to each character, inputting the second feature vector into the feedforward neural network layer of the first encoder to perform feature extraction to obtain the output vector of the first encoder, because the preset BERT model in the embodiment of the invention comprises a plurality of encoders, and the encoders are connected in series end to end, therefore, the output vector of the first coder is input into the second coder for feature extraction to obtain the output vector of the second coder, and finally, determining the output vector of the last encoder as a semantic information vector corresponding to the sample disease information.
204. And inputting the semantic information vector into a preset classifier for classification, and extracting a first distribution characteristic of the sample disease information aiming at different medicine information.
The preset classifier is a neural network model and specifically comprises a full connection layer and a softmax layer.
For the embodiment of the invention, after the semantic information vector corresponding to the sample disease information is determined, the semantic information vector is input into the neural network model for classification, after the semantic information vector is output through a full connection layer in a preset neural network model and before the semantic information vector is input into a softmax layer, the distribution characteristic corresponding to the semantic information vector is extracted, and the distribution characteristic is used as the first distribution characteristic of the sample disease information for different drug information, so that the loss function corresponding to the initial drug recommendation model is constructed by using the first distribution characteristic.
205. And inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information.
The preset initial medicine recommendation model can be a multilayer sensor, and the multilayer sensor is a neural network model and comprises an input layer, a hidden layer and an output layer.
For the embodiment of the present invention, in order to obtain the second distribution characteristics of the sample disease information for the different pieces of drug information and the recommended drug information corresponding to the sample disease information, step 205 specifically includes: inputting the sample disease information into the multilayer perceptron, extracting the characteristics output by the last full-connection layer in the multilayer perceptron, and determining the characteristics output by the last full-connection layer as second distribution characteristics of the sample disease information aiming at different medicine information; and inputting the characteristics output by the last full-connection layer into a softmax layer in the multilayer perceptron to obtain recommended medicine information corresponding to the sample disease information.
Specifically, the embedded vectors corresponding to the characters in the sample disease information are input to a hidden layer through an input layer of the multilayer perceptron, and the result output through the hidden layer is as follows:
f(W1x+b1)
where x is the embedding vector corresponding to each character, w1Weight of hidden layer, which is also the connection coefficient of multi-layer perceptron, b1For the bias coefficients of the hidden layer, the f-function may generally adopt sigmoid function or tanh function, as shown below:
sigmoid(x)＝1/(1+e-x)
tanh(x)＝(ex-e-x)/(e1+e-x)
further, after the result output by the hidden layer is processed by a full connection layer, the result output by the full connection layer is input to an output layer, namely the softmax layer of the multilayer sensor, and drug recommendation is performed through the output layer, wherein the obtained recommendation result is as follows:
softmax(W2f(W1x+b1)+b2)
wherein, the W2f(W1+b1)+b2Namely, after the embedded vectors corresponding to the characters are subjected to full connection of the initial drug recommendation model and before the softmax layer is input, the disease information is the second distribution characteristic W of different drug information2As weight coefficients of the output layer, b2And the output layer of the multilayer perceptron is used for outputting recommended medicine information corresponding to the sample disease information, the recommended medicine information is the classified probability of the recommended medicine corresponding to the sample disease information, and the medicine information corresponding to the maximum classified probability is determined as the recommended medicine information corresponding to the sample disease information.
206. And respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristic and the second distribution characteristic, and the actual drug information and the recommended drug information.
For the embodiment of the present invention, in order to enable the drug recommendation model to meet the requirement of drug recommendation accuracy in the actual recommendation process, a first loss function and a second loss function corresponding to the initial drug recommendation model need to be constructed, and for the first loss function and the second loss function that are used for constructing the initial drug recommendation model, step 206 specifically includes: setting a parameter value of a distillation parameter corresponding to the initial drug recommendation model, wherein the parameter value is greater than 1; respectively adjusting the first distribution characteristic and the second distribution characteristic by using the parameter values to obtain an adjusted first distribution characteristic and an adjusted second distribution characteristic; constructing a first loss function corresponding to the initial drug recommendation model based on the adjusted first distribution characteristic and the adjusted second distribution characteristic; and constructing a second loss function corresponding to the initial drug recommendation model based on the actual drug information and the recommended drug information.
For the embodiment of the invention, after a preset BERT model is used for extracting first distribution characteristics of sample disease information aiming at different drug information and second distribution characteristics of the sample disease information aiming at different drug information in the initial drug recommendation model, a first loss function corresponding to the initial drug recommendation model is constructed according to the first distribution characteristics and the second distribution characteristics, and the specific formula is as follows:
wherein L issoftAs a first loss function (distillation loss function), ziA second distribution characteristic, v, for the sample disease information for different drug informationiAiming at the first distribution characteristics of different drug information for sample disease information, in order to avoid over concentration of the output result of the initial drug recommendation model, a distillation parameter T needs to be introduced, and T is used for constructing a first loss function>1，viT is the adjusted first distribution characteristic, ziand/T is the adjusted second distribution characteristic.
Further, according to actual drug information and recommended drug information corresponding to the sample disease information, a second loss function corresponding to the initial drug recommendation model is constructed, and the specific formula is as follows:
wherein L ishardAs a second loss function, CjIs actual drug information corresponding to the sample disease information, mjAnd (4) outputting recommended medicine information for the initial medicine recommendation model, wherein j represents jth medicine information, and N represents the category number of the medicine information. Therefore, the first loss function and the second loss function corresponding to the initial medicine recommendation model can be constructed according to the method.
207. And performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
For the embodiment of the present invention, after the first loss function and the second loss function corresponding to the initial drug recommendation model are constructed, a total loss function corresponding to the initial drug recommendation model needs to be constructed, and based on the total loss function, the initial drug recommendation model is iteratively trained to construct a preset drug recommendation model, and based on this, step 207 specifically includes: determining weight coefficients corresponding to the first loss function and the second loss function respectively; based on the weight coefficient, adding the first loss function and the second loss function to obtain a total loss function corresponding to the initial drug recommendation model; and performing iterative training on the initial drug recommendation model based on the total loss function to construct the preset drug recommendation model.
The weight coefficients may be specifically set according to actual requirements, specifically, the weight coefficients corresponding to a first loss function and a second loss function corresponding to the initial drug recommendation model are respectively determined, and then the product of the first loss function and the corresponding weight coefficient is added to the product of the second loss function and the corresponding weight coefficient to obtain a total loss function corresponding to the initial drug recommendation model, where the specific formula is as follows:
L＝αLsoft+βLhard
after the total loss function corresponding to the initial drug recommendation model is determined, iterative training is performed on the initial drug recommendation model by using the total loss function, parameters in the initial drug recommendation model are updated, and a preset drug recommendation model is obtained.
Further, after obtaining the preset drug recommendation model, in order to prescribe a corresponding drug for the patient according to the information of the patient with the disease, the method further includes: acquiring disease information of a patient to be predicted; inputting the disease information into a preset medicine recommendation model for medicine recommendation to obtain probability values of the disease information for different medicine information, wherein the parameter value of a distillation parameter corresponding to the preset medicine recommendation model is 1; and determining a maximum probability value from the probability values, and determining the medicine information corresponding to the maximum probability value as recommended medicine information corresponding to the patient to be predicted.
Specifically, after a preset drug recommendation model is constructed, the preset drug recommendation model is deployed on a line, and meanwhile, in order to improve the recommendation precision and recommendation efficiency of the preset drug recommendation model and meet the performance requirements of the on-line preset drug recommendation model, the distillation parameter T of the preset drug recommendation model needs to be set to be 1, when actual drug information is recommended to a patient by using the preset drug recommendation model according to the disease information of the patient to be predicted, firstly, a doctor needs to diagnose the disease of the patient to be predicted to obtain the disease information suffered by the patient to be predicted, then, the disease information is input into the preset drug recommendation model to be recommended to obtain the probability values of the disease information corresponding to different drug information, the maximum probability value is selected from the probability values, and the drug information corresponding to the maximum probability value is determined as the recommended drug information corresponding to the patient to be predicted, for example, the disease information of the patient to be predicted is acute enteritis, the acute enteritis is input into a preset medicine recommendation model for medicine recommendation, and the obtained classification probabilities corresponding to the acute enteritis are respectively that the quinolone antibiotics correspond to 50%, the second-generation cephalosporin corresponds to 20%, and the anti-anaerobe antibiotics correspond to 30%, wherein the probability value of 50% is the maximum, so that the quinolone antibiotics corresponding to the probability value are determined as the recommended medicine information corresponding to the patient to be predicted.
According to another drug recommendation model training method provided by the invention, compared with the mode that a doctor recommends a corresponding drug for a patient by using an offline machine learning technology aiming at the disease of the patient, the method provided by the invention acquires sample disease information of the patient and corresponding actual drug information; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; meanwhile, the sample disease information is input into a preset initial drug recommendation model for drug recommendation, and second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information are obtained; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; finally, iterative training is carried out on the initial drug recommendation model based on the first loss function and the second loss function, a preset drug recommendation model is built, so that a first loss function and a second loss function corresponding to the initial drug recommendation model can be built by obtaining a first distribution characteristic and a second distribution characteristic, and the actual drug information and the recommended drug information, and the preset drug recommendation model is built based on the first loss function and the second loss function, so that the preset drug recommendation model has learning capability of a model under a line, meanwhile, the preset drug recommendation model is light in weight and can be deployed on the line, meanwhile, as the preset drug recommendation model is compared with the model under the line, model parameters are reduced, the preset drug recommendation model can reduce the amount of calculation and improve the drug recommendation efficiency while ensuring the drug recommendation precision, effectively assisting the doctor to perform on-line inquiry, thereby improving the diagnosis efficiency of the doctor.
Further, as a specific implementation of fig. 1, an embodiment of the present invention provides a training apparatus for a drug recommendation model, as shown in fig. 3, the apparatus includes: an acquisition unit 31, an extraction unit 32, a recommendation unit 33, a construction unit 34 and a training unit 35.
The acquiring unit 31 may be configured to acquire sample disease information of a patient and corresponding actual drug information.
The extracting unit 32 may be configured to extract a first distribution feature of the sample disease information for different drug information by using a preset natural language model.
The recommending unit 33 may be configured to input the sample disease information into a preset initial drug recommendation model for drug recommendation, so as to obtain a second distribution characteristic of the sample disease information for the different drug information and recommended drug information corresponding to the sample disease information.
The constructing unit 34 may be configured to construct a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristic and the second distribution characteristic, and the actual drug information and the recommended drug information, respectively.
The training unit 35 may be configured to perform iterative training on the initial drug recommendation model based on the first loss function and the second loss function, and construct a preset drug recommendation model.
In a specific application scenario, in order to determine the first distribution characteristics of the sample disease information for different drug information, as shown in fig. 4, the extracting unit 32 includes a first determining module 321, a first extracting module 322, and a classifying module 323.
The first determining module 321 may be configured to determine each character included in the sample disease information and an embedded vector corresponding to each character.
The extracting module 322 may be configured to input the embedded vector into the preset natural language model to perform semantic information extraction, so as to obtain a semantic information vector corresponding to the sample disease information.
The classification module 323 may be configured to input the semantic information vector into a preset classifier for classification, and extract a first distribution feature of the sample disease information with respect to the different drug information.
In a specific application scenario, in order to obtain a semantic information vector corresponding to the sample disease information, the extraction module 322 includes an extraction sub-module and an addition sub-module.
The extraction submodule may be configured to input the embedded vector to the attention layer to perform feature extraction, so as to obtain a first feature vector corresponding to each character.
The adding submodule may be configured to add the first feature vector and the embedded vector to obtain a second feature vector corresponding to each character.
The extraction submodule may be specifically configured to input the second feature vector to the feedforward neural network layer to perform feature extraction, so as to obtain a semantic information vector corresponding to the sample disease information.
In a specific application scenario, in order to obtain the second distribution characteristics of the sample disease information for different drug information and the recommended drug information corresponding to the sample disease information, the recommending unit 33 includes a second extracting module 331 and a third extracting module 332.
The second extraction module 331 may be configured to input the sample disease information into the multilayer sensor, extract a feature output by a last full connection layer in the multilayer sensor, and determine the feature output by the last full connection layer as a second distribution feature of the sample disease information for the different medicine information.
The third extracting module 332 may be configured to input the feature output by the last full connection layer into a softmax layer in the multilayer sensor, so as to obtain recommended drug information corresponding to the sample disease information.
In a specific application scenario, in order to construct a first loss function and a second loss function corresponding to an initial drug recommendation model, the constructing unit 34 includes a setting module 341, an adjusting module 342, a first constructing module 343, and a second constructing module 344.
The setting module 341 may be configured to set a parameter value of a distillation parameter corresponding to the initial drug recommendation model, where the parameter value is greater than 1.
The adjusting module 342 may be configured to adjust the first distribution characteristic and the second distribution characteristic respectively by using the parameter values, so as to obtain an adjusted first distribution characteristic and an adjusted second distribution characteristic.
The first constructing module 343 may be configured to construct, based on the adjusted first distribution characteristic and the adjusted second distribution characteristic, a first loss function corresponding to the initial drug recommendation model.
The second constructing module 344 may be configured to construct a second loss function corresponding to the initial drug recommendation model based on the actual drug information and the recommended drug information.
In a specific application scenario, in order to construct a preset drug recommendation model, the training unit 35 includes a second determining module 351, an adding module 352, and a training module 353.
The second determining module 351 may be configured to determine weighting coefficients corresponding to the first loss function and the second loss function, respectively.
The adding module 352 may be configured to add the first loss function and the second loss function based on the weight coefficient to obtain a total loss function corresponding to the initial drug recommendation model.
The training module 353 may be configured to perform iterative training on the initial drug recommendation model based on the total loss function, and construct the preset drug recommendation model.
In a specific application scenario, in order to utilize the established preset drug recommendation model to perform drug recommendation, the device further includes: a determination unit 36.
The obtaining unit 31 may be further configured to obtain disease information of a patient to be predicted.
The recommending unit 33 may be further configured to input the disease information into a preset drug recommending model for drug recommendation, so as to obtain probability values of the disease information for the different drug information, where a parameter value of a distillation parameter corresponding to the preset drug recommending model is 1.
The determining unit 36 may be configured to determine a maximum probability value from the probability values, and determine the medicine information corresponding to the maximum probability value as recommended medicine information corresponding to the patient to be predicted.
It should be noted that other corresponding descriptions of the functional modules involved in the training apparatus for a drug recommendation model provided in the embodiment of the present invention may refer to the corresponding descriptions of the method shown in fig. 1, and are not described herein again.
Based on the method shown in fig. 1, correspondingly, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the following steps: acquiring sample disease information of a patient and corresponding actual medicine information; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
Based on the above embodiments of the method shown in fig. 1 and the apparatus shown in fig. 3, an embodiment of the present invention further provides an entity structure diagram of a computer device, as shown in fig. 5, where the computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and executable on the processor, wherein the memory 42 and the processor 41 are both arranged on a bus 43 such that when the processor 41 executes the program, the following steps are performed: acquiring sample disease information of a patient and corresponding actual medicine information; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
According to the technical scheme, the sample disease information of the patient and the corresponding actual medicine information are obtained; extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model; meanwhile, the sample disease information is input into a preset initial drug recommendation model for drug recommendation, and second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information are obtained; respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information; finally, iterative training is carried out on the initial drug recommendation model based on the first loss function and the second loss function, a preset drug recommendation model is built, so that a first loss function and a second loss function corresponding to the initial drug recommendation model can be built by obtaining a first distribution characteristic and a second distribution characteristic, and the actual drug information and the recommended drug information, and the preset drug recommendation model is built based on the first loss function and the second loss function, so that the preset drug recommendation model has learning capability of a model under a line, meanwhile, the preset drug recommendation model is light in weight and can be deployed on the line, meanwhile, as the preset drug recommendation model is compared with the model under the line, model parameters are reduced, the preset drug recommendation model can reduce the amount of calculation and improve the drug recommendation efficiency while ensuring the drug recommendation precision, effectively assisting the doctor to perform on-line inquiry, thereby improving the diagnosis efficiency of the doctor.
It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A training method of a medicine recommendation model is characterized by comprising the following steps:
acquiring sample disease information of a patient and corresponding actual medicine information;
extracting a first distribution characteristic of the sample disease information aiming at different medicine information by using a preset natural language model;
inputting the sample disease information into a preset initial drug recommendation model for drug recommendation, and obtaining a second distribution characteristic of the sample disease information for different drug information and recommended drug information corresponding to the sample disease information;
respectively constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information;
and performing iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
2. The method of claim 1, wherein the extracting the first distribution feature of the sample disease information for different drug information by using a preset natural language model comprises:
determining each character contained in the sample disease information and an embedded vector corresponding to each character;
inputting the embedded vector into the preset natural language model for semantic information extraction to obtain a semantic information vector corresponding to the sample disease information;
and inputting the semantic information vector into a preset classifier for classification, and extracting a first distribution characteristic of the sample disease information aiming at different medicine information.
3. The method according to claim 2, wherein the preset natural language model is a preset BERT model, the preset BERT model includes an attention layer and a feedforward neural network layer, and the inputting the embedded vector into the preset natural language model for semantic information extraction to obtain a semantic information vector corresponding to the sample disease information includes:
inputting the embedded vector to the attention layer for feature extraction to obtain a first feature vector corresponding to each character;
adding the first feature vector and the embedded vector to obtain a second feature vector corresponding to each character;
and inputting the second feature vector to the feedforward neural network layer for feature extraction to obtain a semantic information vector corresponding to the sample disease information.
4. The method according to claim 1, wherein the initial drug recommendation model is a multilayer sensor, and the step of inputting the sample disease information into a preset initial drug recommendation model for drug recommendation to obtain a second distribution characteristic of the sample disease information for the different drug information and recommended drug information corresponding to the sample disease information comprises:
inputting the sample disease information into the multilayer perceptron, extracting the characteristics output by the last full-connection layer in the multilayer perceptron, and determining the characteristics output by the last full-connection layer as second distribution characteristics of the sample disease information aiming at different medicine information;
and inputting the characteristics output by the last full-connection layer into a softmax layer in the multilayer perceptron to obtain recommended medicine information corresponding to the sample disease information.
5. The method of claim 1, wherein the constructing a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristics and the second distribution characteristics, and the actual drug information and the recommended drug information, respectively, comprises:
setting a parameter value of a distillation parameter corresponding to the initial drug recommendation model, wherein the parameter value is greater than 1;
respectively adjusting the first distribution characteristic and the second distribution characteristic by using the parameter values to obtain an adjusted first distribution characteristic and an adjusted second distribution characteristic;
constructing a first loss function corresponding to the initial drug recommendation model based on the adjusted first distribution characteristic and the adjusted second distribution characteristic;
and constructing a second loss function corresponding to the initial drug recommendation model based on the actual drug information and the recommended drug information.
6. The method of claim 1, wherein iteratively training the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model comprises:
determining weight coefficients corresponding to the first loss function and the second loss function respectively;
based on the weight coefficient, adding the first loss function and the second loss function to obtain a total loss function corresponding to the initial drug recommendation model;
and performing iterative training on the initial drug recommendation model based on the total loss function to construct the preset drug recommendation model.
7. The method of any of claims 1-6, wherein after iteratively training the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model, the method further comprises:
acquiring disease information of a patient to be predicted;
inputting the disease information into a preset medicine recommendation model for medicine recommendation to obtain probability values of the disease information for different medicine information, wherein the parameter value of a distillation parameter corresponding to the preset medicine recommendation model is 1;
and determining a maximum probability value from the probability values, and determining the medicine information corresponding to the maximum probability value as recommended medicine information corresponding to the patient to be predicted.
8. A training device for a drug recommendation model, comprising:
the system comprises an acquisition unit, a processing unit and a display unit, wherein the acquisition unit is used for acquiring sample disease information of a patient and corresponding actual medicine information;
the extraction unit is used for extracting first distribution characteristics of the sample disease information aiming at different medicine information by using a preset natural language model;
the recommending unit is used for inputting the sample disease information into a preset initial drug recommending model for drug recommending to obtain second distribution characteristics of the sample disease information aiming at different drug information and recommended drug information corresponding to the sample disease information;
a building unit, configured to respectively build a first loss function and a second loss function corresponding to the initial drug recommendation model based on the first distribution characteristic and the second distribution characteristic, and the actual drug information and the recommended drug information;
and the training unit is used for carrying out iterative training on the initial drug recommendation model based on the first loss function and the second loss function to construct a preset drug recommendation model.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
10. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program realizes the steps of the method of any one of claims 1 to 7 when executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111017415.6A CN113569999A (en) | 2021-08-31 | 2021-08-31 | Training method and device for medicine recommendation model, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111017415.6A CN113569999A (en) | 2021-08-31 | 2021-08-31 | Training method and device for medicine recommendation model, storage medium and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113569999A true CN113569999A (en) | 2021-10-29 |
Family
ID=78173403
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111017415.6A Pending CN113569999A (en) | 2021-08-31 | 2021-08-31 | Training method and device for medicine recommendation model, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113569999A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108906A (en) * | 2023-04-06 | 2023-05-12 | 北京亚信数据有限公司 | Disease drug relation mapping model training and related recommendation and detection methods and devices |
-
2021
- 2021-08-31 CN CN202111017415.6A patent/CN113569999A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108906A (en) * | 2023-04-06 | 2023-05-12 | 北京亚信数据有限公司 | Disease drug relation mapping model training and related recommendation and detection methods and devices |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490251B (en) | Artificial intelligence-based prediction classification model obtaining method and device and storage medium | |
| US11461554B2 (en) | Semantic classification of numerical data in natural language context based on machine learning | |
| CN109637669B (en) | Deep learning-based treatment scheme generation method, device and storage medium | |
| Zhang et al. | Artificial intelligence–based traditional Chinese medicine assistive diagnostic system: validation study | |
| CN110297908A (en) | Diagnosis and treatment program prediction method and device | |
| WO2020228636A1 (en) | Training method and apparatus, dialogue processing method and system, and medium | |
| US20210406687A1 (en) | Method for predicting attribute of target object based on machine learning and related device | |
| CN112100406A (en) | Data processing method, device, equipment and medium | |
| US11977972B2 (en) | Residual semi-recurrent neural networks | |
| CN116364299B (en) | Disease diagnosis and treatment path clustering method and system based on heterogeneous information network | |
| CN113707307A (en) | Disease analysis method and device, electronic equipment and storage medium | |
| JP2022500713A (en) | Machine-assisted dialogue system, as well as medical condition inquiry device and its method | |
| CN111199797A (en) | Auxiliary diagnosis model establishing and auxiliary diagnosis method and device | |
| CN117218498B (en) | Multi-modal large language model training method and system based on multi-modal encoder | |
| Peng et al. | Heart disease prediction using artificial neural networks: a survey | |
| CN114783603A (en) | Multi-source graph neural network fusion-based disease risk prediction method and system | |
| CN113569999A (en) | Training method and device for medicine recommendation model, storage medium and computer equipment | |
| Ma et al. | Construction and evaluation of intelligent medical diagnosis model based on integrated deep neural network | |
| Liu et al. | Evolving deep convolutional neural networks by IP-based marine predator algorithm for COVID-19 diagnosis using chest CT scans | |
| CN114420279A (en) | Medical resource recommendation method, device, equipment and storage medium | |
| CN112216379A (en) | Disease diagnosis system based on intelligent joint learning | |
| CN115631852B (en) | Certificate type recommendation method and device, electronic equipment and nonvolatile storage medium | |
| Valle-Lisboa et al. | Elman topology with sigma–pi units: An application to the modeling of verbal hallucinations in schizophrenia | |
| CN116994695A (en) | Training method, device, equipment and storage medium of report generation model | |
| CN114627993A (en) | Information prediction method, information prediction device, storage medium and computer equipment |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220601 Address after: 518000 China Aviation Center 2901, No. 1018, Huafu Road, Huahang community, Huaqiang North Street, Futian District, Shenzhen, Guangdong Province Applicant after: Shenzhen Ping An medical and Health Technology Service Co.,Ltd. Address before: Room 12G, Area H, 666 Beijing East Road, Huangpu District, Shanghai 200001 Applicant before: PING AN MEDICAL AND HEALTHCARE MANAGEMENT Co.,Ltd. |