CN116307272A - Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning - Google Patents
Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning Download PDFInfo
- Publication number
- CN116307272A CN116307272A CN202310551325.8A CN202310551325A CN116307272A CN 116307272 A CN116307272 A CN 116307272A CN 202310551325 A CN202310551325 A CN 202310551325A CN 116307272 A CN116307272 A CN 116307272A
- Authority
- CN
- China
- Prior art keywords
- appointment
- data
- missing
- training
- outpatient
- 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
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000013135 deep learning Methods 0.000 title claims abstract description 26
- 230000036962 time dependent Effects 0.000 claims abstract description 20
- 238000003745 diagnosis Methods 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 70
- 238000012360 testing method Methods 0.000 claims description 41
- 238000003672 processing method Methods 0.000 claims description 18
- 238000012795 verification Methods 0.000 claims description 13
- 238000011156 evaluation Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 11
- 238000010200 validation analysis Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 238000002372 labelling Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 4
- 230000001419 dependent effect Effects 0.000 description 3
- 238000007477 logistic regression Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/02—Reservations, e.g. for tickets, services or events
-
- 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
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Mathematical Physics (AREA)
- Quality & Reliability (AREA)
- Mathematical Analysis (AREA)
- Health & Medical Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Marketing (AREA)
- Medical Informatics (AREA)
- Computational Mathematics (AREA)
- Development Economics (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Primary Health Care (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Public Health (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Probability & Statistics with Applications (AREA)
- Biomedical Technology (AREA)
Abstract
The application relates to the technical field of deep learning, in particular to a pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning, wherein the method comprises the following steps: receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic values include at least: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant; and inputting the outpatient diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result. Compared with the prior art, the technical scheme in the application includes time dependency features in the feature value information for predicting whether to be refreshing, and the time dependency features comprise at least one of the following features: the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition.
Description
Technical Field
The application relates to the technical field of deep learning, in particular to a pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning.
Background
The change of the traditional medical service mode and the optimization of the medical service flow become important aspects for solving the current medical and health service problems. Compared with adults, the pediatric department has the specificity, and along with the development of science and technology, the Internet plus medical clinic gradually becomes an important measure for solving the problem of uneven distribution of pediatric medical resources among areas, but the online reservation and the refreshing rate of the pediatric clinic based on the Internet plus is 2.65 times of that of offline reservation. This not only causes serious waste of pediatric medical resources, but also may delay diagnosis and treatment of the infant disease.
The prior art 'internet +' outpatient based refreshing model has few researches and poor accuracy, and does not relate to the pediatric field.
Disclosure of Invention
In order to overcome the problems that the related art is less in study and poor in accuracy based on a refreshing model of an 'Internet+' outpatient and does not relate to the pediatric field at least to a certain extent, the application provides a pediatric Internet+outpatient refreshing prediction method and equipment based on deep learning.
The scheme of the application is as follows:
according to a first aspect of embodiments of the present application, there is provided a pediatric internet+outpatient visit prediction method based on deep learning, including:
receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic value at least comprises: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant;
inputting the outpatient diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result;
wherein the time-dependent features include at least one of the following features:
the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition.
Preferably, the method further comprises:
acquiring sample data from a plurality of data sources;
dividing the sample data into a training data set, an internal verification data set and an external verification data set according to sources;
extracting the characteristic value of the training data set, wherein the extracted characteristic value at least comprises: social demographics of children patients, clinical reservation characteristics, reservation time characteristics, time dependence characteristics and refreshing conditions;
carrying out one-hot treatment on the classified variables in all the characteristic values, and carrying out standardization treatment on the continuous variables;
dividing the training data set into a plurality of equal parts, alternately taking 1 part of the training data set as test data and the other parts as training data, training the deep network learning model according to the training data, and obtaining a model evaluation result according to the test data.
Preferably, the method further comprises:
and determining the precision of the deep network learning model according to the average value of the model evaluation result obtained by each training.
Preferably, after obtaining the sample data from the plurality of data sources, the method further comprises:
performing missing value processing on the sample data; the method for processing the missing value is at least one of the following methods: automatic missing data processing method, non-automatic missing data processing method and extra labeling strategy.
Preferably, the automatic missing data processing method includes: the missing data is not processed and filled, and the missing data is automatically processed through the deep network learning model;
the non-automatic missing data processing method comprises the following steps: discarding strategies, mean interpolation strategies, forward/backward interpolation strategies, regression interpolation strategies, and chained equation multiple interpolation strategies;
wherein the discard policy comprises: deleting missing data, and dividing the training data set and the verification data set into two types including missing data and non-missing data according to whether missing values exist or not; the validation data set includes the internal validation data set and the external validation data set;
the form of training and testing according to the discard strategy includes:
training in a training data set containing missing data and testing in a test data set containing missing data;
training in a training data set containing missing data and testing in a test data set not containing missing data;
training in a training data set which does not contain missing data, and testing in a test data set which contains missing data;
training in a training data set which does not contain missing data, and testing in a test data set which does not contain missing data;
the mean value interpolation strategy comprises the following steps: filling the missing data with the mean or mode of the variables;
the forward/backward interpolation strategy includes: performing leak repairing treatment by adopting clinic data of adjacent dates of the child patient;
the regression interpolation strategy includes: establishing a regression equation based on the sample data, substituting a known variable value of the object with the missing value into the regression equation to solve the missing value;
the chain equation multiple interpolation includes: assuming that all missing data are randomly lost, the properties of the missing data are related to non-missing data, and performing conditional modeling interpolation on the missing value of each missing data according to the non-missing data by running a plurality of regression models;
the extra labeling strategy comprises the following steps: a variable is added to the sample data to indicate whether the value of the sample data is missing.
Preferably, the method further comprises:
constructing the deep network learning model in a preset environment, and determining the rear end, the loss function and the gradient descent method of the deep network learning model;
the deep network learning model comprises at least 4 hidden layers and 1 output layer; the activation functions of the hidden layer and the output layer are different;
the deep network learning model is trained based on preset iteration times, and early stop is set.
Preferably, the evaluation of the deep network learning model includes at least one of the following metrics:
confusion matrix, accuracy, recall, precision, F1 value, AUCROC (Area under the Curve of ROC, area enclosed by ROC curve and coordinate axis), AUPRC (Area under the Curve of PRC, area enclosed by PRC curve and coordinate axis), and Cohen Kappa Score.
Preferably, the method further comprises:
determining importance scores of the extracted feature values in the training process of the deep network learning model;
calculating an OR (odds ratio) value and an adjusted P value of each characteristic value through a regression model, and drawing a volcanic chart according to the OR value and the adjusted P value of each characteristic value to determine operability characteristics in each characteristic value;
the influence size of each operability feature is determined according to the importance score of each operability feature.
Preferably, the social demographic characteristics of the infant include at least: age, sex, source and ethnicity of the infant;
the clinical appointment characteristics include at least: reserved departments, reserved doctors, and doctor types;
the reserved time feature at least comprises: the appointment occurrence time, the appointment visit duration, and the difference between the appointment occurrence time and the appointment visit time.
According to a second aspect of embodiments of the present application, there is provided a pediatric internet+outpatient visit forecast apparatus based on deep learning, comprising:
a processor and a memory;
the processor is connected with the memory through a communication bus:
the processor is used for calling and executing the program stored in the memory;
the memory is used for storing a program at least for executing the pediatric internet+outpatient and refreshing prediction method based on deep learning.
The technical scheme that this application provided can include following beneficial effect: the pediatric Internet+outpatient and refreshing prediction method based on deep learning comprises the following steps of: receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic values include at least: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant; and inputting the outpatient diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result. Compared with the prior art, the technical scheme in the application includes time dependency features in the feature value information for predicting whether to be refreshing, and the time dependency features comprise at least one of the following features: the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition. One of the important characteristics of the refreshing of the outpatient is that the time-dependent characteristics of the patient exist, namely, the future refreshing of the patient who is frequently refreshed in the past is more likely, and the prediction accuracy of the model is improved by incorporating the time-dependent characteristics into the deep network learning model.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
Fig. 1 is a schematic flow chart of a pediatric internet+outpatient prediction method based on deep learning according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a training process of a deep web learning model according to an embodiment of the present application;
FIG. 3 is an exemplary diagram of a missing value processing method according to one embodiment of the present application;
FIG. 4 is a schematic diagram of a specific architecture of a deep web learning model according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a pediatric internet+outpatient and refreshing prediction apparatus based on deep learning according to an embodiment of the present application.
Reference numerals: a processor-31; and a memory-32.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as detailed in the accompanying claims.
Example 1
Fig. 1 is a flow chart of a pediatric internet+outpatient and refreshing prediction method based on deep learning according to an embodiment of the present application, referring to fig. 1, a pediatric internet+outpatient and refreshing prediction method based on deep learning includes:
s11: receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic values include at least: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant;
s12: inputting the clinic diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result;
wherein the time-dependent features include at least one of the following features:
the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition.
It should be noted that, the technical scheme in this embodiment relates to the technical field of deep learning, and in particular relates to an application of the deep learning technology in pediatric internet+outpatient and refreshing prediction.
It should be noted that, the current outpatient refreshing prediction model has certain limitations:
the accuracy of the existing outpatient refreshing prediction model is to be improved. The most common of the existing patient refreshing prediction models is to adopt a regression model, and then a tree model, a Bayesian model and an integrated learning algorithm. Among the above prediction models, only one prediction model has an AUCROC greater than 0.9, and the other prediction models have an AUCROC lower than 0.85, so that the accuracy of the existing outpatient contract prediction model is still not ideal as a whole.
The existing model is still to be promoted for screening time-dependent features in the outpatient refreshing prediction model. One of the important characteristics of outpatient refreshing patients is that there is a greater likelihood that the patient will have a time dependent characteristic, i.e., future refreshing of patients who were in frequent refreshing.
In specific practice, the model prediction effect of the time-dependent characteristics, which does not contain the previous online/offline refreshing rate and the online/offline refreshing condition of the last appointment visit, is the worst; the model prediction effect of the online/offline refreshing condition of the last appointment visit is slightly improved in the time dependency characteristics; the prediction effect of the model with the 'past online up/down refreshing rate' is obviously improved in the time dependency characteristics; the model prediction effect is optimal in the time dependency characteristics, which comprise both the previous online/offline appointment rate and the online/offline appointment condition of the last appointment visit.
It will be appreciated that the above practices may illustrate that the time-dependent features are of great significance to model predictive effect optimization.
It can be appreciated that the pediatric internet+outpatient and refreshing prediction method based on deep learning in this embodiment includes: receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic values include at least: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant; and inputting the outpatient diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result. Compared with the prior art, the technical solution in this embodiment includes a time-dependent feature in the feature value information for predicting whether to get refreshing, where the time-dependent feature includes at least one of the following features: the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition. One of the important features of the outpatient visit is that there is a time-dependent feature of the patient, that is, the future visit of the patient who is frequently refreshed in the past is more likely, and the prediction accuracy of the model is improved by incorporating the time-dependent feature into the deep network learning model.
Example two
The training process of the deep web learning model, referring to fig. 2, includes:
s21: acquiring sample data from a plurality of data sources;
s22: dividing the sample data into a training data set, an internal verification data set and an external verification data set according to the source;
s23: the training data set is subjected to characteristic value extraction, and the extracted characteristic values at least comprise: social demographics of children patients, clinical reservation characteristics, reservation time characteristics, time dependence characteristics and refreshing conditions;
s24: carrying out one-hot treatment on the classified variables in all the characteristic values, and carrying out standardization treatment on the continuous variables;
s25: dividing the training data set into a plurality of equal parts, taking 1 part of the training data set as test data and the other parts of the training data set as training data in turn, training a deep network learning model according to the training data, and obtaining a model evaluation result according to the test data.
In specific practice, reservation registration data of different hospitals or women and young health care hospitals for self-starting pediatric 'Internet+' outpatient diagnosis and treatment to date are collected as sample data.
The prediction efficiency of the model is also affected due to the off-line outpatient conditions of the pediatric "internet+" outpatient users. Therefore, in this embodiment, the "internet+" outpatient user's contemporaneous "offline" line outpatient appointment registration data is also collected, so as to form "offline" outpatient data, which is also used as sample data.
Sample data from one source is divided into a training data set and an internal verification data set, and sample data from other sources is used as an external verification data set, so that the external prediction performance of the sample data can be verified through the external verification data set.
It should be noted that, the feature values extracted in this embodiment at least include: social demographics of the infant, clinical appointment characteristics, appointment time characteristics, time-dependent characteristics and appointment conditions.
Wherein, the social demographic characteristics of the infant at least comprise: child age, gender, origin (province/city/district/county) and ethnicity;
the clinical appointment characteristics include at least: reserved departments, reserved doctors, and doctor types;
the reservation time feature includes at least: appointment occurrence time (year/month/day, whether or not it is holiday), appointment visit duration, difference between appointment occurrence time and appointment visit time.
The refreshing condition is whether the reservation is refreshing or not.
It should be noted that, in this embodiment, a ten-fold cross-validation method is adopted to divide the training data set into 10 equal parts, 1 part of the data is used as test data in turn, and the other 9 parts of the data are used as training data, and the deep network learning model is trained according to the training data. And obtaining a corresponding model evaluation result in each training process, and taking the average value of the evaluation results of 10 times as the estimation of algorithm precision, namely the precision of the deep network learning model.
In specific practice, model type screening is also performed: based on sample data, constructing a plurality of prediction models such as a depth network, a random forest, a regression model, a support vector machine, a Bayesian model, an Xgboost model and the like, then obtaining the prediction precision of each prediction model by using a ten-fold cross verification method, comprehensively screening model types which are most suitable for predicting the refreshing of pediatric 'Internet+' outpatients, and verifying the external prediction efficiency through an external data set. The finally determined prediction model is a deep network learning model.
It should be noted that, the evaluation of the deep network learning model includes at least one of the following indexes:
confusion matrix, accuracy, recall, precision, F1 value, AUCROC, AUPRC, and Cohen Kappa Score.
Example III
It should be noted that the prediction model in the prior art does not evaluate the influence of missing data on model prediction. Due to various factors such as patient privacy considerations, low willingness of patients to share information, and non-mandatory appointment information filling of hospital outpatient information systems, missing data is very common in outpatient appointment systems. Therefore, in model construction, whether to be based on the complete data set containing missing data or the complete data set containing non-missing data, and how to optimize interpolation of existing missing data, are all issues that should be considered in predictive model construction and optimization.
Based on this, the pediatric internet+outpatient contract prediction method based on deep learning in this embodiment, after acquiring sample data through a plurality of data sources, further includes:
carrying out missing value processing on the sample data; the method for processing the missing value is at least one of the following methods: automatic missing data processing method, non-automatic missing data processing method and extra labeling strategy.
FIG. 3 is an exemplary diagram of a missing value processing method, and referring to FIG. 3, in particular:
the automatic missing data processing method comprises the following steps: the missing data is not processed and filled, and the missing data is automatically processed through a deep network learning model;
the non-automatic missing data processing method comprises the following steps: discarding strategies, mean interpolation strategies, forward/backward interpolation strategies, regression interpolation strategies, and chained equation multiple interpolation strategies;
wherein the discard policy comprises: deleting missing data, and dividing the training data set and the verification data set into two types including missing data and non-missing data according to whether missing values exist or not; the validation data set includes an internal validation data set and an external validation data set;
the form of training and testing according to the discard strategy includes:
training in a training data set containing missing data and testing in a test data set containing missing data;
training in a training data set containing missing data and testing in a test data set not containing missing data;
training in a training data set which does not contain missing data, and testing in a test data set which contains missing data;
training in a training data set which does not contain missing data, and testing in a test data set which does not contain missing data;
the mean value interpolation strategy comprises the following steps: filling the missing data with the mean or mode of the variables;
the forward/backward interpolation strategy includes: performing leak repairing treatment by adopting clinic data of adjacent dates of the child patient;
the regression interpolation strategy includes: establishing a regression equation based on the sample data, substituting a known variable value of the object with the missing value into the regression equation to solve the missing value;
the chain equation multiple interpolation includes: assuming that all missing data are randomly lost, the properties of the missing data are related to non-missing data, and performing conditional modeling interpolation on the missing value of each missing data according to the non-missing data by running a plurality of regression models;
the labeling strategy comprises the following steps: a variable is added to the sample data to indicate whether the value of the sample data is missing.
In specific practice, the preferred missing data processing method is a discarding strategy in a non-automatic missing data processing method, and because training and testing according to the discarding strategy includes four forms, in this embodiment, a ten-fold cross-validation method is adopted by applying a deep network learning model, pediatric 'internet+' outpatient refreshing model prediction is performed through a training set and a testing set of the four forms, AUROC and AUPRC are used as model prediction effect evaluation indexes, and the obtained result is that the prediction performance of the deep network learning model is best when training is performed on the training data set containing missing data and testing is performed on the testing data set containing missing data.
Therefore, in this embodiment, the discard policy in the non-automatic missing data processing method is used as the missing data optimization processing method, and when the discard policy is executed, training is performed on the training data set containing the missing data, and testing is performed on the test data set containing the missing data.
Example IV
It should be noted that the method further includes:
constructing a deep network learning model in a preset environment, and determining the rear end, a loss function and a gradient descent method of the deep network learning model;
the deep network learning model comprises at least 4 hidden layers and 1 output layer; the activation functions of the hidden layer and the output layer are different;
the deep network learning model is trained based on preset iteration times, and early stop is set.
In specific practice, a deep network learning model is built in a Keras 2.9.0 environment, and a gradient descent method with TensorFlow2.9.1 as a back end, cross-entcopy as a loss function and Adam as a default setting is used. The deep network learning model consists of 4 hidden layers, and the output layer has 1 neuro. The activation functions for the hidden layer and the output layer are "ReLu" and "Sigmoid", respectively. A specific architecture of the deep web learning model is shown in fig. 4.
The iteration number of the deep network learning model can be set to be 50, and early stop is carried out when the 12 th iteration is set, so that the model is prevented from being over fitted.
Example five
It should be noted that the method further includes:
determining importance scores of the extracted feature values in the training process of the deep network learning model;
calculating an OR value and an adjusted P value of each characteristic value through a regression model, and drawing a volcanic chart according to the OR value and the adjusted P value of each characteristic value to determine operability characteristics in each characteristic value;
the influence size of each operability feature is determined according to the importance score of each operability feature.
In specific practice, model feature importance scores are calculated by deep (Deep Learning Important Fea Tures, important features for deep learning):
the method for calculating the feature importance scores of the model by deep is a conventional technical means in the prior art, and is not described herein.
In this embodiment, the operability feature may include: scheduled visit time, scheduled doctor, scheduled visit duration, etc. These factors are internal management activities that hospitals can take through scheduling, system settings, etc., thereby reducing the patient's cool down operability characteristics.
In this embodiment, the operability features are further sorted based on the importance scores of the existing feature values: and (3) combining the logistic regression model to calculate the OR value of the characteristic value and the adjusted P value, and drawing a volcanic chart to determine the operability characteristic in each characteristic value.
In the logistic regression analysis, both the P value and the OR value are important indicators. The P-value is an indicator used to test statistical assumptions, indicating whether the fit of the model is significant at a significant level (typically 0.05). And the OR value is an index that measures the association between two factors and is used to represent the strength and direction of the relationship between the dependent and independent variables. In performing logistic regression analysis, it is necessary to first pay attention to whether the P value is significant. If the P-value is less than a significant level (typically 0.05), then the fitting of the model is interpreted as significant, i.e., the effect of the independent variable on the dependent variable is statistically significant. In this case, further attention may be paid to the OR value to understand the specific effect of the independent variable on the dependent variable.
It can be understood that the influence of each operability feature is determined according to the importance scores of the operability features, and further, an improved item is provided according to the influence of each operability feature, so that the hospital is assisted in making relevant decisions, the occurrence probability of the refreshing behavior of the patient is reduced, the running efficiency of the outpatient service of the hospital is improved, and the social benefit of the hospital is improved.
Example six
Pediatric internet+outpatient and refreshing prediction equipment based on deep learning, comprising:
a processor 31 and a memory 32;
the processor 31 is connected to the memory 32 via a communication bus:
wherein the processor 31 is used for calling and executing the program stored in the memory 32;
the memory 32 is configured to store a program, where the program is configured to perform at least one of the pediatric internet+outpatient prediction methods based on deep learning in any of the above embodiments.
It is to be understood that the same or similar parts in the above embodiments may be referred to each other, and that in some embodiments, the same or similar parts in other embodiments may be referred to.
It should be noted that in the description of the present application, the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Furthermore, in the description of the present application, unless otherwise indicated, the meaning of "plurality" means at least two.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and further implementations are included within the scope of the preferred embodiment of the present application in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the embodiments of the present application.
It is to be understood that portions of the present application may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
Those of ordinary skill in the art will appreciate that all or a portion of the steps carried out in the method of the above-described embodiments may be implemented by a program to instruct related hardware, where the program may be stored in a computer readable storage medium, and where the program, when executed, includes one or a combination of the steps of the method embodiments.
In addition, each functional unit in each embodiment of the present application may be integrated in one processing module, or each unit may exist alone physically, or two or more units may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules. The integrated modules may also be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product.
The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, or the like.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Although embodiments of the present application have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the application, and that variations, modifications, alternatives, and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the application.
Claims (10)
1. The pediatric Internet+outpatient and refreshing prediction method based on deep learning is characterized by comprising the following steps of:
receiving an outpatient appointment request, and determining outpatient diagnosis and treatment characteristic value information according to the outpatient appointment request; the characteristic value at least comprises: social demographics, clinical appointment characteristics, appointment time characteristics, and time-dependent characteristics of the infant;
inputting the outpatient diagnosis and treatment characteristic value information into a pre-trained deep network learning model to obtain a refreshing prediction result;
wherein the time-dependent features include at least one of the following features:
the online/down appointment rate, the online/down appointment advance cancellation time, the online/down clinic arrival rate, the online/down appointment condition of last appointment, the previous appointment advance cancellation time and the online/down clinic arrival condition.
2. The method according to claim 1, wherein the method further comprises:
acquiring sample data from a plurality of data sources;
dividing the sample data into a training data set, an internal verification data set and an external verification data set according to sources;
extracting the characteristic value of the training data set, wherein the extracted characteristic value at least comprises: social demographics of children patients, clinical reservation characteristics, reservation time characteristics, time dependence characteristics and refreshing conditions;
carrying out one-hot treatment on the classified variables in all the characteristic values, and carrying out standardization treatment on the continuous variables;
dividing the training data set into a plurality of equal parts, alternately taking 1 part of the training data set as test data and the other parts as training data, training the deep network learning model according to the training data, and obtaining a model evaluation result according to the test data.
3. The method according to claim 2, wherein the method further comprises:
and determining the precision of the deep network learning model according to the average value of the model evaluation result obtained by each training.
4. The method of claim 2, wherein after obtaining the sample data from the plurality of data sources, the method further comprises:
performing missing value processing on the sample data; the method for processing the missing value is at least one of the following methods: automatic missing data processing method, non-automatic missing data processing method and extra labeling strategy.
5. The method of claim 4, wherein the automatic missing data processing method comprises: the missing data is not processed and filled, and the missing data is automatically processed through the deep network learning model;
the non-automatic missing data processing method comprises the following steps: discarding strategies, mean interpolation strategies, forward/backward interpolation strategies, regression interpolation strategies, and chained equation multiple interpolation strategies;
wherein the discard policy comprises: deleting missing data, and dividing the training data set and the verification data set into two types including missing data and non-missing data according to whether missing values exist or not; the validation data set includes the internal validation data set and the external validation data set;
the form of training and testing according to the discard strategy includes:
training in a training data set containing missing data and testing in a test data set containing missing data;
training in a training data set containing missing data and testing in a test data set not containing missing data;
training in a training data set which does not contain missing data, and testing in a test data set which contains missing data;
training in a training data set which does not contain missing data, and testing in a test data set which does not contain missing data;
the mean value interpolation strategy comprises the following steps: filling the missing data with the mean or mode of the variables;
the forward/backward interpolation strategy includes: performing leak repairing treatment by adopting clinic data of adjacent dates of the child patient;
the regression interpolation strategy includes: establishing a regression equation based on the sample data, substituting a known variable value of the object with the missing value into the regression equation to solve the missing value;
the chain equation multiple interpolation includes: assuming that all missing data are randomly lost, the properties of the missing data are related to non-missing data, and performing conditional modeling interpolation on the missing value of each missing data according to the non-missing data by running a plurality of regression models;
the extra labeling strategy comprises the following steps: a variable is added to the sample data to indicate whether the value of the sample data is missing.
6. The method according to claim 2, wherein the method further comprises:
constructing the deep network learning model in a preset environment, and determining the rear end, the loss function and the gradient descent method of the deep network learning model;
the deep network learning model comprises at least 4 hidden layers and 1 output layer; the activation functions of the hidden layer and the output layer are different;
the deep network learning model is trained based on preset iteration times, and early stop is set.
7. The method of claim 2, wherein the evaluation of the deep web learning model comprises at least one of the following metrics:
confusion matrix, accuracy, recall, precision, F1 value, AUCROC, AUPRC, and Cohen Kappa Score.
8. The method according to claim 2, wherein the method further comprises:
determining importance scores of the extracted feature values in the training process of the deep network learning model;
calculating an OR value and an adjusted P value of each characteristic value through a regression model, and drawing a volcanic chart according to the OR value and the adjusted P value of each characteristic value to determine operability characteristics in each characteristic value;
the influence size of each operability feature is determined according to the importance score of each operability feature.
9. The method according to claim 2, wherein the social demographic characteristics of the infant patient include at least: age, sex, source and ethnicity of the infant;
the clinical appointment characteristics include at least: reserved departments, reserved doctors, and doctor types;
the reserved time feature at least comprises: the appointment occurrence time, the appointment visit duration, and the difference between the appointment occurrence time and the appointment visit time.
10. Pediatric internet+outpatient and refreshing prediction device based on deep learning, which is characterized by comprising:
a processor and a memory;
the processor is connected with the memory through a communication bus:
the processor is used for calling and executing the program stored in the memory;
the memory for storing a program for performing at least one deep learning based pediatric internet + outpatient prediction method of any one of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310551325.8A CN116307272A (en) | 2023-05-17 | 2023-05-17 | Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310551325.8A CN116307272A (en) | 2023-05-17 | 2023-05-17 | Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116307272A true CN116307272A (en) | 2023-06-23 |
Family
ID=86796262
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310551325.8A Pending CN116307272A (en) | 2023-05-17 | 2023-05-17 | Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116307272A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105224989A (en) * | 2014-06-05 | 2016-01-06 | 上海宝信软件股份有限公司 | Automobile leasing based on the prediction of motion interval historical data is super orders management system |
| CN106992942A (en) * | 2017-03-28 | 2017-07-28 | 浙江工商大学 | A kind of SDN resource pricing methods based on resource load and user's request |
| CN107278304A (en) * | 2015-02-27 | 2017-10-20 | 皇家飞利浦有限公司 | The system that health care reservation is dispatched for failing to keep an appointment probability based on patient |
| JP2019200750A (en) * | 2018-05-18 | 2019-11-21 | ヤフー株式会社 | Information processor, method for processing information, and information processing program |
| CN113223688A (en) * | 2020-08-11 | 2021-08-06 | 株式会社迈梦特 | Outpatient service management service platform and system for providing outpatient service management service platform |
-
2023
- 2023-05-17 CN CN202310551325.8A patent/CN116307272A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105224989A (en) * | 2014-06-05 | 2016-01-06 | 上海宝信软件股份有限公司 | Automobile leasing based on the prediction of motion interval historical data is super orders management system |
| CN107278304A (en) * | 2015-02-27 | 2017-10-20 | 皇家飞利浦有限公司 | The system that health care reservation is dispatched for failing to keep an appointment probability based on patient |
| CN106992942A (en) * | 2017-03-28 | 2017-07-28 | 浙江工商大学 | A kind of SDN resource pricing methods based on resource load and user's request |
| JP2019200750A (en) * | 2018-05-18 | 2019-11-21 | ヤフー株式会社 | Information processor, method for processing information, and information processing program |
| CN113223688A (en) * | 2020-08-11 | 2021-08-06 | 株式会社迈梦特 | Outpatient service management service platform and system for providing outpatient service management service platform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Coorey et al. | The health digital twin to tackle cardiovascular disease—a review of an emerging interdisciplinary field | |
| Shickel et al. | DeepSOFA: a continuous acuity score for critically ill patients using clinically interpretable deep learning | |
| US20220310267A1 (en) | Evaluating Risk of a Patient Based on a Patient Registry and Performing Mitigating Actions Based on Risk | |
| Chan et al. | Cardiac performance measure compliance in outpatients: the American College of Cardiology and National Cardiovascular Data Registry's PINNACLE (practice innovation and clinical excellence) program | |
| CN104166667B (en) | Analysis system and public health work support method | |
| US10014080B2 (en) | Evidence based medical record | |
| Chang et al. | Assessment of critical exposure and outcome windows in time-to-event analysis with application to air pollution and preterm birth study | |
| JP6159872B2 (en) | Medical data analysis system, medical data analysis method, and storage medium | |
| See et al. | Does happiness matter to health system efficiency? A performance analysis | |
| Sutradhar et al. | Multistate analysis of interval-censored longitudinal data: application to a cohort study on performance status among patients diagnosed with cancer | |
| Viana et al. | A multi-paradigm, whole system view of health and social care for age-related macular degeneration | |
| Faries et al. | Real world health care data analysis: causal methods and implementation using SAS | |
| D'Souza et al. | Clinical decision analysis in perinatology | |
| Buja et al. | Systematic age-related differences in chronic disease management in a population-based cohort study: a new paradigm of primary care is required | |
| Merz et al. | Exploring the antecedents for value cocreation during healthcare service provision | |
| Lin et al. | Time-to-event predictive modeling for chronic conditions using electronic health records | |
| Bhatia et al. | Prediction of severity of diabetes mellitus using fuzzy cognitive maps | |
| Drouin et al. | Cost-effectiveness of a smoking cessation intervention for parents in pediatric primary care | |
| Riley et al. | Prognostic model research | |
| Tripathy et al. | Innovative classification, regression model for predicting various diseases | |
| CN116307272A (en) | Pediatric Internet+outpatient and refreshing prediction method and equipment based on deep learning | |
| US7877341B2 (en) | Self-adaptive data pre-fetch by artificial neuron network | |
| Shickel et al. | DeepSOFA: a real-time continuous acuity score framework using deep learning | |
| Goldhaber-Fiebert et al. | Modeling and calibration for exposure to time-varying, modifiable risk factors: the example of smoking behavior in India | |
| CN113223677A (en) | Doctor matching method and device for patient |
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 |