CN116798630B - Myopia physiotherapy compliance prediction method, device and medium based on machine learning - Google Patents
Myopia physiotherapy compliance prediction method, device and medium based on machine learning Download PDFInfo
- Publication number
- CN116798630B CN116798630B CN202310820057.5A CN202310820057A CN116798630B CN 116798630 B CN116798630 B CN 116798630B CN 202310820057 A CN202310820057 A CN 202310820057A CN 116798630 B CN116798630 B CN 116798630B
- Authority
- CN
- China
- Prior art keywords
- data
- model
- feature
- patient data
- graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000000554 physical therapy Methods 0.000 title claims abstract description 39
- 208000001491 myopia Diseases 0.000 title claims abstract description 15
- 230000004379 myopia Effects 0.000 title claims abstract description 15
- 238000010801 machine learning Methods 0.000 title claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 38
- 230000004927 fusion Effects 0.000 claims abstract description 21
- 238000007637 random forest analysis Methods 0.000 claims abstract description 14
- 238000013145 classification model Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 11
- 238000012300 Sequence Analysis Methods 0.000 claims abstract description 10
- 230000007246 mechanism Effects 0.000 claims abstract description 10
- 238000012731 temporal analysis Methods 0.000 claims abstract description 7
- 238000000700 time series analysis Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 18
- 230000004913 activation Effects 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000004140 cleaning Methods 0.000 claims description 5
- 238000002790 cross-validation Methods 0.000 claims description 4
- 230000004424 eye movement Effects 0.000 claims description 4
- 210000001747 pupil Anatomy 0.000 claims description 4
- 230000004931 aggregating effect Effects 0.000 claims description 2
- 210000002569 neuron Anatomy 0.000 claims description 2
- 230000036962 time dependent Effects 0.000 claims description 2
- 238000002560 therapeutic procedure Methods 0.000 claims 4
- 230000008859 change Effects 0.000 description 5
- 238000003066 decision tree Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000001225 therapeutic effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000017531 blood circulation Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 230000005802 health problem Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- 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
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
-
- 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
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Theoretical Computer Science (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Pathology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a myopia physiotherapy compliance prediction method, equipment and medium based on machine learning, wherein the method comprises the following steps: collecting patient data, the patient data including physiotherapy data and personal data; extracting data features from the patient data using a multi-branch feature extraction network, the multi-branch feature extraction network comprising a self-encoder, a time series analysis model and a graph-based feature extraction model; adopting an attention mechanism to perform feature fusion on the data features; training the self-adaptive random forest model by using the fusion characteristics to obtain a classification model; and inputting patient data to be predicted into the trained classification model to obtain a prediction result. According to the invention, the data characteristics are extracted from the patient data through three different angles by the self-encoder, the time sequence analysis model and the graph-based characteristic extraction model in the multi-branch characteristic extraction network, so that the accuracy and the reliability of prediction are improved.
Description
Technical Field
The invention belongs to the technical field of medicine, and particularly relates to a myopia physiotherapy compliance prediction method, equipment and medium based on machine learning.
Background
Myopia has become a common eye health problem worldwide, and affects the quality of life of people. There are many methods for correcting myopia, and red light irradiation has been attracting attention as a non-drug treatment method. The red light irradiation can improve visual fatigue, promote blood circulation of eyes and has a certain myopia control effect. However, patient compliance is critical to the effectiveness of the treatment, but there is currently a lack of effective methods to assess patient compliance with myopia treatment.
Furthermore, existing compliance assessment techniques suffer from the following drawbacks:
the data processing capacity is insufficient: the prior art may not adequately process complex medical data, such as time dependencies, nonlinearities, inherent correlations of data, complex structures, and the like. If these complex data cannot be processed efficiently, it may result in a decrease in the accuracy and reliability of the prediction.
The feature extraction and fusion are insufficient: traditional methods may employ only one or a limited feature extraction and fusion approach, which may ignore the diversity of the data and reduce the generalization ability of the model.
Classifier optimization problem: existing classifiers, such as conventional random forests, generally assume that the weights of all trees are the same, which may not be optimal in the face of different problems and data.
In recent years, machine learning has been increasingly used in the medical field. Machine learning is an artificial intelligence technique that automatically predicts and makes decisions by learning rules and patterns from data. In compliance assessment, machine learning can identify features and patterns related to patient compliance by analyzing a large number of data samples, thereby providing more accurate assessment and prediction.
Disclosure of Invention
In order to overcome the technical defects, the invention provides a myopia physiotherapy compliance prediction method based on machine learning, which is used for extracting data features from patient data through three different angles by a self-encoder, a time sequence analysis model and a graph-based feature extraction model in a multi-branch feature extraction network, so that the accuracy and reliability of prediction are improved.
In order to solve the problems, the first aspect of the invention discloses a myopia physiotherapy compliance prediction method based on machine learning, which comprises the following steps:
collecting patient data, the patient data including physiotherapy data and personal data;
extracting data features from the patient data using a multi-branch feature extraction network, the multi-branch feature extraction network comprising a self-encoder, a time series analysis model and a graph-based feature extraction model;
adopting an attention mechanism to perform feature fusion on the data features;
training the self-adaptive random forest model by using the fusion characteristics to obtain a classification model;
and inputting patient data to be predicted into the trained classification model to obtain a prediction result.
Further, the physiotherapy data comprises physiotherapy duration, physiotherapy frequency, physiotherapy intensity, eye movement track, pupil size change and/or blink frequency; the personal data includes age and/or gender.
Further, before the step of extracting the data features from the patient data using the multi-branch feature extraction network, the method comprises the steps of:
the patient data is preprocessed, and preprocessing content comprises data cleaning, missing value processing and/or outlier processing.
Further, the steps of extracting data features from patient data using a multi-branch feature extraction network include the steps of:
the patient data is processed using the self-encoder to extract a first data feature.
Further, the steps of extracting data features from patient data using a multi-branch feature extraction network include the steps of:
and processing the patient data by adopting a time sequence analysis method, and extracting second data characteristics.
Further, the steps of extracting data features from patient data using a multi-branch feature extraction network include the steps of:
the patient data is processed based on the feature extraction of the graph, and a third data feature is extracted.
Further, the step of performing feature fusion on the data features by adopting an attention mechanism comprises the following steps:
inputting data features into a feature embedding layer:
h i =E i (x i )
wherein h is i Is the embedded data feature, x i Is the original data of type i;
the attention score of the data feature is calculated by adopting the full connection layer:
a i =σ(W a h i +b a )
wherein W is a And b a Is a parameter of the full connection layer, σ () is Sigmoid activation function, a i Is the attention score of data feature i;
and carrying out weighted summation according to the attention scores, and calculating fusion characteristics.
Further, the step of inputting patient data to be predicted into a trained classification model, and before obtaining a prediction result, comprising the steps of:
and evaluating the trained model by using a K-fold cross validation method.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses a myopia physiotherapy compliance prediction method based on machine learning, which is characterized in that data features are extracted from patient data through three different angles by a self-encoder, a time sequence analysis model and a graph-based feature extraction model in a multi-branch feature extraction network, so that the accuracy and reliability of prediction are improved; in addition, the invention also improves the accuracy of prediction by adopting an attention mechanism to perform feature fusion and adopting a self-adaptive random forest classifier.
In a second aspect the invention discloses a computer device comprising a memory storing a computer program and a processor implementing the steps of the method described above when executing the computer program.
A third aspect of the present invention discloses a computer readable storage medium having stored thereon a computer program which when executed by a processor realizes the steps of the above method.
Drawings
The invention is described in further detail below with reference to the attached drawing figures, wherein:
FIG. 1 is a flowchart of a compliance prediction method described in example 1;
fig. 2 is a schematic diagram of the computer device described in embodiment 2.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.
For a better understanding of the embodiments of the present application, the related art of the embodiments of the present application will be described first.
Random forests are an ensemble learning algorithm that consists of a number of decision trees, with the final prediction result determined by voting or averaging. Each tree is built based on a part of the training data randomly selected feature subsets, so that the occurrence of the over-fitting phenomenon can be reduced, and the generalization capability of the model is improved. In random forests, each decision tree randomly samples the data set and each time a division occurs, the best division point is selected in a random subset of features. Thus, each tree has certain independence, and the problem of over fitting is avoided. In the prediction process, the prediction results of all trees in the random forest are integrated in a voting or average number mode and the like, so that a final prediction result is obtained. Random forests generally have better performance and higher accuracy than single decision trees. It is suitable for classification and regression tasks and can process datasets with a large number of features and samples.
Example 1
As shown in fig. 1, the embodiment discloses a myopia physiotherapy compliance prediction method based on machine learning, which comprises the following steps:
s1, collecting patient data, wherein the patient data comprise physiotherapy data and personal data, and the physiotherapy data comprise physiotherapy duration, physiotherapy frequency, physiotherapy intensity, eye movement track, pupil size change and/or blink frequency; the personal data includes age and/or gender.
In the above embodiment, the step S2 is preceded by the following steps:
and preprocessing the patient data, wherein preprocessing contents comprise data cleaning, missing value processing and/or abnormal value processing, so that the quality and consistency of the data are ensured.
S2, extracting data features from the patient data by adopting a multi-branch feature extraction network, and extracting features from different types of data. The branch feature extraction network comprises a self-encoder, a time sequence analysis model and a graph-based feature extraction model.
Specifically, step S2 includes the following steps:
processing patient data using a self-encoder to extract a first data feature:
the objective function of the self-encoder is:
wherein ||x-g (phi (f (theta (x))))) ||j 2 Is a reconstruction error;is a sparsity regularization term; λ is a regularization parameter for controlling the strength of the sparsity constraint; ρ is a sparsity parameter representing the expected average activation value of hidden layer neurons; />Is the actual average activation value.
Inputting patient data into the objective function of the self-encoder to obtain a hidden representation h=f θ (x) H is taken as the first data characteristic.
Processing the patient data by adopting a time sequence analysis model, and extracting second data characteristics:
the time-dependent features in the physiotherapy data are processed by a time-series analysis model. For time series { x } in physiotherapy data t An autoregressive model is represented as:
wherein c is a constant, phi i Is a model parameter, epsilon t Is the error term and p is the time series length. Inputting the time sequence in the physiotherapy data into an autoregressive model to obtain model parametersAs a second data feature.
Processing the patient data based on the feature extraction method of the graph, and extracting a third data feature:
in complex medical data, the inherent relevance of the data and the complex structure can generally be better expressed by the graph structure. Patient data is converted into a graph structure in which nodes represent patients and edges represent similarities between patients. The node characteristics include physiotherapy data and personal data.
Extracting complex patterns of the graph on node characteristics and graph structures through a graph convolution network, wherein the basic operation of the graph convolution network is as follows:
H (l+1) =σ(D -1 AH (l) W (l) )
wherein A is the adjacency matrix of the graph, D is the node degree matrix, H (l) Is a node feature of the first layer, W (l) Is the weight matrix of the first layer, σ is a nonlinear activation function.
Inputting patient data to obtain new characteristics of each node, and aggregating the characteristics to obtain the characteristic H of the whole graph Tu As a third data feature.
And S3, carrying out feature fusion on each data feature by adopting an attention mechanism.
Specifically, step S3 includes the following steps:
inputting each data feature obtained through step S2 to the feature embedding layer:
h i =E i (x i )
wherein h is i Is the embedded data feature, x i Is the original data of the data characteristic type i. Mapping each data feature to a shared hidden space through a feature embedding layer; for each data feature type i, there is a separate embedded network E i Implemented with a fully connected layer or a small neural network.
Feature fusion is carried out according to the importance of each data feature, and attention scores of the data features are calculated by adopting a full connection layer firstly:
a i =σ(W a h i +b a )
wherein W is a And b a Is a parameter of the full connection layer, σ () is Sigmoid activation function, a i Is the attention score of the data feature i.
And carrying out weighted summation on each data characteristic according to the attention score, and calculating a fusion characteristic.
The feature fusion can increase the diversity and information richness of features, improve the expression capacity of the model and enhance the modeling capacity of complex data relations. By fusing a plurality of feature sources, more comprehensive and accurate feature representation can be obtained, so that the performance and prediction capability of the machine learning model are improved.
And S4, training the self-adaptive random forest model by using the fusion characteristics to obtain a classification model.
Specifically, step S4 includes the following steps:
each tree is assigned the same weight:
where n is the total number of trees in the random forest.
After each training period is finished, calculating the accuracy acc of each tree i The weight of each tree is then updated with the following formula:
w i =w i ·acc i
to ensure that the sum of the weights of all trees is 1, the weights of all trees are normalized:
in the prediction, the voting weight of each tree is determined by the weight of each tree.
Specifically, let c j Represents class j, p ij Representing the ith tree pre-runThe probability of class j is measured, then the final predicted probability is:
finally, the class with the highest probability is selected as the prediction result.
In the above embodiment, the step S5 further includes the steps of:
and evaluating the trained model by using a K-fold cross validation method. Patient data is divided into a plurality of folds, training and validation is performed on each fold, and the model is evaluated by an evaluation index. The evaluation index comprises accuracy, precision, recall, F1 score and the like.
S5, inputting patient data to be predicted into the trained classification model, and obtaining a prediction result about patient compliance. The predicted outcome may be a label, such as "compliance" or "non-compliance"; or probability values, indicating the likelihood that the patient belongs to a certain category. The medical professional can assess patient compliance based on the prediction and take appropriate intervention and therapeutic measures.
In the above embodiment, the method further includes the steps of:
and acquiring input feedback data, and iterating and optimizing the model. Wherein the feedback data comprises actual compliance of the patient and treatment effect, and the optimization content comprises adjustment of feature selection, adjustment of model parameters, improvement of data collection and the like.
And the model is optimized according to the prediction result and the actual condition, so that the accuracy and the practicability of the model are improved.
The following is further described by the specific implementation procedure:
patient data is acquired, wherein the patient data comprises physiotherapy data and personal data, and the physiotherapy data comprises physiotherapy duration, physiotherapy frequency, physiotherapy intensity, eye movement track, pupil size change and/or blink frequency; the personal data includes age and/or gender.
And performing data cleaning, missing value processing and abnormal value processing on the patient data to ensure the quality and consistency of the data.
The multi-branch feature extraction network is adopted to extract data features from patient data through three different angles, such as extracting certain data patterns which are difficult to observe from the data by a self-encoder, finding a certain trend between physiotherapy duration and physiotherapy frequency by a time sequence analysis model, and finding the relationship between physiotherapy intensity and other factors by a feature extraction model based on a graph. The data features from different sources are fused by using an attention mechanism, so that the model can better understand and utilize the data features, and the prediction accuracy is improved.
The self-adaptive random forest model is used for training, and after each training period is finished, the weight of each tree is self-adaptively adjusted according to the prediction performance, so that the model can be used for various conditions, and the prediction accuracy is improved.
The patient A is a myopia crowd, red light irradiation in myopia physiotherapy needs to be regularly received, after model training is finished, new data of the patient A are input into the model when physiotherapy is finished, compliance evaluation results of the patient A are obtained, and if 80% of the patient A is predicted to be treated according to doctor's advice. Based on the compliance assessment results, the medical professional can assess compliance of patient a and take appropriate intervention and therapeutic measures.
In addition, the actual compliance condition and treatment effect of the inputted patient A are obtained, and if the patient A is not actually treated according to the estimated compliance, model parameters or optimization feature selection and the like can be adjusted so as to improve the accuracy and the practicability of the model.
According to the invention, the quality and consistency of the data are ensured by preprocessing the acquired data, so that the accuracy and reliability of the model can be increased; the data characteristics are extracted from the patient data through three different angles through a self-encoder, a time sequence analysis model and a graph-based characteristic extraction model in a multi-branch characteristic extraction network, so that the accuracy and the reliability of prediction are improved, and the expressive power and the generalization capability of the model are improved; the feature fusion is carried out through the attention mechanism, so that the model can be helped to better understand and utilize different types of data, and the prediction accuracy is improved; in addition, the invention also adopts the self-adaptive random forest classifier to improve the prediction accuracy.
Example 2
As shown in fig. 2, the present embodiment discloses a computer device, including a memory and a processor, the memory storing a computer program, the processor being configured to execute the computer program, the processor implementing the following steps when executing the computer program:
collecting patient data, the patient data including physiotherapy data and personal data;
extracting data features from the patient data using a multi-branch feature extraction network, the multi-branch feature extraction network comprising a self-encoder, a time series analysis model and a graph-based feature extraction model;
adopting an attention mechanism to perform feature fusion on the data features;
training the self-adaptive random forest model by using the fusion characteristics to obtain a classification model;
and inputting patient data to be predicted into the trained classification model to obtain a prediction result.
Specifically, the processor when executing the computer program further implements the following steps:
the patient data is preprocessed, and preprocessing content comprises data cleaning, missing value processing and/or outlier processing.
Specifically, the processor when executing the computer program further implements the following steps:
the patient data is processed using the self-encoder to extract a first data feature.
Specifically, the processor when executing the computer program further implements the following steps:
and processing the patient data by adopting a time sequence analysis method, and extracting second data characteristics.
Specifically, the processor when executing the computer program further implements the following steps:
the patient data is processed based on the feature extraction of the graph, and a third data feature is extracted.
Specifically, the processor when executing the computer program further implements the following steps:
inputting data features into a feature embedding layer:
h i =E i (x i )
wherein h is i Is the embedded data feature, x i Is the original data of type i;
the attention score of the data feature is calculated by adopting the full connection layer:
a i =σ(W a h i +b a )
wherein W is a And b a Is a parameter of the full connection layer, σ () is Sigmoid activation function, a i Is the attention score of data feature i;
and carrying out weighted summation according to the attention scores, and calculating fusion characteristics.
Specifically, the processor when executing the computer program further implements the following steps:
and evaluating the trained model by using a K-fold cross validation method.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile memory may include Read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high density embedded nonvolatile memory, resistive random access memory (ReRAM), magnetic random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic RandomAccess Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.
Example 3
The present embodiment discloses a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the method described in embodiment 1.
The present invention is not limited to the preferred embodiments, and any modifications, equivalent variations and modifications made to the above embodiments according to the technical principles of the present invention are within the scope of the technical proposal of the present invention.
Claims (7)
1. A myopia physiotherapy compliance prediction method based on machine learning is characterized by comprising the following steps:
collecting patient data, the patient data including physiotherapy data and personal data;
extracting data features from the patient data using a multi-branch feature extraction network, the multi-branch feature extraction network comprising a self-encoder, a time series analysis model and a graph-based feature extraction model;
adopting an attention mechanism to perform feature fusion on the data features;
training the self-adaptive random forest model by using the fusion characteristics to obtain a classification model;
inputting patient data to be predicted into a trained classification model to obtain a prediction result;
the step of extracting data features from patient data using a multi-branch feature extraction network includes the steps of:
processing patient data by using a self-encoder to extract first data features;
processing the patient data by adopting a time sequence analysis model, and extracting second data characteristics;
processing the patient data based on the feature extraction method of the graph, and extracting a third data feature;
the steps of processing patient data using a self-encoder to extract a first data feature, comprising:
the objective function of the self-encoder is:
wherein ||x-g (phi (f (theta (x))))) ||j 2 Is a reconstruction error;is a sparsity regularization term; λ is a regularization parameter for controlling the strength of the sparsity constraint; ρ is a sparsity parameter representing the expected average activation value of hidden layer neurons; />Is the actual average activation value;
inputting patient data into the objective function of the self-encoder to obtain a hidden representation h=f θ (x) Taking h as a first data characteristic;
processing the patient data by using a time series analysis model, extracting second data features, including:
processing time-dependent features in physiotherapy data by a time-series analysis model, time-series { x } in physiotherapy data t An autoregressive model is represented as:
wherein c is a constant, phi i Is a model parameter, epsilon t Is an error term, p is the time sequence length, and the time sequence in the physiotherapy data is input into an autoregressive model to obtain model parametersAs a second data feature;
the step of processing the patient data based on the feature extraction of the graph, extracting a third data feature, comprising:
extracting complex patterns of the graph on node characteristics and graph structures through a graph convolution network, wherein the basic operation of the graph convolution network is as follows:
H (l+1) =σ(D -1 AH (l) W (l) )
wherein A is the adjacency matrix of the graph, D is the node degree matrix, H (l) Is a node feature of the first layer, W (l) Is the weight matrix of the first layer, sigma is a nonlinear activation function;
inputting patient data to obtain new characteristics of each node, and aggregating the characteristics to obtain the characteristic H of the whole graph Tu As a third data feature.
2. The compliance prediction method of claim 1, wherein the therapy data comprises therapy duration, therapy frequency, therapy intensity, eye movement trajectory, pupil size variation, and/or blink frequency; the personal data includes age and/or gender.
3. The compliance prediction method according to claim 1, wherein before the step of extracting data features from patient data using the multi-branch feature extraction network, comprising the steps of:
the patient data is preprocessed, and preprocessing content comprises data cleaning, missing value processing and/or outlier processing.
4. The compliance prediction method according to claim 1, wherein the step of feature fusing the data features using an attention mechanism comprises the steps of:
inputting data features into a feature embedding layer:
h i =E i (x i )
wherein h is i Is the embedded data feature, x i Is the original data of type i;
the attention score of the data feature is calculated by adopting the full connection layer:
a i =σ(W a h i +b a )
wherein W is a And b a Is a parameter of the full connection layer, σ () is Sigmoid activation function, a i Is the attention score of data feature i;
and carrying out weighted summation according to the attention scores, and calculating fusion characteristics.
5. The compliance prediction method according to claim 1, wherein the step of inputting patient data to be predicted into a trained classification model, before obtaining a prediction result, comprises the steps of:
and evaluating the trained model by using a K-fold cross validation method.
6. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
7. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310820057.5A CN116798630B (en) | 2023-07-05 | 2023-07-05 | Myopia physiotherapy compliance prediction method, device and medium based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310820057.5A CN116798630B (en) | 2023-07-05 | 2023-07-05 | Myopia physiotherapy compliance prediction method, device and medium based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116798630A CN116798630A (en) | 2023-09-22 |
CN116798630B true CN116798630B (en) | 2024-03-08 |
Family
ID=88040014
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310820057.5A Active CN116798630B (en) | 2023-07-05 | 2023-07-05 | Myopia physiotherapy compliance prediction method, device and medium based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116798630B (en) |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109637669A (en) * | 2018-11-22 | 2019-04-16 | 中山大学 | Generation method, device and the storage medium of therapeutic scheme based on deep learning |
CN112002410A (en) * | 2020-08-20 | 2020-11-27 | 医渡云(北京)技术有限公司 | Infectious disease state prediction method and device, storage medium, and electronic device |
CN112861978A (en) * | 2021-02-20 | 2021-05-28 | 齐齐哈尔大学 | Multi-branch feature fusion remote sensing scene image classification method based on attention mechanism |
CN113763965A (en) * | 2021-08-26 | 2021-12-07 | 江苏大学 | Speaker identification method with multiple attention characteristics fused |
CN113902186A (en) * | 2021-09-30 | 2022-01-07 | 中南大学 | Patient death risk prediction method, system, terminal and readable storage medium based on electronic medical record |
CN115062538A (en) * | 2022-06-07 | 2022-09-16 | 武汉大学 | Converter fault diagnosis method and system based on attention mechanism and convolution NN |
CN115188429A (en) * | 2022-07-27 | 2022-10-14 | 华东理工大学 | Catalytic cracking unit key index modeling method integrating time sequence feature extraction |
CN115985515A (en) * | 2023-03-20 | 2023-04-18 | 广东工业大学 | Amblyopia correction effect prediction method, device and equipment based on machine learning |
CN116167617A (en) * | 2022-12-29 | 2023-05-26 | 福州大学 | Geological disaster risk assessment method and system integrating random forest and attention |
CN116259415A (en) * | 2022-10-25 | 2023-06-13 | 西安理工大学 | Patient medicine taking compliance prediction method based on machine learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11676719B2 (en) * | 2018-12-20 | 2023-06-13 | Oregon Health & Science University | Subtyping heterogeneous disorders using functional random forest models |
-
2023
- 2023-07-05 CN CN202310820057.5A patent/CN116798630B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109637669A (en) * | 2018-11-22 | 2019-04-16 | 中山大学 | Generation method, device and the storage medium of therapeutic scheme based on deep learning |
CN112002410A (en) * | 2020-08-20 | 2020-11-27 | 医渡云(北京)技术有限公司 | Infectious disease state prediction method and device, storage medium, and electronic device |
CN112861978A (en) * | 2021-02-20 | 2021-05-28 | 齐齐哈尔大学 | Multi-branch feature fusion remote sensing scene image classification method based on attention mechanism |
CN113763965A (en) * | 2021-08-26 | 2021-12-07 | 江苏大学 | Speaker identification method with multiple attention characteristics fused |
CN113902186A (en) * | 2021-09-30 | 2022-01-07 | 中南大学 | Patient death risk prediction method, system, terminal and readable storage medium based on electronic medical record |
CN115062538A (en) * | 2022-06-07 | 2022-09-16 | 武汉大学 | Converter fault diagnosis method and system based on attention mechanism and convolution NN |
CN115188429A (en) * | 2022-07-27 | 2022-10-14 | 华东理工大学 | Catalytic cracking unit key index modeling method integrating time sequence feature extraction |
CN116259415A (en) * | 2022-10-25 | 2023-06-13 | 西安理工大学 | Patient medicine taking compliance prediction method based on machine learning |
CN116167617A (en) * | 2022-12-29 | 2023-05-26 | 福州大学 | Geological disaster risk assessment method and system integrating random forest and attention |
CN115985515A (en) * | 2023-03-20 | 2023-04-18 | 广东工业大学 | Amblyopia correction effect prediction method, device and equipment based on machine learning |
Also Published As
Publication number | Publication date |
---|---|
CN116798630A (en) | 2023-09-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Gadekallu et al. | Deep neural networks to predict diabetic retinopathy | |
Maity et al. | Machine learning for improved diagnosis and prognosis in healthcare | |
Gil et al. | Predicting seminal quality with artificial intelligence methods | |
Sarveshvar et al. | Performance of different machine learning techniques for the prediction of heart diseases | |
Selvakuberan et al. | An efficient feature selection method for classification in health care systems using machine learning techniques | |
WO2020234388A1 (en) | Propensity score based assessment of patient data | |
Fakih et al. | An efficient prediction of diabetes using artificial neural networks | |
Sharma et al. | A linearly adaptive Sine–cosine algorithm with application in deep neural network for feature optimization in arrhythmia classification using ECG signals | |
Ahmadi et al. | Eye State Identification Utilizing EEG Signals: A Combined Method Using Self‐Organizing Map and Deep Belief Network | |
Soğukkuyu et al. | Diagnosing cervical cancer using machine learning methods | |
Patel | Predicting a risk of diabetes at early stage using machine learning approach | |
Trivedi et al. | Machine learning based evaluations of stress, depression, and anxiety | |
Kumar et al. | IBM auto AI bot: diabetes mellitus prediction using machine learning algorithms | |
CN116798630B (en) | Myopia physiotherapy compliance prediction method, device and medium based on machine learning | |
Nouei et al. | Developing a genetic fuzzy system for risk assessment of mortality after cardiac surgery | |
Pious et al. | Heart disease prediction using machine learning algorithms | |
Barry | A deep learning approach to diagnosing schizophrenia | |
El-Shafiey et al. | Heart-Disease Prediction Method Using Random Forest and Genetic Algorithms | |
Kumar et al. | Chronic Kidney Disease Prediction Using Gradient Boosting and KNN Classifier | |
Fidan et al. | Classification of dermatological data with self organizing maps and support vector machine | |
Saranya et al. | Bd-Mdl: bipolar disorder detection using machine leanring and deep learning | |
Gabrys | Machine learning using radiomics and dosiomics for normal tissue complication probability modeling of radiation-induced xerostomia | |
Mehrno et al. | Feature selection of Thyroid disease using Deep Learning: A Literature survey | |
Varma et al. | Early Detection of Alzheimer’s Disease Using SVM, Random Forest & FNN Algorithms | |
Kavitha et al. | An Efficient Boruta-Based Feature Selection and Classification of Gene Expression Data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |