CN114530228B - Blood glucose prediction method and system based on smoothing and fusion and medical equipment - Google Patents
Blood glucose prediction method and system based on smoothing and fusion and medical equipment Download PDFInfo
- Publication number
- CN114530228B CN114530228B CN202210109388.3A CN202210109388A CN114530228B CN 114530228 B CN114530228 B CN 114530228B CN 202210109388 A CN202210109388 A CN 202210109388A CN 114530228 B CN114530228 B CN 114530228B
- Authority
- CN
- China
- Prior art keywords
- blood sugar
- matrix
- smooth
- characteristic
- values
- 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
- 239000008280 blood Substances 0.000 title claims abstract description 206
- 210000004369 blood Anatomy 0.000 title claims abstract description 206
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000009499 grossing Methods 0.000 title claims abstract description 39
- 230000004927 fusion Effects 0.000 title claims abstract description 20
- WQZGKKKJIJFFOK-GASJEMHNSA-N Glucose Natural products OC[C@H]1OC(O)[C@H](O)[C@@H](O)[C@@H]1O WQZGKKKJIJFFOK-GASJEMHNSA-N 0.000 title claims description 49
- 239000008103 glucose Substances 0.000 title claims description 49
- 239000011159 matrix material Substances 0.000 claims abstract description 117
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000012795 verification Methods 0.000 claims abstract description 30
- 238000012360 testing method Methods 0.000 claims abstract description 26
- 238000012545 processing Methods 0.000 claims abstract description 17
- 235000021061 dietary behavior Nutrition 0.000 claims abstract description 8
- 238000007637 random forest analysis Methods 0.000 claims description 38
- 238000010606 normalization Methods 0.000 claims description 15
- 230000020595 eating behavior Effects 0.000 claims description 13
- 238000003066 decision tree Methods 0.000 claims description 10
- 230000006399 behavior Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000036772 blood pressure Effects 0.000 claims description 3
- 235000021070 high sugar diet Nutrition 0.000 claims description 3
- 235000020887 ketogenic diet Nutrition 0.000 claims description 3
- 235000021590 normal diet Nutrition 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- 206010012601 diabetes mellitus Diseases 0.000 abstract description 17
- 230000000875 corresponding effect Effects 0.000 description 26
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 150000001720 carbohydrates Chemical class 0.000 description 3
- 235000014633 carbohydrates Nutrition 0.000 description 3
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000037149 energy metabolism Effects 0.000 description 2
- 230000037406 food intake Effects 0.000 description 2
- 238000001727 in vivo Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000000056 organ Anatomy 0.000 description 2
- 208000035977 Rare disease Diseases 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000037081 physical activity Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/60—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to nutrition control, e.g. diets
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—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
-
- 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
- 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
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Epidemiology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Primary Health Care (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Nutrition Science (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention provides a blood sugar prediction method, a system and medical equipment based on smoothing processing and fusion, relating to the technical field of medical equipment, firstly collecting characteristic values of original data, normalizing the characteristic values, smoothing the normalized characteristic values through correlation coefficients to make the relation between the characteristic values and blood sugar measured values more relevant, forming a characteristic matrix as a data set, dividing the characteristic matrix into a training set, a verification set and a test set, combining the characteristic values with different characteristic contribution rates, fusing blood sugar prediction result matrixes obtained by different blood sugar prediction models by using the optimal correlation coefficients to obtain a final blood sugar prediction result, fully considering diet behavior change, reducing the deviation of the diet behavior change on the prediction result, and improving the accuracy of the blood sugar prediction result, wherein the method can be packaged in a system to further form the application of the medical equipment, provides guidance for diabetic patients to adopt self-intervention measures in daily life.
Description
Technical Field
The invention relates to the technical field of medical instruments, in particular to a blood sugar prediction method and system based on smoothing and fusion and medical equipment.
Background
In recent years, with the improvement of living standard of people, the living style of people is continuously changed, the working intensity is increased, the physical activity is obviously reduced, the life rhythm is accelerated, people are in a stress environment for a long time, the number of people suffering from diabetes is increased rapidly, and the diabetes is changed into epidemic disease from rare disease.
Diabetes is a lifelong disease, and the current medical level cannot eradicate the disease, so that the monitoring of the blood sugar of a diabetic patient is particularly important. The diabetes patients have poor blood sugar regulation capability, and when the ingestion behavior of exogenous carbohydrates is changed, the great fluctuation of blood sugar in vivo can be caused, and the blood sugar fluctuation can cause the damage of body organs and accelerate the occurrence of complications. Therefore, the thinking is brought forward, if the exogenous intake behavior of the diabetic can be positively intervened, the diabetic can be reminded to take carbohydrate in an effective time in time, the method is very important for the diabetic to enjoy normal life, and the blood sugar can be accurately predicted, so that the purpose is realized, and the blood sugar basis is conveniently provided for the diabetic in daily life.
The scheme is based on a random forest algorithm and a neural network prediction algorithm, firstly, energy metabolism parameters of a finger tip of an individual to be measured are measured and sent to a measurement host, the measurement host calls the random forest algorithm to predict the blood sugar category of the energy metabolism parameters of the finger tip of the individual to be measured, and calls a corresponding neural network prediction algorithm to calculate the blood sugar value of the finger tip of the individual to be measured, so that the noninvasive prediction of the blood sugar is realized.
Disclosure of Invention
In order to solve the problem that the accuracy of blood sugar prediction is low due to the fact that the conventional blood sugar prediction method cannot adapt to changes of individual eating behaviors, the invention provides a blood sugar prediction method, a blood sugar prediction system and medical equipment based on smoothing processing and fusion, so that the deviation of the individual eating behavior changes to the prediction result is reduced, the accuracy of the blood sugar prediction result is improved, and references are provided for diabetic patients to self-monitor blood sugar levels and adopt self-intervention measures in daily life.
In order to achieve the technical effects, the technical scheme of the invention is as follows:
a blood sugar prediction method based on smoothing and fusion comprises the following steps:
s1, determining individual dietary behaviors and a data acquisition time period, acquiring physiological index data of N times of individuals under different dietary behaviors as original characteristic values, and actually measuring blood glucose values corresponding to the physiological index data synchronously to serve as blood glucose reference values;
s2, forming an original characteristic value matrix based on the original characteristic values, and carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and form a new characteristic value matrix;
s3, calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
s4, introducing a random forest model, calculating the characteristic contribution rate of the smooth characteristic values by using the random forest model, sequencing, and selecting the smooth characteristic values with the characteristic contribution rate larger than zero;
s5, selecting and extracting smooth characteristic values ranked at the front e position from the training set to serve as a fixed set, uniformly dividing the remaining smooth characteristic values into f sets, combining the uniformly divided f sets and the e-position smooth characteristic values into a new characteristic vector, and forming f characteristic matrixes;
s6, training a random forest model by using the f feature matrices and blood sugar reference values synchronously corresponding to the f feature matrices to obtain f blood sugar prediction models;
s7, sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
s8, calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
and S9, sequentially inputting the test set into the f blood sugar prediction models to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient obtained in the S8 to obtain a final blood sugar prediction value.
Preferably, the individual eating behavior of step S1 includes: normal diet, fasting behavior, ketogenic diet, high-sugar diet; the data acquisition period comprises: morning, noon, afternoon and evening; the physiological indicator data of the individual includes: the method comprises the steps that the weight, the blood pressure, the blood fat, the heart rate, the PPG electrocardiosignal and the ECG pulse signal of an individual are taken as original characteristic values reflecting the physiological characteristics of the individual, the physiological index data participate in blood sugar prediction, and N sample data sets are formed by data acquired for N times.
Preferably, the raw eigenvalues are arranged to form an N × K raw eigenvalue matrix:
each row in the original eigenvalue matrix represents that each data sample in the N data samples contains K eigenvalues, and a certain column represents the same type eigenvalue of the N data samples and is marked as y; normalizing each y by the formula:
wherein y 'represents a new eigenvalue normalized by the eigenvalue of the same type of a certain column representing N data samples, and a new eigenvalue matrix X' is formed after normalization:
in order to eliminate the dimensional influence between the indexes, data standardization processing is required to solve the comparability between the data indexes, which is beneficial to ensuring the accuracy of the subsequent blood glucose prediction.
Preferably, a certain column represents the same type of features of N data samplesThe new characteristic value after value normalization is y ', y' is equal to the IR N×1 Setting the correlation coefficient of the same type eigenvalue under different data samples in the new eigenvalue matrix X' as w 1 The column corresponding to y' represents that the smooth characteristic value obtained after the smoothing processing of the characteristic values of the same type of the N data samples is x, and x belongs to the IR N×1 Using x ═ y'. w 1 Smoothing the feature values y' of the same type of the N data samples, wherein the process is as follows:
establishing an objective function:
in that
At maximum, i.e.:
when x T y′ T w 1 >0，w 1 T y′y′ T w 1 C is a constant, and the correlation coefficient is solved to w 1 According to x ═ y'. w 1 Obtaining the column smooth characteristic value X, repeating the above processes until obtaining the smooth characteristic values of the same type characteristic values represented by all columns in the new characteristic value matrix to form a smooth characteristic value matrix X, wherein the table formula is as follows:
dividing the N smooth sample data sets after the N sample data sets are subjected to smoothing processing into a training set containing a smooth samples, a verification set containing b smooth samples and a test set containing c smooth samples, wherein N is a + b + c.
The characteristic value is smoothed by the correlation coefficient, so that the relationship between the characteristic value and the corresponding blood glucose measured value is more correlated.
Preferably, a total of n decision trees in the random forest model are set, and the importance value of the jth characteristic value in the ith decision tree is calculated, wherein the calculation expression is as follows:
wherein,
an importance value representing the jth smoothed feature value,
and expressing the importance value of the jth smooth characteristic value in the ith decision tree, wherein the expression is as follows:
wherein Gini m 、Gini l 、Gini r Respectively representing the node m, the node l and the node r in the random forest model, wherein the calculation expression of the node r of any one node is as follows:
gini represents the kini index of any node in the random forest model; k represents the number of the smooth characteristic values, and K represents the K-th smooth characteristic value calculated currently; p is a radical of mk Representing the proportion of the kth smooth characteristic value in the node m in the random forest model; the expression for the feature contribution ratio of the smoothed feature value is:
wherein K represents the number of smooth characteristic values; VIM j Representing the contribution of the jth smoothed feature value,
indicating the importance of all smooth feature valuesSumming;
let d denote the number of smooth eigenvalues greater than zero in each data sample, and select the original smooth eigenvalue matrix X composed of d eigenvalues to be expressed as X ″:
preferably, in step S5, the smooth feature value ranked at the top e bit is selected from the training set for extraction, and is characterized as: [ x ] of 1 ，x 2 ，…，x e ]The remaining smooth feature values are d-e characterized as [ x ] e+1 ，…，x d ]Equally divided into f sets characterized as:
combining the averaged f-group set and the e-bit smooth feature values into a new feature vector to form f feature matrixes X 1 ，…,X u ,…,X f Wherein any one of the feature matrices X u Expressed as:
feature matrix X 1 ，…，X u ，…，X f Has a number of measured blood glucose values corresponding to each other in synchronization, and in step S6, the feature matrix X is used 1 ，…,X u ,…,X f And inputting each feature matrix and a actually measured blood sugar values synchronously corresponding to the feature matrix into the random forest model respectively, wherein each data sample in the feature matrix takes the actual blood sugar value as output and each feature matrix as input, and takes the error between the output and the input as a training target, and when the training target is optimally converged, the random forest model corresponding to each feature matrix is trained to be used as f blood sugar prediction models.
Here, the feature values having different feature contribution rates are combined, and the influence of the feature values on the prediction result is sufficiently considered.
Preferably, in step S7, the verification set includes b data samples, and the process of sequentially inputting the verification set into the f blood glucose prediction models to obtain the first blood glucose prediction result matrix is as follows:
the verification set is ranked from large to small according to the feature contribution rate to obtain the top
Characteristic value of each data sample, forming
Is expressed as:
sequentially inputting the characteristic matrix into F blood sugar prediction models to obtain a first blood sugar prediction result matrix F of b multiplied by F, wherein the first blood sugar prediction result matrix F is expressed as:
wherein, the result of the prediction vector output by each blood sugar prediction model is expressed as: [ F ] h1 ,F h2 ，…,F hf ],h＝1,…,b；
Let the correlation coefficient of f blood sugar prediction models be w 2 Let r be a matrix of measured blood glucose values corresponding to the b smoothed sample validation sets, and in step S8, based on the first blood glucose prediction result matrix F, F ∈ IR b×f ，r∈IR b Using F.w 2 R to obtain w 2 ∈IR f Solving for the 2 norm
Correlation coefficient w at the time of obtaining minimum value 2 Calculated w 2 I.e. the optimum correlation coefficient w 2 * 。
Preferably, the specific process of step S9 is:
sequentially inputting a test set containing c smooth samples into f blood sugar prediction models, wherein a second prediction result matrix obtained by each blood sugar prediction model is characterized as follows: [ F ] g1 ,F g2 ，…,F gf ]，g＝1,…,c；
And combining the second prediction result matrixes obtained by each blood sugar prediction model to obtain an integral second blood sugar prediction result matrix, wherein the second blood sugar prediction result matrix is characterized by comprising the following steps:
optimum correlation coefficient w obtained by S8 2 * And fusing the integral second blood sugar prediction result matrix to obtain a final blood sugar prediction value, wherein the expression is as follows:
wherein Z1, …, Zc represents the final predicted blood glucose value corresponding to the test set containing c smooth samples.
In the method, the result matrixes obtained by different prediction models are fused by using the optimal correlation coefficient to obtain the final blood sugar prediction result, so that the deviation of behavior change on the prediction result is greatly reduced, and the accuracy of the blood sugar prediction result is improved.
The invention also provides a blood sugar prediction system based on smoothing and fusion, which is used for realizing the blood sugar prediction method and comprises the following steps:
the data acquisition unit is used for determining individual eating behaviors and data acquisition time periods, acquiring physiological index data of the individuals under different eating behaviors for N times as original characteristic values, and actually measuring blood sugar values corresponding to the physiological index data synchronously to serve as blood sugar reference values;
the normalization unit is used for forming an original characteristic value matrix based on the original characteristic values, and carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and form a new characteristic value matrix;
the smoothing unit is used for calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
the characteristic contribution rate calculating unit introduces a random forest model, calculates the characteristic contribution rate of the smooth characteristic value by using the random forest model, sorts the characteristic contribution rate and selects the smooth characteristic value with the characteristic contribution rate larger than zero;
the characteristic combination unit is used for selecting and extracting smooth characteristic values ranked at the front e bit from the training set to serve as a fixed set, uniformly dividing the residual smooth characteristic values into f sets, combining the uniformly divided f sets and the e bit smooth characteristic values into a new characteristic vector and forming f characteristic matrixes;
the training unit is used for training a random forest model by using the f characteristic matrixes and blood sugar reference values synchronously corresponding to the f characteristic matrixes to obtain f blood sugar prediction models;
the verification unit is used for sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
the correlation coefficient solving unit is used for calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
and the test unit is used for sequentially inputting the test set into the f blood sugar prediction models to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient to obtain a final blood sugar prediction value.
The invention also provides a medical device which comprises the blood sugar prediction system based on the smoothing and fusion.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides a blood sugar prediction method, a system and medical equipment based on smoothing processing and fusion, which are characterized in that original data characteristic values of individuals are collected from different eating behaviors, normalization processing is carried out on the original data characteristic values in order to eliminate dimensional influence among data, then the normalized characteristic values are smoothed through correlation coefficients, so that the relation between the characteristic values and corresponding blood sugar measured values is more relevant, a formed characteristic matrix is used as a data set and is divided into a training set, a verification set and a test set, the characteristic values with different characteristic contribution rates are combined, the influence of the characteristic values on blood sugar prediction results is fully considered, the optimal correlation coefficients are utilized to fuse the blood sugar prediction result matrixes obtained by different blood sugar prediction models, the final blood sugar prediction results are obtained, and the influence caused by behavior change is fully considered, the method can be packaged in a system, further forms the application of medical equipment, can provide guidance for a diabetic to self-predict the blood sugar and take self-intervention measures in daily life, and has practical significance.
Drawings
Fig. 1 is a schematic flow chart of a blood glucose prediction method based on smoothing and fusion according to embodiment 1 of the present invention;
FIG. 2 is a schematic diagram showing the process of forming f blood glucose prediction models proposed in embodiment 1 of the present invention;
FIG. 3 is a schematic diagram of a blood glucose prediction system based on smoothing and fusion according to embodiment 2 of the present invention;
fig. 4 is a schematic structural view of a medical apparatus proposed in embodiment 3 of the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for better illustration of the present embodiment, certain parts of the drawings may be omitted, enlarged or reduced, and do not represent actual dimensions;
it will be understood by those skilled in the art that certain well-known descriptions of the figures may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
The positional relationships depicted in the drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
example 1
As shown in fig. 1, the present embodiment proposes a blood glucose prediction method based on smoothing and fusion, and referring to fig. 1, the method is based on the consideration of poor blood glucose regulation capability of a diabetic, when the ingestion behavior of exogenous carbohydrates changes, large fluctuations of blood glucose in vivo may be caused, and blood glucose fluctuations may cause damage to body organs, accelerate the occurrence of complications, and consider how to approach the life of the diabetic, and design a way to make the diabetic self-predict blood glucose level and take self-intervention measures in life on the premise that the blood glucose prediction accuracy is guaranteed, which is the most fundamental, the method is developed around the following implementation steps:
s1, determining individual dietary behaviors and a data acquisition time period, acquiring physiological index data of N times of individuals under different dietary behaviors as original characteristic values, and actually measuring blood glucose values corresponding to the physiological index data synchronously to serve as blood glucose reference values;
wherein the individual eating behavior comprises: normal diet, fasting behavior, ketogenic diet, high-sugar diet; the data acquisition period comprises: morning, noon, afternoon and evening; the physiological index data of the individual includes: the method comprises the steps that the weight, the blood pressure, the blood fat, the heart rate, the PPG electrocardiosignals and the ECG pulse signals of an individual are taken as original characteristic values reflecting the physiological characteristics of the individual, blood sugar prediction is participated in, N times of collected data form N sample data sets, the N times of collected data are collected physiological index data (original characteristic values) of the individual under different eating behaviors, and blood sugar values corresponding to the measured physiological index data in a synchronous mode are obtained.
S2, forming an original characteristic value matrix based on the original characteristic values, and carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and form a new characteristic value matrix;
setting original eigenvalues to form an N multiplied by K original eigenvalue matrix:
each row in the original eigenvalue matrix represents that each data sample in the N data samples contains K eigenvalues, and a certain column represents the same type eigenvalue of the N data samples and is marked as y; considering that different physiological index data can affect the result of data analysis, in order to eliminate the dimensional influence between indexes, data standardization processing is needed to solve the comparability between data indexes, which is beneficial to ensuring the accuracy of subsequent blood glucose prediction, specifically, each y is normalized, and the normalization formula is as follows:
wherein y 'represents a new eigenvalue normalized by the eigenvalue of the same type of a certain column representing N data samples, and a new eigenvalue matrix X' is formed after normalization:
s3, calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
smoothing the characteristic value through a correlation coefficient to enable the relation between the characteristic value and the corresponding blood sugar measured value to be more relevant, wherein a certain column represents the new characteristic value of N data samples after the characteristic values of the same type are normalized to be y ', y' belongs to the IR N×1 The specific method comprises the following steps:
let the correlation coefficient of the same type eigenvalue under different data samples in the new eigenvalue matrix X' be w 1 The column corresponding to y' represents that the smooth characteristic value obtained after the smoothing processing of the characteristic values of the same type of the N data samples is x, and x belongs to the IR N×1 Using x ═ y' · w 1 Same type of N data samplesSmoothing the characteristic value y', wherein the process is as follows:
establishing an objective function:
in that
At maximum, i.e.:
when x T y′ T w 1 >0，w 1 T y′y′ T w 1 When C is a constant, the correlation coefficient is solved to w 1 According to x ═ y' · w 1 Finding the column of smoothed eigenvalues x, more specifically:
is provided with
S B ＝y′y′ T Are combined with
To obtain
Is provided with
Due to S A And S B All are full rank matrices, then Q ═ uv can be made T ,uv T ξ＝λξ，u＝ξ,v T ξ＝λ
Is provided with
v＝Q(1,:) T
By combining C and D, w can be solved 1 And according to x ═ y' · w 1 And calculating x.
Repeating the above processes until the smooth eigenvalues of the same type represented by all columns in the new eigenvalue matrix are solved to form a smooth eigenvalue matrix X, wherein the table formula is as follows:
dividing the N smooth sample data sets after the N sample data sets are subjected to smoothing processing into a training set containing a smooth samples, a verification set containing b smooth samples and a test set containing c smooth samples, wherein N is a + b + c.
S4, introducing a random forest model, calculating the characteristic contribution rate of the smooth characteristic values by using the random forest model, sequencing, and selecting the smooth characteristic values with the characteristic contribution rate larger than zero;
in this embodiment, a random forest model is constructed by a self-help resampling technique, a plurality of training set samples are repeatedly and randomly extracted from a training set in a replaced manner by using a bootstrap method to generate new training set samples, a decision tree is established for each new training set sample, a prediction result of each decision tree of the random forest model is used as a prediction vote, and a minority obedience majority is used as a prediction result.
Calculating the importance value of the jth characteristic value in the ith decision tree, wherein the calculation expression is as follows:
wherein,
an importance value representing the jth smoothed feature value,
to representThe importance value of the jth smooth characteristic value in the ith decision tree is expressed as:
wherein Gini m 、Gini l 、Gini r Respectively representing the node m, the node l and the node r in the random forest model, wherein the calculation expression of the node r of any one node is as follows:
gini represents the kini index of any node in the random forest model; k represents the number of the smooth characteristic values, and K represents the K-th smooth characteristic value calculated currently; p is a radical of mk Representing the proportion of the kth smooth characteristic value in the node m in the random forest model; the expression for the feature contribution ratio of the smoothed feature value is:
wherein K represents the number of smooth characteristic values; VIM j Representing the contribution of the jth smoothed feature value,
representing the sum of the importance of all the smooth feature values;
setting d to represent the number of the smooth characteristic values in each data sample larger than zero, and selecting an original smooth characteristic value matrix X consisting of the d characteristic values to represent as X':
s5, selecting and extracting smooth characteristic values ranked at the front e position from the training set to serve as a fixed set, uniformly dividing the remaining smooth characteristic values into f sets, combining the uniformly divided f sets and the e-position smooth characteristic values into a new characteristic vector, and forming f characteristic matrixes;
in this embodiment, feature values with different feature contribution rates are combined, the influence of the feature values on the prediction result is fully considered, and in step S5, the smooth feature values ranked at the top e are selected from the training set and extracted, and are characterized as a fixed set: [ x ] of 1 ，x 2 ，…，x e ]The remaining smooth feature values are d-e characterized as [ x ] e+1 ，…，x d ]Equally dividing into f groups of sets, and characterizing as:
Combining the f groups of the averaged set and the e-bit smooth characteristic values into a new characteristic vector to form f characteristic matrixes X 1 ，…,X u ,…,X f Wherein any one of the feature matrices X u Expressed as:
feature matrix X 1 ，…，X u ，…，X f Each of the samples has a synchronously corresponding a actual blood sugar value, that is, assuming that N is 10, each sample contains 14 feature values, taking the first 5 feature values as a fixed set, averagely dividing the remaining 9 feature values into 3 groups, and recombining into 3 feature matrices:
s6, training a random forest model by using the f feature matrices and blood sugar reference values synchronously corresponding to the f feature matrices to obtain f blood sugar prediction models;
as shown in FIG. 2, the feature matrix X 1 ，…,X u ,…,X f Respectively inputting each feature matrix and a actually measured blood sugar values synchronously corresponding to the feature matrix into a randomAnd in the forest model, each data sample in the feature matrix takes an actually measured blood sugar value as output, each feature matrix as input, the error between the output and the input is taken as a training target, and when the training target is optimally converged, the random forest model corresponding to each feature matrix is trained to be used as f blood sugar prediction models.
S7, sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
b data samples are contained in the verification set, the verification set is sequentially input into f blood sugar prediction models, and the process of obtaining a first blood sugar prediction result matrix is as follows:
the verification set is ranked from large to small according to the feature contribution rate to obtain the top
Characteristic value of each data sample, forming
Is expressed as:
sequentially inputting the characteristic matrix into F blood sugar prediction models to obtain a first blood sugar prediction result matrix F of b multiplied by F, wherein the first blood sugar prediction result matrix F is expressed as:
the result of the prediction vector output by each blood sugar prediction model is expressed as: [ F ] h1 ,F h2 ，…,F hf ],h＝1,…,b；
S8, calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
let the correlation coefficient of f blood sugar prediction models be w 2 Let r be the actual measurement corresponding to b smooth sample verification setsThe matrix of blood glucose values is based on the first blood glucose prediction result matrix F, F ∈ IR in step S8 b×f ，r∈IR b Using F.w 2 R to obtain w 2 ∈IR f Solving for the 2 norm
Correlation coefficient w at the time of obtaining minimum value 2 W obtained by 2 I.e. the optimum correlation coefficient w 2 * Specifically, the method comprises the following steps:
let J (w) 2 )＝w 2 T F T Fw 2 -2r T Fw 2 +r T r, order
Get w 2 * ＝(F T F) -1 Fr。
And S9, sequentially inputting the test set into the f blood sugar prediction models to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient obtained in the S8 to obtain a final blood sugar prediction value.
When the step is implemented, the optimal correlation coefficient is utilized to fuse result matrixes obtained by different prediction models to obtain a final blood sugar prediction result, so that the deviation of behavior change on the prediction result can be reduced, and the accuracy of the blood sugar prediction result is improved, wherein the specific process comprises the following steps of:
sequentially inputting a test set containing c smooth samples into f blood sugar prediction models, wherein a second prediction result matrix obtained by each blood sugar prediction model is characterized as follows: [ F ] g1 ,F g2 ，…,F gf ]，g＝1,…,c；
And combining the second prediction result matrixes obtained by each blood sugar prediction model to obtain an integral second blood sugar prediction result matrix, wherein the second blood sugar prediction result matrix is characterized by comprising the following steps:
optimum correlation obtained by S8Number w 2 * And fusing the integral second blood sugar prediction result matrix to obtain a final blood sugar prediction value, wherein the expression is as follows:
wherein Z1, …, Zc represents the final predicted blood glucose value corresponding to the test set containing c smooth samples.
Example 2
As shown in fig. 3, the present embodiment provides a blood glucose prediction system based on smoothing and fusion, the system is used to implement the blood glucose prediction method described in embodiment 1, and referring to fig. 3, the system includes:
the data acquisition unit 1 is used for determining individual eating behaviors and data acquisition time periods, acquiring physiological index data of the individual under different eating behaviors for N times as an original characteristic value, and actually measuring blood sugar values synchronously corresponding to the physiological index data to serve as blood sugar reference values;
the normalization unit 2 is used for forming an original characteristic value matrix based on the original characteristic values, carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and forming a new characteristic value matrix;
the smoothing unit 3 is used for calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
the characteristic contribution rate calculating unit 4 introduces a random forest model, calculates the characteristic contribution rate of the smooth characteristic value by using the random forest model, sorts the characteristic contribution rate, and selects the smooth characteristic value with the characteristic contribution rate larger than zero;
the feature combination unit 5 is used for selecting and extracting smooth feature values ranked at the front e bit from the training set to serve as a fixed set, uniformly dividing the remaining smooth feature values into f sets, combining the uniformly divided f sets and the e bit smooth feature values into a new feature vector and forming f feature matrices;
the training unit 6 is used for training a random forest model by using the f feature matrices and blood sugar reference values synchronously corresponding to the f feature matrices to obtain f blood sugar prediction models;
the verification unit 7 is used for sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
the correlation coefficient solving unit 8 is used for calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
and the test unit 9 is used for sequentially inputting the test set into the f blood sugar prediction models to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient to obtain a final blood sugar prediction value.
In the system provided by this embodiment, each unit is used to package the algorithm implementation process corresponding to the method provided by embodiment 1, and then the system is further integrated, so that the auxiliary application in the medical field can be realized.
Example 3
Referring to fig. 4, the present embodiment proposes a medical device, which includes the blood glucose prediction system based on smoothing and fusion proposed in embodiment 2, and the blood glucose prediction result performed by the medical device in the daily life of the diabetic can provide guidance for the diabetic to self-monitor the blood glucose level and take self-intervention measures in the daily life.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (7)
1. A blood glucose prediction method based on smoothing and fusion is characterized by comprising the following steps:
s1, determining individual dietary behaviors and a data acquisition time period, acquiring physiological index data of N times of individuals under different dietary behaviors as original characteristic values, and actually measuring blood glucose values corresponding to the physiological index data synchronously to serve as blood glucose reference values;
s2, forming an original characteristic value matrix based on the original characteristic values, and carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and form a new characteristic value matrix;
s3, calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
setting the original characteristic values to form an N multiplied by K original characteristic value matrix:
each row in the original eigenvalue matrix represents that each data sample in the N data samples contains K eigenvalues, and a certain column represents the same type eigenvalue of the N data samples and is marked as y; normalizing each y by the formula:
wherein y 'represents a new eigenvalue normalized by the eigenvalue of the same type of a certain column representing N data samples, and a new eigenvalue matrix X' is formed after normalization:
the new characteristic value of a certain column representing the normalized characteristic values of the same type of N data samples is y ', y' is equal to the IR N×1 Setting the correlation coefficient of the same type eigenvalue under different data samples in the new eigenvalue matrix X' as w 1 The column corresponding to y' represents N data samplesThe smooth characteristic value obtained after the same type of characteristic value smoothing is x, x belongs to IR N×1 Using x ═ y' · w 1 Smoothing the feature values y' of the same type of the N data samples, wherein the process is as follows:
establishing an objective function:
in that
At maximum, i.e.:
when x T y′ T w 1 >0，w 1 T y′y′ T w 1 C is a constant, and the correlation coefficient is solved to w 1 According to x ═ y'. w 1 Obtaining the column smooth characteristic value X, repeating the above processes until obtaining the smooth characteristic values of the same type characteristic values represented by all columns in the new characteristic value matrix to form a smooth characteristic value matrix X, wherein the table formula is as follows:
dividing N smooth sample data sets after smoothing processing of N sample data sets into a training set containing a smooth samples, a verification set containing b smooth samples and a test set containing c smooth samples, wherein N is a + b + c;
s4, introducing a random forest model, calculating and sequencing the feature contribution rate of the smooth feature value by using the random forest model, and selecting the smooth feature value with the feature contribution rate larger than zero;
s5, selecting and extracting smooth characteristic values ranked at the front e position from the training set to serve as a fixed set, uniformly dividing the remaining smooth characteristic values into f sets, combining the uniformly divided f sets and the e-position smooth characteristic values into a new characteristic vector, and forming f characteristic matrixes;
s6, training a random forest model by using the f feature matrices and blood glucose reference values synchronously corresponding to the f feature matrices to obtain f blood glucose prediction models;
s7, sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
s8, calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
s9, inputting the test set into f blood sugar prediction models in sequence to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient obtained in the S8 to obtain a final blood sugar prediction value;
let the correlation coefficient of f blood sugar prediction models be w 2 Let r be a matrix of measured blood glucose values corresponding to the b smoothed sample validation sets, and in step S8, based on the first blood glucose prediction result matrix F, F ∈ IR b×f ，r∈IR b Using F.w 2 R to obtain w 2 ∈IR f Solving for the 2 norm
Correlation coefficient w at the time of obtaining minimum value 2 W obtained by 2 I.e. the optimum correlation coefficient w 2 * ；
The specific process of step S9 is:
sequentially inputting a test set containing c smooth samples into f blood sugar prediction models, wherein a second prediction result matrix obtained by each blood sugar prediction model is characterized as follows: [ F ] g1 ,F g2 ，…,F gf ]，g＝1,…,c；
And combining the second prediction result matrixes obtained by each blood sugar prediction model to obtain an integral second blood sugar prediction result matrix, wherein the second blood sugar prediction result matrix is characterized by comprising the following steps:
optimum correlation coefficient w obtained by S8 2 * Second blood glucose prediction result matrix for the whole bodyAnd fusing to obtain a final blood sugar predicted value, wherein the expression is as follows:
wherein Z1, …, Zc represents the final predicted blood glucose value corresponding to the test set containing c smooth samples.
2. The smoothing and fusion-based blood glucose prediction method of claim 1, wherein the individual eating behavior of step S1 comprises: normal diet, fasting behavior, ketogenic diet, high-sugar diet; the data acquisition period comprises: morning, noon, afternoon and evening; the physiological indicator data of the individual includes: the method comprises the steps that the weight, the blood pressure, the blood fat, the heart rate, the PPG electrocardiosignal and the ECG pulse signal of an individual are taken as original characteristic values reflecting the physiological characteristics of the individual, the physiological index data participate in blood sugar prediction, and N sample data sets are formed by data acquired for N times.
3. The blood sugar prediction method based on smoothing and fusion as claimed in claim 2, wherein n decision trees in the random forest model are set, and the importance value of the jth eigenvalue in the ith decision tree is calculated by the following calculation expression:
wherein,
an importance value representing the jth smoothed feature value,
and expressing the importance value of the jth smooth characteristic value in the ith decision tree, wherein the expression is as follows:
wherein Gini m 、Gini l 、Gini r Respectively representing the node m, the node l and the node r in the random forest model, wherein the calculation expression of the node r of any one node is as follows:
gini represents the kini index of any node in the random forest model; k represents the number of the smooth characteristic values, and K represents the K-th smooth characteristic value calculated currently; p is a radical of mk Representing the proportion of the kth smooth characteristic value in the node m in the random forest model; the expression for the feature contribution ratio of the smoothed feature value is:
wherein K represents the number of smooth characteristic values; VIM j Representing the contribution of the jth smoothed feature value,
representing the sum of the importance of all the smooth feature values;
setting d to represent the number of the smooth characteristic values in each data sample larger than zero, and selecting an original smooth characteristic value matrix X consisting of the d characteristic values to represent as X':
4. the smoothing and fusion-based blood glucose prediction method of claim 3, wherein in step S5, the training is performed from trainingAnd (3) centrally selecting smooth characteristic values ranked at the top e bit for extraction, and characterizing as a fixed set: [ x ] of 1 ，x 2 ，…，x e ]The remaining smooth feature values are d-e characterized as [ x ] e+1 ，…，x d ]Equally dividing into f groups of sets, and characterizing as:
Combining the averaged f-group set and the e-bit smooth feature values into a new feature vector to form f feature matrixes X 1 ，…,X u ,…,X f Wherein any one of the feature matrices X u Expressed as:
feature matrix X 1 ，…，X u ，…，X f Has a number of measured blood glucose values corresponding to each other in synchronization, and in step S6, the feature matrix X is used 1 ，…,X u ,…,X f And inputting each feature matrix and a actually measured blood sugar values synchronously corresponding to the feature matrix into the random forest model respectively, wherein each data sample in the feature matrix takes the actual blood sugar value as output and each feature matrix as input, and takes the error between the output and the input as a training target, and when the training target is optimally converged, the random forest model corresponding to each feature matrix is trained to be used as f blood sugar prediction models.
5. The method of claim 4, wherein in step S7, the verification set comprises b data samples, and the first blood glucose prediction matrix is obtained by sequentially inputting the verification set into f blood glucose prediction models by:
the verification set is ranked from large to small according to the feature contribution rate to obtain the top
Characteristic value of each data sample, forming
Is expressed as:
sequentially inputting the characteristic matrix into F blood sugar prediction models to obtain a first blood sugar prediction result matrix F of b multiplied by F, wherein the first blood sugar prediction result matrix F is expressed as follows:
wherein, the result of the prediction vector output by each blood sugar prediction model is expressed as: [ F ] h1 ,F h2 ，…,F hf ],h＝1,…,b。
6. A blood glucose prediction system based on smoothing and fusion is characterized in that the system is used for realizing the blood glucose prediction method of any one of claims 1-5, and comprises the following steps:
the data acquisition unit is used for determining individual eating behaviors and data acquisition time periods, acquiring physiological index data of the individuals under different eating behaviors for N times as original characteristic values, and actually measuring blood sugar values corresponding to the physiological index data synchronously to serve as blood sugar reference values;
the normalization unit is used for forming an original characteristic value matrix based on the original characteristic values, and carrying out normalization processing on the original characteristic values in the original characteristic value matrix to obtain new characteristic values and form a new characteristic value matrix;
the smoothing unit is used for calculating a correlation coefficient of the new characteristic value, smoothing the new characteristic value based on the correlation coefficient to obtain a smooth characteristic value, and dividing the smooth characteristic value into a training set, a verification set and a test set;
the characteristic contribution rate calculating unit introduces a random forest model, calculates the characteristic contribution rate of the smooth characteristic value by using the random forest model, sorts the characteristic contribution rate and selects the smooth characteristic value with the characteristic contribution rate larger than zero;
the characteristic combination unit is used for selecting and extracting smooth characteristic values ranked at the front e bit from the training set to serve as a fixed set, uniformly dividing the rest smooth characteristic values into f sets, combining the uniformly divided f sets and the e bit smooth characteristic values into a new characteristic vector and forming f characteristic matrixes;
the training unit is used for training a random forest model by using the f characteristic matrixes and the blood sugar reference values synchronously corresponding to the characteristic matrixes to obtain f blood sugar prediction models;
the verification unit is used for sequentially inputting the verification set into f blood sugar prediction models to obtain a first blood sugar prediction result matrix;
the correlation coefficient solving unit is used for calculating the optimal correlation coefficients of the f blood sugar prediction models based on the first blood sugar prediction result matrix;
and the test unit is used for sequentially inputting the test set into the f blood sugar prediction models to obtain a second blood sugar prediction result matrix, and fusing the second blood sugar prediction result matrix by using the optimal correlation coefficient to obtain a final blood sugar prediction value.
7. A medical device comprising the smoothing and fusion based blood glucose prediction system of claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210109388.3A CN114530228B (en) | 2022-01-28 | 2022-01-28 | Blood glucose prediction method and system based on smoothing and fusion and medical equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210109388.3A CN114530228B (en) | 2022-01-28 | 2022-01-28 | Blood glucose prediction method and system based on smoothing and fusion and medical equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114530228A CN114530228A (en) | 2022-05-24 |
| CN114530228B true CN114530228B (en) | 2022-09-27 |
Family
ID=81622594
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210109388.3A Active CN114530228B (en) | 2022-01-28 | 2022-01-28 | Blood glucose prediction method and system based on smoothing and fusion and medical equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114530228B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117612737B (en) * | 2024-01-24 | 2024-05-03 | 胜利油田中心医院 | Intelligent optimization method for diabetes care data |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831326A (en) * | 2012-09-06 | 2012-12-19 | 南方医科大学 | Mean amplitude of glucose excursions (MAGE) calculation method |
| CN106446566A (en) * | 2016-09-29 | 2017-02-22 | 北京理工大学 | Elderly cognitive function classification method based on random forest |
| CN107403072A (en) * | 2017-08-07 | 2017-11-28 | 北京工业大学 | A kind of diabetes B prediction and warning method based on machine learning |
| CN107463766A (en) * | 2017-06-23 | 2017-12-12 | 深圳市中识创新科技有限公司 | Generation method, device and the computer-readable recording medium of blood glucose prediction model |
| CN109378072A (en) * | 2018-10-13 | 2019-02-22 | 中山大学 | A kind of abnormal fasting blood sugar method for early warning based on integrated study Fusion Model |
| CN111898666A (en) * | 2020-07-23 | 2020-11-06 | 中南大学 | Random forest algorithm and module population combined data variable selection method |
-
2022
- 2022-01-28 CN CN202210109388.3A patent/CN114530228B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831326A (en) * | 2012-09-06 | 2012-12-19 | 南方医科大学 | Mean amplitude of glucose excursions (MAGE) calculation method |
| CN106446566A (en) * | 2016-09-29 | 2017-02-22 | 北京理工大学 | Elderly cognitive function classification method based on random forest |
| CN107463766A (en) * | 2017-06-23 | 2017-12-12 | 深圳市中识创新科技有限公司 | Generation method, device and the computer-readable recording medium of blood glucose prediction model |
| CN107403072A (en) * | 2017-08-07 | 2017-11-28 | 北京工业大学 | A kind of diabetes B prediction and warning method based on machine learning |
| CN109378072A (en) * | 2018-10-13 | 2019-02-22 | 中山大学 | A kind of abnormal fasting blood sugar method for early warning based on integrated study Fusion Model |
| CN111898666A (en) * | 2020-07-23 | 2020-11-06 | 中南大学 | Random forest algorithm and module population combined data variable selection method |
Non-Patent Citations (2)
| Title |
|---|
| Design of Near Allpass Strictly Stable Minimal Phase Real Valued Rational IIR Filters;Charlotte Yuk-Fan Ho等;《IEEE Transactions on Circuits and Sytem II:Express Briefs》;20080801;第55卷(第8期);第781-785页 * |
| 基于机器学习的糖尿病预测模型的研究;张熙;《中国优秀硕士学位论文全文数据库-医学卫生科技辑》;20210915(第9期);第E065-29页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114530228A (en) | 2022-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US6540686B2 (en) | Measurement relating to human body | |
| US20210338135A1 (en) | Determining device and mapping system for origin of arrhythmia | |
| CN109758160B (en) | LSTM-RNN model-based noninvasive blood glucose prediction method | |
| US20230107787A1 (en) | Blood pressure prediction method and device | |
| US6416473B1 (en) | Methods and apparatus for providing an indicator of autonomic nervous system function | |
| CN114548158A (en) | Data processing method for blood sugar prediction | |
| CN109431492A (en) | ECG lead signals based on neural network algorithm simulate method for reconstructing | |
| CN113509186B (en) | ECG classification system and method based on deep convolutional neural network | |
| CN114530228B (en) | Blood glucose prediction method and system based on smoothing and fusion and medical equipment | |
| WO2021184802A1 (en) | Blood pressure classification prediction method and apparatus | |
| CN114420301B (en) | Method, system and storage medium for predicting blood glucose based on segmented domain RF modeling | |
| Zhang et al. | A hybrid model for blood pressure prediction from a PPG signal based on MIV and GA-BP neural network | |
| CN1275179C (en) | Computer evaluating method for human body sub-health status | |
| TWI688371B (en) | Intelligent device for atrial fibrillation signal pattern acquisition and auxiliary diagnosis | |
| CN116434979A (en) | Physiological state cloud monitoring method, monitoring system and storage medium | |
| CN114403866B (en) | Noninvasive blood sugar prediction device based on near-infrared light wavelength conversion | |
| US20230309866A1 (en) | Blood sugar level estimation device, blood sugar level estimation method, and computer program | |
| CN113066547A (en) | ARDS early dynamic early warning method and system based on conventional noninvasive parameters | |
| EP4383274A1 (en) | Hba1c risk estimation device, hba1c risk estimation method, and program | |
| CN116959742B (en) | Blood glucose data processing method and system based on spherical coordinate kernel principal component analysis | |
| US20240008814A1 (en) | Blood neutral fat estimation device, blood neutral fat estimation method, and computer program | |
| CN113782210B (en) | Method for predicting treatment failure probability of noninvasive ventilator | |
| EP4383275A1 (en) | Creatinine risk estimation device, creatinine risk estimation method, and program | |
| US20230317218A1 (en) | Uric acid level estimation device, uric acid level estimation method, and computer program | |
| US20240156377A1 (en) | yGT ESTIMATION DEVICE, yGT ESTIMATION METHOD, AND COMPUTER PROGRAM |
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 |