CN107169284A - A kind of biomedical determinant attribute system of selection - Google Patents
A kind of biomedical determinant attribute system of selection Download PDFInfo
- Publication number
- CN107169284A CN107169284A CN201710332543.7A CN201710332543A CN107169284A CN 107169284 A CN107169284 A CN 107169284A CN 201710332543 A CN201710332543 A CN 201710332543A CN 107169284 A CN107169284 A CN 107169284A
- Authority
- CN
- China
- Prior art keywords
- attribute
- mrow
- intervention effect
- data
- determinant
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention relates to a kind of biomedical determinant attribute system of selection, belong to field of biomedicine technology.The present invention extracts the important attribute of influence goal in research first by the importance of boruta Algorithm Analysis attribute to be selected;Then attribute construction logic regression model to be selected is used, successive Regression is carried out using AIC criterion, the attribute being had a significant impact to goal in research is obtained;Attribute is obtained for two methods screening, with reference to expert opinion, the method sorted out using occuring simultaneously carries out attribute fusion, obtains final determinant attribute.The present invention is selected the attribute for influenceing goal in research using two distinct methods, and difference is obvious between algorithm, it is to avoid the limitation that single method is brought, and improves the generalization of determinant attribute.
Description
Technical field
The present invention relates to a kind of biomedical determinant attribute system of selection.Said from the angle of application, belong to biological doctor
Learn technical field;For the angle realized from technology, computer science and Bioinformatics technical field are also belonged to.
Background technology
In recent years, biomedical research is developed rapidly, particularly the raising of measuring instrument technology and hospital information system
Promote so that a large amount of medical informations can be recorded accurately, so as to cause medical data data explosive increase.But
While a large amount of data for enriching complexity bring sufficient data to research work, also the analyzing and processing to researcher is brought more
Big challenge.Data mining overall process includes data acquisition, data prediction, knowledge excavation, model evaluation and knowledge application,
It can be good at handling mass data, and can be from potentially useful knowledge is extracted, just because of the characteristic, data mining technology is from carrying
Go out and be just applied to field of biomedical research, and achieved considerable success.
For field of biomedical research, often cause original without specific specific goal in research during information gathering
A large amount of attributes are included in beginning data set, it is necessary to carry out Attributions selection to initial data before data analysis, obtaining one has
Representational attribute set, its main purpose has:Uncorrelated attribute, redundant attributes are removed, storage efficiency is improved;Remove synteny
Attribute and noise attribute, reduce the interference and influence on data analysis;Improve the Generalization Capability and operational efficiency of model;Obtain more
Plus the learning model simply and readily understood, improve the interpretation of model.
Attributions selection is carried out during biomedical research, although had the method for many Attributions selections now, but
It is and is not present a kind of attribute selection method suitable for any problem.Feature Selection Algorithm is distinguished according to evaluation criterion, mainly
It is divided into two classes:
1. filtering type Attributions selection (Filter)
Filtering type Attributions selection is a kind of higher method of computational efficiency, and it is obtained by the inwardness of data set in itself
Evaluation criterion, it is unrelated with specific learning algorithm, with preferable versatility.The evaluation criterion of filtering type Attributions selection is divided into four
Class:I.e. distance metric, measure information, the degree of association are measured and consistency metric.
(1) distance metric includes geometric distance measurement and probability metrics measurement.Wherein, the evaluation criterion of geometric distance is usual
It is divided into class and class scatter matrix.Scatter Matrix represents that each sample point surrounds the distribution situation of average, class scatter square in class
In the distribution situation in space between each quasi-mode of matrix representation.The result of Attributions selection should make the mark of scatter matrix in class smaller more
Good, the mark of scatter matrix is the bigger the better between class.Evaluation criterion based on probability metrics has Kullback-Leibler distances, also known as
Relative entropy, weighs the difference condition of two probability distribution in similar events space, because it needs the general of each known classification
Rate density function, therefore with significant limitation.
(2) measure information is to use the evaluation criterion based on entropy, such as minimum description length (Minimum in information theory
Description length), mutual information (Mutual Information), information gain (Information Gain) etc..
These evaluation criterions describe the complexity of attribute, represent the size that attribute includes information content, and Attributions selection often selects complexity
More attribute.
(3) correlation mainly investigates the degree of association between attribute, i.e. correlation and redundancy.Wherein linear correlation is wired
Property coefficient correlation (Pearson's coefficient and Spearman's correlation coefficient) etc., nonlinear dependence is associated with the mutual information based on comentropy, right
Uncertainty of title etc..
(4) consistency metric attempts to find the minimal feature subset with complete or collected works' same category ability, and inconsistency is defined as
If in selected character subset value identical sample, but belonging to different classes.
2. packaged type Attributions selection (Wrapper)
Packaged type Attributions selection comes the quality of evaluation attributes subset, that is, package method using the performance indications of learning algorithm
Grader is trained using attribute set to be evaluated, the performance further according to grader is evaluated the attribute set.
The package method learning algorithm good and bad to evaluation attributes subset is diversified, is wanted for the learning algorithm used
Ask very low, most of sorting algorithm can use package method, for example decision tree, neutral net, Bayes classifier, support to
Amount machine and nearest neighbour method etc..
The highly versatile of filtration method, eliminates model training step, and algorithm complex is low, it is adaptable to large-scale dataset,
A large amount of uncorrelated attributes can be quickly removed, but this method, independently of specific learning algorithm, classification accuracy is relatively low.And seal
The attribute set classification performance that although dress method obtains is preferable, but feature versatility is not strong, and algorithm computation complexity is higher,
For large-scale dataset, the execution time of algorithm is very long.
In summary, existing Feature Selection Algorithm selects attribute set only for specific evaluation index, it is impossible to take into account general
Property and algorithm complex, it is relatively low to the treatment effeciency of large-scale dataset;And the effect of Attributions selection is carried out to biomedical data
Fruit is also not satisfactory, still treats further raising.
The content of the invention
The problem of purpose of the present invention is to solve Attributions selection in biomedical data, proposes a kind of based on boruta-
The attribute selection method of logistic regression.
The present invention design principle be:First, using the importance of boruta Algorithm Analysis attribute to be selected, influence is extracted
The important attribute of goal in research;Then attribute construction logic regression model to be selected is used, successive Regression is carried out using AIC criterion,
Obtain having a significant impact attribute to goal in research;Determinant attribute is obtained for two methods screening, with reference to expert opinion, using friendship
The method that collection is sorted out carries out attribute fusion, obtains final determinant attribute.The present invention is studied influence using two distinct methods
The attribute of target is selected, and difference is obvious between algorithm, it is to avoid the limitation that single method is brought, and improves the general of determinant attribute
The property changed.
The technical scheme is that be achieved by the steps of:
Step 1, data set S includes N number of sample, and the attribute to be selected of M dimension influence diabetes B intervention effects is used
Boruta algorithms are fitted to the intervention effect of crowd, obtain influenceing the determinant attribute of intervention effect, and attribute to be selected
Importance ranking, concrete methods of realizing is:
Step 1.1, data set S copy attribute is created, rearrangement of going forward side by side obtains recombination data collection S', increases data-oriented
The randomness of collection;
Step 1.2, the number for building tree is n (0<N≤N) in data set S', put back to ground using bootstrap modes
N number of new sample set is randomly selected, the data volume of each sample set is about 2N/3, and each sample set is registered as Di (0<i≤
N), the data record not being extracted is the outer data Oi (0 of bag<I≤n), build the Random Forest model for including n tree;
Step 1.3, post-class processing is built, calculating each sets the outer data mean square error of corresponding bag, is designated as MSEi, then n
The outer data mean square error vector of original bag of tree is [MSE1,MSE2,…,MSEn];
Step 1.4, the mean square error vector [MSE obtained based on step 2.21,MSE2,…,MSEn], with corresponding attribute meter
Obtained Z values, determine the maximum copy attribute of Z values, while the attribute bigger than its value is filtered out as important attribute, and it is right
Answer the small attribute of its Z value to be labeled as insignificant attribute, and deleted from data set;
Step 1.5, above step 1.1 to 1.4 is repeated, until reaching default end condition;
Step 1.6, the mean square error of attribute represents the importance of attribute, the forward attribute of Importance of Attributes sequence, for 2
Patients with type Ⅰ DM intervention effect has the attribute of material impact.
Step 2, based on raw data set S, discrimination model is returned according to the intervention effect construction logic of crowd, logic is used
Regression algorithm is fitted to the intervention effect of crowd, obtains having the attribute significantly affected, the side of implementing to intervention effect
Method is:
Step 2.1, tie up attribute to be selected to M to be normalized, computational methods are:
Wherein,It is i-th of sample l dimension attribute original value,It is the normalized value of i-th of sample l dimension attribute,WithIt is then the maximum and minimum value of sample l dimension attributes respectively;
Step 2.2, construction logic regression model, each attribute coefficients, regression equation are calculated using maximum Likelihood
It is as follows:
F (x)=b0+b1x1+b2x2+...+bMxM
Wherein, xl(0<L≤M) l dimension attributes are represented, M is the dimension of attribute, bl(0<L≤M) represent Logic Regression Models
In each attribute weight;
Step 2.3, successive Regression is carried out using AIC criterion, obtains combinations of attributes during AIC minimums, construction logic is returned
Model;
Step 2.4, significance test is carried out to M dimension attributes weight in Logic Regression Models, using significance as 0.05,
Screen has the attribute significantly affected to intervention effect.
Step 3, obtain patrolling in the attribute that there is intervention effect material impact, step 2 based on boruta algorithms in step 1
Collect regression algorithm and obtain that there is the attribute significantly affected on intervention effect, with reference to expert opinion, the method sorted out using occuring simultaneously is obtained
To the determinant attribute of influence intervention effect, concrete methods of realizing is:
Step 3.1, with reference to expertise opinion, it is considered to the collection difficulty of attribute, selection in step 1 and step 2 is obtained
Attribute further screen;
Step 3.2, the attribute to intervention effect with material impact, and logistic regression algorithm are obtained based on boruta algorithms
Obtain that there is the attribute significantly affected on intervention effect, the attribute that selection occurs simultaneously, crucial as influence intervention effect belongs to
Property.
Beneficial effect
Biomedical determinant attribute system of selection proposed by the present invention based on boruta algorithms and logistic regression algorithm, leads to
Cross logistic regression and obtain effect degree of the different attribute to intervention effect, selection is with the attribute significantly affected, and boruta is calculated
Method can quantify effect sensitivity level of the different attribute to intervention effect, and the determinant attribute that comprehensive two methods are obtained both ensured
The credibility of determinant attribute, causes determinant attribute to have difference between material impact, algorithm obvious to intervention effect again, it is to avoid
The limitation that single method is brought, instructs measure to provide instruction for specific aim adjustment intervention.
Brief description of the drawings
Fig. 1 is biomedical determinant attribute system of selection schematic diagram proposed by the present invention;
Fig. 2 is in embodiment, the Importance of Attributes based on boruta algorithms sorts.
Embodiment
In order to better illustrate objects and advantages of the present invention, with reference to the accompanying drawings and examples to the reality of the inventive method
The mode of applying is described in further details.
All tests are completed on same computer below, and concrete configuration is:Intel double-core CPU (dominant frequency
2.53G), 4G internal memories, the operating systems of Windows 7.
Test is strengthened using data source from the diabetes B high-risk individuals cooperated in Beijing Hospital's Gerontological Research Center
Lifestyle modification manages data, and group state that enters to interviewee in intervention group carries out RSD judgements, selects into group people at highest risk's number
According to totally 452 samples.
RSD risk status judgements are carried out again to intervening data after half a year, and it is to be divided into 0 class, risk status to keep high-risk
What is declined is divided into 1 class (i.e. intervening measure is effective), and as the label of two classification based training data sources, totally 20 dimension different attribute, is wrapped
Including 7 dimension intervening measures, (total run duration, effective exercise amount, effective exercise time, effective exercise number of times, effective dose are than body weight, reality
Border intake, aequum), 10 tie up (body weight, BMI, waistline, systolic pressure, diastolic pressure, blood glucose, cholesterol, sweet into a group body index
Oily three esters, HDL, low-density lipoprotein), 3-dimensional basic index (age, sex, Diabetes family history).
1st, the important attribute based on boruta algorithms
By the intervention time of half a year different changes occur for people at highest risk, by boruta algorithms, according to importance ranking
As a result the important attribute of selection influence intervention effect, schematic diagram is as shown in figure 1, specific implementation step is as follows:
Step 1, the data of variable are replicated, builds and resets copy attribute, growth data collection is obtained;
Step 2, based on growth data collection, using bootstrap mode sample drawn collection, the conduct bag not being drawn into is outer
Data, build random forest grader;
Step 3, post-class processing is trained, and calculates the mean square error MSE of the outer data of each tree correspondence bag1,MSE2,…,
MSEi, wherein (0<I≤n), then the outer data mean square error vector of the original bag of m tree can be expressed as [MSE1,MSE2,…,
MSEn];
Step 4, data mean square error vector [MSE outside the original bag obtained based on step 31,MSE2,…,MSEn], calculate
The Z values of correspondence attribute, the maximum copy attribute of screening Z values, and the attribute bigger than the attribute value, fusion obtain important category
Property, the attribute smaller than the attribute value is classified as insignificant attribute, insignificant attribute and copy attribute is deleted;
Step 5, step 1 is repeated to step 4, untill reaching preset termination condition;
Step 6, according to the result of calculation of boruta algorithms, the mean square error of attribute represents the importance of attribute, attribute weight
The forward attribute of the property wanted sequence, to have the attribute of material impact to diabetes B intervention effect.
When intervening half a year, 20 dimension attributes carry out Importance of Attributes sequence according to data mean square error outside bag, as a result see Fig. 2.
Compare the Z values of attribute by iteration, it is final to determine that 5 tie up important attributes, it is that effective exercise amount, effective dose compare body respectively
Weight, waistline, effective exercise time and BMI, the tentative uncertainty attribute of 1 dimension, is low-density lipoprotein, and remaining 14 dimension attribute is not weigh
Want attribute.
2nd, the notable attribute that logic-based is returned
Different intervention time crowd situations of change are directed to respectively, and by logistic regression algorithm, weight represents each attribute to dry
The influence degree of pre- effect, sig table shows the result significantly examined, and as sig≤0.05, represents that the attribute has aobvious to intervention effect
Work property influence, specific implementation step is as follows:
Step 1, attribute to be selected is normalized, eliminates influence of the different dimensions of attribute to result;
Step 2, construction logic regression model, each attribute coefficients are calculated using maximum Likelihood;
Step 3, successive Regression is carried out using AIC criterion, obtains combinations of attributes during AIC minimums, construction logic returns mould
Type;
Step 4, significance test is carried out to attribute weight in Logic Regression Models, using significance as 0.05, screening
There is the attribute significantly affected on intervention effect.
Experimental result, is shown in Table 1:
The logistic regression Importance of Attributes ranking results of table 1
According to experimental result, it is defined by significance for 0.05, selection effective dose is than body weight, effective exercise time, BMI
For notable attribute.
3rd, attribute is merged
The attribute to intervention effect with material impact is obtained based on boruta algorithms, and logistic regression algorithm is obtained to dry
Pre- effect has the attribute significantly affected, selects the attribute occurred simultaneously, as the determinant attribute of influence intervention effect, final true
Effective dose is determined than the determinant attribute of body weight, effective exercise time, BMI for influence intervention effect.
Claims (4)
1. a kind of biomedical determinant attribute system of selection, it is characterised in that the described method comprises the following steps:
Step 1, data set S includes N number of sample, and the attribute to be selected of M dimension influence diabetes B intervention effects is calculated using boruta
Method is fitted to the intervention effect of crowd, obtains influenceing the determinant attribute of intervention effect, and the importance of attribute to be selected to arrange
Sequence;
Step 2, based on raw data set S, discrimination model is returned according to the intervention effect construction logic of crowd, logistic regression is used
Algorithm is fitted to the intervention effect of crowd, obtains having the attribute significantly affected to intervention effect;
Step 3, logic in the attribute that there is intervention effect material impact, step 2 is obtained based on boruta algorithms in step 1 to return
Reduction method obtains having the attribute significantly affected to intervention effect, with reference to expert opinion, and the method sorted out using occuring simultaneously obtains shadow
Ring the determinant attribute of intervention effect.
2. according to the method described in claim 1, it is characterised in that the step of the use boruta algorithms selection important attributes
Specifically include:
Step 2.1, data set S copy attribute is created, rearrangement of going forward side by side obtains recombination data collection S, increase data-oriented collection
Randomness;
Step 2.2, the number for building tree is n (0<N N) in data set S, taken out at random with putting back to using bootstrap modes
N number of new sample set is taken, the data volume of each sample set is about 2N/3, and each sample set is registered as Di (0<I n), do not have
The data record being extracted is the outer data Oi (0 of bag<In), the Random Forest model for including n tree is built;
Step 2.3, post-class processing is built, calculating each sets the outer data mean square error of corresponding bag, is designated as MSEi, then n tree
The outer data mean square error vector of original bag be [MSE1, MSE2 ..., MSEn];
Step 2.4, the mean square error vector [MSE1, MSE2 ..., MSEn] obtained based on step 2.2, is calculated with corresponding attribute
The Z values arrived, determine the maximum copy attribute of Z values, while filtering out the attribute bigger than its value as important attribute, and correspond to it
The small attribute of Z values is labeled as insignificant attribute, and is deleted from data set;
Step 2.5, above step 1.1 to 1.4 is repeated, until reaching default end condition;
Step 2.6, the mean square error of attribute represents the importance of attribute, the forward attribute of Importance of Attributes sequence, for 2 types sugar
The sick intervention effect of urine has the attribute of material impact.
3. according to the method described in claim 1, it is characterised in that have the step of attribute notable using logistic regression algorithms selection
Body includes:
Step 3.1, tie up attribute to be selected to M to be normalized, computational methods are:
<mrow>
<msubsup>
<mi>sx</mi>
<mi>i</mi>
<mi>l</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>x</mi>
<mi>i</mi>
<mi>l</mi>
</msubsup>
<mo>-</mo>
<mi>min</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>l</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>max</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>l</mi>
</msup>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>min</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>l</mi>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
Wherein,It is i-th of sample l dimension attribute original value,It is the normalized value of i-th of sample l dimension attribute,WithIt is then the maximum and minimum value of sample l dimension attributes respectively;
Step 3.2, construction logic regression model, each attribute coefficients are calculated using maximum Likelihood, and regression equation is as follows:
F (x)=b0+b1x1+b2x2+...+bMxM
Wherein, xl(0<L≤M) l dimension attributes are represented, M is the dimension of attribute, bl(0<L≤M) represent each in Logic Regression Models
The weight of attribute;
Step 3.3, successive Regression is carried out using AIC criterion, obtains combinations of attributes during AIC minimums, construction logic regression model;
Step 3.4, significance test is carried out to M dimension attributes weight in Logic Regression Models, using significance as 0.05, screening
There is the attribute significantly affected on intervention effect.
4. according to the method described in claim 1, it is characterised in that the step of attribute fusion obtains determinant attribute specifically includes:
Step 4.1, with reference to expertise opinion, it is considered to the collection difficulty of attribute, to the category for selecting to obtain in step 1 and step 2
Property is further screened;
Step 4.2, the attribute to intervention effect with material impact is obtained based on boruta algorithms, and logistic regression algorithm is obtained
There is the attribute significantly affected on intervention effect, the attribute that selection occurs simultaneously is used as the determinant attribute of influence intervention effect.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710332543.7A CN107169284A (en) | 2017-05-12 | 2017-05-12 | A kind of biomedical determinant attribute system of selection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710332543.7A CN107169284A (en) | 2017-05-12 | 2017-05-12 | A kind of biomedical determinant attribute system of selection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107169284A true CN107169284A (en) | 2017-09-15 |
Family
ID=59814993
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710332543.7A Pending CN107169284A (en) | 2017-05-12 | 2017-05-12 | A kind of biomedical determinant attribute system of selection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107169284A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280236A (en) * | 2018-02-28 | 2018-07-13 | 福州大学 | A kind of random forest visualization data analysing method based on LargeVis |
CN108921222A (en) * | 2018-07-05 | 2018-11-30 | 四川泰立智汇科技有限公司 | A kind of air-conditioning energy consumption feature selection approach based on big data |
CN109243561A (en) * | 2018-08-10 | 2019-01-18 | 上海交通大学 | The model optimization method and system of therapeutic scheme recommender system |
CN109785973A (en) * | 2019-01-25 | 2019-05-21 | 中国医学科学院阜外医院 | The quantitative forecasting technique of complication occurs in peri-operation period for a kind of patient with operation |
CN109961077A (en) * | 2017-12-22 | 2019-07-02 | 广东欧珀移动通信有限公司 | Gender prediction's method, apparatus, storage medium and electronic equipment |
CN110019166A (en) * | 2017-12-25 | 2019-07-16 | 大连楼兰科技股份有限公司 | Screen the method and customer defection early warning method of attribute data |
CN111178449A (en) * | 2019-12-31 | 2020-05-19 | 浙江大学 | Liver cancer image classification method and device combining computer vision characteristics and imaging omics characteristics |
CN111242174A (en) * | 2019-12-31 | 2020-06-05 | 浙江大学 | Liver cancer image feature extraction and pathological classification method and device based on imaging omics |
CN111243751A (en) * | 2020-01-17 | 2020-06-05 | 河北工业大学 | Heart disease prediction method based on dual feature selection and XGboost algorithm |
CN112101416A (en) * | 2020-08-13 | 2020-12-18 | 北京理工大学 | Method for extracting key attributes of security risk rating of ICT supply chain nodes |
CN112101414A (en) * | 2020-08-13 | 2020-12-18 | 北京理工大学 | ICT supply chain network key node identification attribute extraction method |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101326439A (en) * | 2005-10-11 | 2008-12-17 | 特提斯生物科学公司 | Diabetes-associated markers and methods of use thereof |
CN102930163A (en) * | 2012-11-01 | 2013-02-13 | 北京理工大学 | Method for judging 2 type diabetes mellitus risk state |
CN103150611A (en) * | 2013-03-08 | 2013-06-12 | 北京理工大学 | Hierarchical prediction method of II type diabetes mellitus incidence probability |
CN106203813A (en) * | 2016-07-05 | 2016-12-07 | 北京理工大学 | Life of elderly person self-care ability Quantitative Calculation Method |
CN106407706A (en) * | 2016-09-29 | 2017-02-15 | 北京理工大学 | Boruta algorithm-based multi-level old people physical state quantization level calculation method |
-
2017
- 2017-05-12 CN CN201710332543.7A patent/CN107169284A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101326439A (en) * | 2005-10-11 | 2008-12-17 | 特提斯生物科学公司 | Diabetes-associated markers and methods of use thereof |
CN102930163A (en) * | 2012-11-01 | 2013-02-13 | 北京理工大学 | Method for judging 2 type diabetes mellitus risk state |
CN103150611A (en) * | 2013-03-08 | 2013-06-12 | 北京理工大学 | Hierarchical prediction method of II type diabetes mellitus incidence probability |
CN106203813A (en) * | 2016-07-05 | 2016-12-07 | 北京理工大学 | Life of elderly person self-care ability Quantitative Calculation Method |
CN106407706A (en) * | 2016-09-29 | 2017-02-15 | 北京理工大学 | Boruta algorithm-based multi-level old people physical state quantization level calculation method |
Non-Patent Citations (1)
Title |
---|
何晓群: "《应用多元统计分析》", 31 August 2015 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109961077A (en) * | 2017-12-22 | 2019-07-02 | 广东欧珀移动通信有限公司 | Gender prediction's method, apparatus, storage medium and electronic equipment |
CN110019166A (en) * | 2017-12-25 | 2019-07-16 | 大连楼兰科技股份有限公司 | Screen the method and customer defection early warning method of attribute data |
CN108280236B (en) * | 2018-02-28 | 2022-03-15 | 福州大学 | Method for analyzing random forest visual data based on LargeVis |
CN108280236A (en) * | 2018-02-28 | 2018-07-13 | 福州大学 | A kind of random forest visualization data analysing method based on LargeVis |
CN108921222A (en) * | 2018-07-05 | 2018-11-30 | 四川泰立智汇科技有限公司 | A kind of air-conditioning energy consumption feature selection approach based on big data |
CN109243561B (en) * | 2018-08-10 | 2020-07-28 | 上海交通大学 | Model optimization method and system of treatment scheme recommendation system |
CN109243561A (en) * | 2018-08-10 | 2019-01-18 | 上海交通大学 | The model optimization method and system of therapeutic scheme recommender system |
CN109785973B (en) * | 2019-01-25 | 2021-02-09 | 中国医学科学院阜外医院 | Quantitative prediction method for complications of surgical patients in perioperative period |
CN109785973A (en) * | 2019-01-25 | 2019-05-21 | 中国医学科学院阜外医院 | The quantitative forecasting technique of complication occurs in peri-operation period for a kind of patient with operation |
CN111242174A (en) * | 2019-12-31 | 2020-06-05 | 浙江大学 | Liver cancer image feature extraction and pathological classification method and device based on imaging omics |
CN111178449A (en) * | 2019-12-31 | 2020-05-19 | 浙江大学 | Liver cancer image classification method and device combining computer vision characteristics and imaging omics characteristics |
CN111178449B (en) * | 2019-12-31 | 2021-11-05 | 浙江大学 | Liver cancer image classification method combining computer vision characteristics and imaging omics characteristics |
CN111243751A (en) * | 2020-01-17 | 2020-06-05 | 河北工业大学 | Heart disease prediction method based on dual feature selection and XGboost algorithm |
CN111243751B (en) * | 2020-01-17 | 2022-04-22 | 河北工业大学 | Heart disease prediction method based on dual feature selection and XGboost algorithm |
CN112101416A (en) * | 2020-08-13 | 2020-12-18 | 北京理工大学 | Method for extracting key attributes of security risk rating of ICT supply chain nodes |
CN112101414A (en) * | 2020-08-13 | 2020-12-18 | 北京理工大学 | ICT supply chain network key node identification attribute extraction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107169284A (en) | A kind of biomedical determinant attribute system of selection | |
CN108389201B (en) | Lung nodule benign and malignant classification method based on 3D convolutional neural network and deep learning | |
Kaur et al. | A systematic review on metaheuristic optimization techniques for feature selections in disease diagnosis: open issues and challenges | |
EP3306500A1 (en) | Method for analysing medical treatment data based on deep learning, and intelligent analyser thereof | |
CN109920547A (en) | A kind of diabetes prediction model construction method based on electronic health record data mining | |
CN107822622A (en) | Electrocardiographic diagnosis method and system based on depth convolutional neural networks | |
CN112472048B (en) | Method for realizing neural network for identifying pulse condition of cardiovascular disease patient | |
CN110021397A (en) | Method and storage medium based on human body physiological parameter prediction dosage | |
CN108304887A (en) | Naive Bayesian data processing system and method based on the synthesis of minority class sample | |
CN108511056A (en) | Therapeutic scheme based on patients with cerebral apoplexy similarity analysis recommends method and system | |
CN116364299B (en) | Disease diagnosis and treatment path clustering method and system based on heterogeneous information network | |
CN111951975A (en) | Sepsis early warning method based on deep learning model GPT-2 | |
Wang et al. | Application of fuzzy cluster analysis for medical image data mining | |
CN115221926A (en) | Heart beat signal classification method based on CNN-GRU network model | |
Shoaib et al. | Hybrid classification structures for automatic COVID-19 detection | |
Nagadeepa et al. | Artificial Intelligence based Cervical Cancer Risk Prediction Using M1 Algorithms | |
Dutta | Detecting Lung Cancer Using Machine Learning Techniques. | |
CN113539473A (en) | Method and system for diagnosing brucellosis only by using blood routine test data | |
CN112967803A (en) | Early mortality prediction method and system for emergency patients based on integrated model | |
CN117370565A (en) | Information retrieval method and system | |
Kabir et al. | Association rule mining based on ethnic groups and classification using super learning | |
CN109300545B (en) | Risk early warning method for thalassemia based on RF | |
Shchetinin et al. | Cardiac arrhythmia disorders detection with deep learning models | |
Ahmed et al. | Improving prediction of plant disease using k-efficient clustering and classification algorithms | |
Du et al. | Prediction of pregnancy diabetes based on machine learning |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170915 |
|
WD01 | Invention patent application deemed withdrawn after publication |