| Title: | Kernel_lasso Expansion |
| Version: | 1.0.0 |
| Maintainer: | Zongrui Dai <dzr17723980497@gmail.com> |
| Description: | Provides the function to calculate the kernel-lasso expansion, Z-score, and max-min-scale standardization.It can increase the dimension of existed dataset and remove abundant features by lasso. Z Dai, L Jiayi, T Gong, C Wang (2021) <doi:10.1088/1742-6596/1955/1/012047>. |
| License: | GPL-2 |
| URL: | https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.1.1.9001 |
| Depends: | glmnet (≥ 4.1-2) |
| Imports: | graphics, stats |
| NeedsCompilation: | no |
| Packaged: | 2021-08-20 12:30:59 UTC; 10979 |
| Author: | Zongrui Dai  [aut,
cre] |
| Repository: | CRAN |
| Date/Publication: | 2021-08-21 09:30:08 UTC |
Z_score standardization
Description
Z-score method is used to calculate the standardization value of data.The formula is (x-mean(x))/var(x). It can compress the data into the (0,1).
Usage
Z_score(data, dataframe = FALSE)
Arguments
data |
Your input data, which can be numerci or data.frame |
dataframe |
Wether the data is dataframe. The default is False(numeric) |
Value
Calculate the Z_score standardization of the dataset by the formula: (x-mean(x))/var(x)
Author(s)
Zongrui Dai
Source
https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion
Examples
##For the numeric data
data(iris,package = 'datasets')
w<-Z_score(iris[,1])
print(w)
##For the data.frame data
w1<-Z_score(iris[,-5],dataframe=TRUE)
print(w1)
Gauss function
Description
Gauss function
Usage
gauss(d1, d2, sigma = 0.5)
Arguments
d1 |
vector1 |
d2 |
vector2 |
sigma |
The hyperparameter of RBF kernel function, which indicates the width. |
Value
Calculate the Gauss function
Author(s)
Zongrui Dai
Source
https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion
Examples
##
data(iris,package = 'datasets')
w<-gauss(iris[,1],iris[,2])
print(w)
kernel_lasso_expansion
Description
Kernel_lasso is one feature selection method, which combines the feature expansion and lasso regression together. Kernel function will increase the dimensions of the existed data and then reduce the features by lasso. 'glmnet' package should be higher than 4.1-2.
Arguments
x |
Your input features, which have to be data.frame with at least two variables. |
y |
The dependent variable |
sigma |
The hyperparameter of RBF kernel function, which indicates the width. |
dataframe |
Wether the data is dataframe. The default is TURE |
standard |
Using 'max_min_scale' or 'Z_score' method to standardize the data. NULL means no standardization |
Value
The result is stored in one list which contains the orignial dataset, amplified dataset, final features, and lasso output.
Author(s)
Zongrui Dai
Source
https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion
References
Z. Dai, J. Li, T. Gong, C. Wang (2021), Kernel_lasso feature expansion method: boosting the prediction ability of machine learning in heart attack,” 2021 IEEE. About Kernel-lasso feature expansion method: boosting the prediction ability of machine learning in heart attack” 2021 IEEE.
Examples
##Regression (MSE)
data(attenu,package = 'datasets')
result<-kernel_lasso_expansion(x=attenu[,-c(3,5)],y=attenu[,5],
standard = 'max_min',sigma=0.01,control = lasso.control(nfolds=3,type.measure = 'mse'))
summary(result)
#Plot the lasso
plot(result$lasso)
#Result
result$original ##The original feature space
result$expansion ##The feature space after expansion
result$final_feature ##The name of the final feature
result$final_data ##The dataframe of final feature
lasso.control
Description
The same function from glmnet, which controls the training of lasso.
Usage
lasso.control(nfolds = 10, trace.it = 1, type.measure = "auc")
Arguments
nfolds |
n-fold cross-validation. |
trace.it |
Whether to plot the training process |
type.measure |
Choose the loss funcrion. |
Value
Will return the lasso training setting
Author(s)
Zongrui Dai
Source
https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion
Examples
##10-fold Cross-validation with MSE as loss function
c<-lasso.control(nfolds=10,type.measure='mse')
max_min_scale
Description
max_min_scale is used to calculate the standardization value of data.The formula is (x-min(x))/(max(x)-min(x)). It can compress the data into the (0,1).
Arguments
data |
Your input data, which can be numerci or data.frame |
dataframe |
Wether the data is dataframe. The default is False(numeric) |
Value
Calculate the max-min standardization of the dataset by the formula: (max(x)-x)/(max(x)-min(x))
Author(s)
Zongrui Dai
Source
https://github.com/Zongrui-Dai/Kernel-lasso-feature-expansion
Examples
##For the numeric data
data(iris,package = 'datasets')
w<-max_min_scale(iris[,1])
print(w)
##For the data.frame data
w1<-max_min_scale(iris[,-5],dataframe=TRUE)
print(w1)