filterVarImp: Calculation of filter-based variable importance

In caret: Classification and Regression Training

Description

Specific engines for variable importance on a model by model basis.

Usage

filterVarImp(x, y, nonpara = FALSE, ...) 

Arguments

x	A matrix or data frame of predictor data
y	A vector (numeric or factor) of outcomes)
nonpara	should nonparametric methods be used to assess the relationship between the features and response
...	options to pass to either lm or loess

Details

The importance of each predictor is evaluated individually using a “filter” approach.

For classification, ROC curve analysis is conducted on each predictor. For two class problems, a series of cutoffs is applied to the predictor data to predict the class. The sensitivity and specificity are computed for each cutoff and the ROC curve is computed. The trapezoidal rule is used to compute the area under the ROC curve. This area is used as the measure of variable importance. For multi–class outcomes, the problem is decomposed into all pair-wise problems and the area under the curve is calculated for each class pair (i.e class 1 vs. class 2, class 2 vs. class 3 etc.). For a specific class, the maximum area under the curve across the relevant pair–wise AUC's is used as the variable importance measure.

For regression, the relationship between each predictor and the outcome is evaluated. An argument, nonpara, is used to pick the model fitting technique. When nonpara = FALSE, a linear model is fit and the absolute value of the $t$–value for the slope of the predictor is used. Otherwise, a loess smoother is fit between the outcome and the predictor. The $R^2$ statistic is calculated for this model against the intercept only null model.

Value

A data frame with variable importances. Column names depend on the problem type. For regression, the data frame contains one column: "Overall" for the importance values.

Author(s)

Max Kuhn

Examples

data(mdrr)
filterVarImp(mdrrDescr[, 1:5], mdrrClass)

data(BloodBrain)

filterVarImp(bbbDescr[, 1:5], logBBB, nonpara = FALSE)
apply(bbbDescr[, 1:5],
      2,
      function(x, y) summary(lm(y~x))$coefficients[2,3],
      y = logBBB)

filterVarImp(bbbDescr[, 1:5], logBBB, nonpara = TRUE)

Example output

Loading required package: lattice
Loading required package: ggplot2
       Active  Inactive
MW  0.7892982 0.7892982
AMW 0.5004158 0.5004158
Sv  0.8252991 0.8252991
Se  0.7946819 0.7946819
Sp  0.8276116 0.8276116
           Overall
tpsa     9.6270929
nbasic   0.5097859
negative 1.0435202
vsa_hyd  4.4681448
a_aro    2.6664055
      tpsa     nbasic   negative    vsa_hyd      a_aro 
-9.6270929 -0.5097859 -1.0435202  4.4681448  2.6664055 
             Overall
tpsa     0.358428014
nbasic   0.001259972
negative 0.005258293
vsa_hyd  0.105859830
a_aro    0.033361774

caret documentation built on May 2, 2019, 5:47 p.m.