Fig. 11
From: SVM-RFE: selection and visualization of the most relevant features through non-linear kernels
Visualization of RFE-pseudo-samples results for Scenario 2. Results for Scenario 2 (in which variables 29 and 30 were the relevant variables) over all 100 simulated datasets, all 30 variables, and first iteration of the RFE-pseudo-samples algorithm. The pseudo-samples distribution for each variable is shown with a non-parametric local regression estimation (LOESS) with the corresponding 95% confidence interval