Skip to main content

Table 1 Mean misclassification rate of feature selection methods applied to simulated test data

From: Elastic SCAD as a novel penalization method for SVM classification tasks in high-dimensional data

FS method

r = 10

r = 50

r = 100

r = 200

L2 SVM

34.8(2.2)

33.1(2.0)

33.3 (2.1)

32.8 (1.9)

L1 SVM

28.3(2.8)

28.6 (3.0)

32.4 (2.2)

32.9 (2.1)

SCAD SVM

18.0 (2.2)

27.2 (4.4)

35.3 (3.4)

34.7 (4.1)

Elastic Net SVM

19.4 (2.0)

24.7 (3.0)

31.3 (2.3)

33.1 (2.7)

Elastic SCAD SVM

20.8 (4.5)

26.8 (4.2)

33.1 (2.7)

34.2 (4.1)

  1. Training and test data with 1000 features and 500 samples were simulated. The number of relative features (r) were increased from r = 10 to r = 200 in four steps. Each simulation step was based on 100 simulations of training and test data. In bold - the significant best method(s) according to the MCB test at the family-wise significance level α = 0.05 and non-inferiority margin of Δ = 5%.