Skip to main content

Table 12 Classification accuracy and number of features selected with five classifiers and six feature selection algorithms

From: Improving feature selection performance using pairwise pre-evaluation

  Forward Backward Relief FSDD Chi-squared MRMR
KNN Original 0.781/01 0.785/41 0.716/10 0.776/05 0.795/05 0.767/15
Modified 0.775/10 0.755/23 0.713/05 0.793/15 0.770/10 0.748/15
SVM Original 0.801/01 0.796/41 0.799/10 0.805/15 0.797/05 0.800/10
Modified 0.812/11 0.820/23 0.798/10 0.823/15 0.802/30 0.809/20
NB Original 0.793/01 0.777/41 0.790/20 0.776/05 0.763/05 0.805/15
Modified 0.819/12 0.754/23 0.813/05 0.813/05 0.788/10 0.786/20
RF Original 0.783/01 0.791/41 0.808/30 0.806/20 0.825/15 0.830/15
Modified 0.843/12 0.820/23 0.823/30 0.851/05 0.801/15 0.819/20
NN Original 0.795/01 NA 0.771/30 0.791/05 0.771/15 0.782/10
Modified 0.744/11 0.734/23 0.759/25 0.764/15 0.765/05 0.771/30
  1. (NB Naive Bayes, RF Random Forest, NN Neural Network)