Skip to main content

Table 2 Performance metrics of different ML algorithms using negative samples generated through random (top) and K-means (bottom) sampling

From: Improved cytokine–receptor interaction prediction by exploiting the negative sample space

  SE SP ACC MCC AUC g-means
Random (loocv)
 NB 52.3 59.3 55.8 0.119 0.554 54.9
 A1DE 70.5 53.3 62.9 0.264 0.707 62.5
 SMO-RBF 79.1 74.5 76.7 0.536 0.767 76.7
 SMO-PolyK 62.8 50.9 56.9 0.199 0.569 56.4
 SMO-PuK 73.0 83.5 78.2 0.568 0.782 78.0
 IBK 81.4 71.9 76.6 0.536 0.766 76.4
 Bagging 66.6 63.5 65.1 0.303 0.718 65.0
 RF 73.2 71.6 72.4 0.449 0.807 72.3
K-means (loocv)
 NB 73.3 51.3 61.3 0.253 0.713 63.0
 A1DE 83.6 84.1 83.9 0.679 0.908 83.8
 SMO-RBF 90.2 90.8 90.5 0.814 0.905 90.5
 SMO-PolyK 80.5 87.9 84.2 0.686 0.842 84.0
 SMO-PuK 85.7 91.7 88.4 0.770 0.887 88.3
 IBK 86.1 86.1 86.1 0.722 0.861 86.0
 Bagging 86.4 92.3 89.4 0.789 0.945 89.3
 RF 89.7 96.3 93.0 0.862 0.965 92.9
  1. The averages of ten runs of loocv are reported. The best performance for each metric is shown in italic