Skip to main content

Table 5 Results of 10-fold cross validations of SVM and random forest on 6 datasets with different P:N ratios of positive to negative instances

From: Predicting protein-binding regions in RNA using nucleotide profiles and compositions

P:N

Sensitivity

Specificity

Accuracy

PPV

NPV

MCC

SVM

      

1:1

91.61%

92.39%

92.02%

91.69%

92.31%

0.840

1:2

91.37%

92.17%

91.91%

84.53%

95.80%

0.819

1:4

91.13%

92.33%

92.09%

74.64%

97.68%

0.777

1:6

91.22%

91.95%

91.84%

66.71%

98.34%

0.736

1:8

91.22%

91.92%

91.83%

62.52%

98.61%

0.713

1:10

91.19%

91.54%

91.50%

58.11%

98.78%

0.686

Random forest

      

1:1

91.13%

92.06%

91.62%

91.32%

91.89%

0.832

1:2

85.44%

95.21%

92.09%

89.31%

93.32%

0.816

1:4

80.40%

97.18%

93.85%

87.59%

95.24%

0.802

1:6

77.88%

97.77%

94.78%

86.01%

96.16%

0.788

1:8

76.01%

98.01%

95.18%

84.95%

96.51%

0.777

1:10

75.24%

98.14%

95.53%

83.90%

96.86%

0.770

  1. PPV positive prediction value, NPV negative prediction value, MCC Matthews correlation coefficient