Skip to main content

Table 4 Choice of learning algorithms. Learning algorithms tested: LR – logistic regression; NN – neural network of 5 hidden nodes; NNE – bagging ensemble of 10 NNs; SVM/linear – linear support vector machine (inner-product kernel), with C = 0.5 for VSL2-S and C = 1 for VSL2-L; SVM/RBF – nonlinear support vector machine (radius-based-function kernel), with C = 2, gamma = 2-4 for VSL2-S and C = 1, gamma = 2-2 for VSL2-L. All 54 features were included to build the predictor models. The SVM parameters were selected by embedded 5-fold cross-validation (see Predictor model).

From: Length-dependent prediction of protein intrinsic disorder

  Learning algorithm SN S SN L SP ACC S /ACC L
  LR 81.8 ± 1.1 70.2 ± 1.9 81.9 ± 0.3 81.8 ± 0.6
VSL2-S NN 81.1 ± 1.1 68.6 ± 1.9 82.0 ± 0.3 81.5 ± 0.6
  NNE 81.5 ± 1.1 68.8 ± 2.0 83.3 ± 0.3 82.4 ± 0.6
  SVM/linear 82.0 ± 1.1 70.7 ± 1.9 81.5 ± 0.3 81.7 ± 0.6
  SVM/RBF 81.0 ± 1.1 70.0 ± 1.9 81.0 ± 0.3 81.0 ± 0.6
  LR 42.1 ± 1.8 80.3 ± 2.2 87.8 ± 0.5 84.0 ± 1.2
VSL2-L NN 31.3 ± 1.7 76.6 ± 2.4 90.3 ± 0.5 83.4 ± 1.2
  NNE 31.9 ± 1.7 76.1 ± 2.4 91.9 ± 0.5 84.0 ± 1.2
  SVM/linear 44.1 ± 1.8 82.1 ± 2.2 87.3 ± 0.6 84.7 ± 1.1
  SVM/RBF 38.2 ± 1.7 80.2 ± 2.2 87.7 ± 0.6 83.9 ± 1.1