Skip to main content

Table 4 Comparison of the prediction performance for three machine learning algorithms on new data set.

From: NBA-Palm: prediction of palmitoylation site implemented in Naïve Bayes algorithm

Algorithm Window length 3-fold cross-validation 8-fold cross-validation Jack-Knife validation Average MCC Max MCC difference
   Ac (%) Sn (%) Sp (%) MCC Ac (%) Sn (%) Sp (%) MCC Ac (%) Sn (%) Sp (%) MCC   
Naïve Bayes 1 3 84.64% 43.95% 94.85% 0.4629 85.08% 44.08% 95.36% 0.4767 85.27% 45.31% 95.29% 0.4857 0.4751 0.0228
  4 85.43% 49.52% 94.44% 0.5017 85.62% 51.29% 94.23% 0.512 85.76% 51.43% 94.37% 0.5162 0.51 0.0145
  5 85.49% 51.97% 93.89% 0.5101 86.17% 53.74% 94.30% 0.5339 86.58% 54.69% 94.58% 0.5479 0.5306 0.0378
  6 85.79% 54.01% 93.76% 0.5241 86.72% 57.28% 94.10% 0.5582 86.74% 58.37% 93.86% 0.5618 0.548 0.0377
  7 85.65% 54.42% 93.48% 0.5216 86.52% 56.87% 93.96% 0.5519 86.58% 57.14% 93.96% 0.5541 0.5425 0.0325
  8 85.79% 55.37% 93.42% 0.528 86.20% 57.55% 93.38% 0.545 86.25% 56.73% 93.65% 0.5442 0.5391 0.0170
RBF Network 2 3 84.12% 45.17% 93.89% 0.4516 85.05% 45.44% 94.98% 0.4794 84.62% 46.12% 94.27% 0.4684 0.4665 0.0278
  4 84.83% 46.39% 94.47% 0.4755 85.57% 49.12% 94.71% 0.5046 85.68% 49.80% 94.68% 0.5095 0.4965 0.0340
  5 85.32% 48.98% 94.44% 0.4971 85.35% 48.57% 94.58% 0.4967 86.91% 54.29% 95.09% 0.5565 0.5168 0.0598
  6 85.38% 51.29% 93.93% 0.5051 85.76% 50.61% 94.58% 0.514 85.92% 50.20% 94.88% 0.5178 0.5123 0.0127
  7 85.08% 51.16% 93.59% 0.4965 86.33% 53.47% 94.58% 0.5379 86.99% 53.88% 95.29% 0.558 0.5308 0.0615
  8 85.43% 50.48% 94.20% 0.5043 86.42% 54.01% 94.54% 0.5416 87.15% 55.51% 95.09% 0.5664 0.5374 0.0621
SVM 3 3 84.72% 48.44% 93.82% 0.4785 85.73% 48.57% 95.05% 0.5081 85.35% 47.35% 94.88% 0.4935 0.4934 0.0296
  4 82.84% 49.66% 91.16% 0.4349 84.89% 51.16% 93.35% 0.4914 84.78% 52.24% 92.94% 0.4919 0.4727 0.0570
  5 83.17% 52.65% 90.82% 0.4541 85.32% 55.24% 92.87% 0.5155 86.82% 60.41% 93.45% 0.5694 0.513 0.1153
  6 82.87% 52.79% 90.41% 0.4478 85.30% 57.55% 92.26% 0.5221 84.37% 57.55% 91.10% 0.4999 0.4899 0.0743
  7 81.53% 53.33% 88.60% 0.4213 84.70% 58.37% 91.30% 0.5104 88.46% 64.49% 94.47% 0.6234 0.5184 0.2021
  8 81.01% 53.33% 87.96% 0.4108 84.48% 60.54% 90.48% 0.5132 85.52% 61.22% 91.61% 0.5393 0.4878 0.1285
  1. 1. Default parameters of WEKA software package were used.
  2. 2. Ridge value was set to 10E-8.
  3. 3. Polynomial kernel, exponent was set to 1, complexity value C to 0.1.