Skip to main content

Table 2 CM presents the confusion matrix of a run. The first row captures the number of true positives and false positives. The second row presents the number of false negatives and true negatives. The test results in % right of CM and algorithmic parameters used for the classification of the data determined with Weka

From: Application of an interpretable classification model on Early Folding Residues during protein folding

CM

 

CA

PR

RE

F1

auROC

Naive Bayes

187

195

72.8

23.9

38.8

29.6

70.9

195

2190

     

Random Forest

192

290

82.1

39.6

39.8

39.7

64.7

293

2491

     

Support Vector Machine

134

348

87.0

63.2

27.8

38.6

62.5

78

2706

     

GMLVQ with 1 prototype per class

Run 1

320

162

69.6

27.8

66.4

39.2

67.7

830

1954

     

Run 2

351

162

68.7

28.3

72.8

40.7

73.7

890

1954

     

Run 3

348

134

68.6

28.1

72.2

40.4

76.6

891

1893

     

GMLVQ with 5 prototype per class

Run 4

187

295

77.4

29.7

38.8

33.6

69.4

443

2341

     

Run 5

288

194

69.0

26.0

59.8

36.2

70.5

819

1965

     

Run 6

274

208

70.3

26.4

56.8

36.1

70.3

763

2021

     
  1. Additionally, we marked the best values for the single evaluation measured bold. If not stated otherwise, default setup was used. SVM with RBF-kernel (σ=5) which results in 1193 number of support vectors. Weights for weighted accuracy: 0.75 and 0.25. F β-measure with β=1 (F1)