Skip to main content

Table 1 Classification accuracies. Results from machine learning for six different classification model types when using the first 3 largest PCs from the PCA and a 5-fold cross-validation, between all groups for tissue and non-dried (fresh) serum. SVM = Support Vector Machine, KNN = k-nearest neighbors algorithm

From: Glycosylation spectral signatures for glioma grade discrimination using Raman spectroscopy

Classification accuracy (%)

Linear SVM

Linear discriminant

Cosine KNN

Logistic regression

Bilayed neural network

Narrow neural network

Tissue

      

II vs. III

80%

75%

70%

70%

40%

50%

II vs. IV

75%

70%

60%

70%

45%

55%

III vs. IV

85%

60%

75%

60%

55%

50%

All grades

63.3%

60%

53.3%

x

36.7%

30%

Fresh serum

      

CTRL vs. III

75%

75%

80%

85%

90%

80%

CTRL vs. IV

60%

65%

70%

85%

85%

85%

III vs. IV

60%

65%

65%

75%

90%

85%

All grades

56.7

53.3%

60%

x

63.3%

66.7%