Skip to main content

Table 8 Median values (and standard deviation in parenthesis) obtained in test set for the five classification methods using fifty most frequently genes pre-selected by LR with iTwiner penalization

From: Identification of biomarkers predictive of metastasis development in early-stage colorectal cancer using network-based regularization

 

Acc

Miscl/FN

Sensitivity

Specificity

AUC

DT

D1

0.78(0.096)

4(1.721)/2(1.226)

0.78(0.136)

0.78(0.146)

0.78(0.096)

D2

0.71(0.085)

5(1.453)/3(1.344)

0.63(0.168)

0.78(0.143)

0.70(0.085)

D3

0.65(0.111)

6(1.882)/3(1.496)

0.63(0.187)

0.78(0.159)

0.65(0.108)

\(\bar{x}\)

0.71

0.68

0.78

0.71

svmL

D1

0.83(0.071)

3(1.284)/3(1.288)

0.67(0.143)

1.00(0.022)

0.83(0.071)

D2

0.76(0.089)

4(1.152)/3(1.218)

0.63(0.152)

1.00(0.085)

0.75(0.092)

D3

0.71(0.092)

5(1.568)/4(1.256)

0.50(0.157)

0.78(0.130)

0.69(0.084)

\(\bar{x}\)

0.77

0.60

0.93

0.76

svmR

D1

0.78(0.089)

4(1.602)/2(1.015)

0.78(0.113)

0.78(0.153)

0.78(0.089)

D2

0.71(0.097)

5(1.656)/2(1.326)

0.88(0.120)

0.78(0.188)

0.72(0.096)

D3

0.59(0.109)

7(1.487)/4(2.259)

0.50(0.282)

0.78(0.154)

0.58(0.107)

\(\bar{x}\)

0.69

0.72

0.78

0.69

LR

D1

0.72(0.094)

5(1.687)/3(1.431)

0.67(0.159)

0.78(0.130)

0.72(0.094)

D2

0.65(0.097)

6(1.652)/4(1.406)

0.50(0.176)

0.89(0.151)

0.64(0.097)

D3

0.65(0.096)

6(1.633)/4(1.456)

0.50(0.182)

0.67(0.173)

0.63(0.092)

\(\bar{x}\)

0.67

0.56

0.78

0.66

RF

D1

0.86(0.063)

3(1.132)/2(1.104)

0.78(0.123)

1.00(0.025)

0.86(0.063)

D2

0.82(0.058)

3(0.983)/3(1.003)

0.63(0.125)

1.00(0.040)

0.81(0.061)

D3

0.76(0.075)

4(1.267)/4(1.135)

0.50(0.142)

1.00(0.102)

0.75(0.076)

\(\bar{x}\)

0.81

0.64

1

0.81

  1. DT—decision trees; svmL—linear support vector machine; svmR—radial support vector machine; LR—logistic regression; RF—random forest; D1—DATASET1; D2—DATASET2; D3—DATASET3; \(\bar{x}\)—datasets mean; Acc—accuracy; Miscl—misclassifications; FN—false negatives; Sensitivity—fraction of actual positive cases (P); Specificity—fraction of actual negative cases (PM); AUC—area under the ROC curve