Skip to main content

Table 3 Average AUCs (+ standard deviation) across 100 repetitions of cross validation from the models corresponding to different combinations of three classifiers (column 1) and four feature ranking methods (column 2) for five different classification tasks (columns 3–6: Experiment 1A, 1B, 1C, 2A, and 2B)

From: Quantitative nuclear histomorphometric features are predictive of Oncotype DX risk categories in ductal carcinoma in situ: preliminary findings

Classifiers

Feature ranking methods

High vs. low

High vs. intermediate

Intermediate vs. low

High + intermediate vs. low

High vs. low + intermediate

SVM

Covariance

0.66 ± 0.11

0.64 ± 0.19

0.56 ± 0.15

0.62 ± 0.08

0.65 ± 0.12

WRST

0.64 ± 0.11

0.61 ± 0.16

0.55 ± 0.14

0.61 ± 0.08

0.66 ± 0.12

MRMR-mid

0.68 ± 0.12

0.67 ± 0.18

0.53 ± 0.17

0.61 ± 0.10

0.63 ± 0.13

MRMR-miq

0.67 ± 0.13

0.65 ± 0.17

0.54 ± 0.17

0.62 ± 0.10

0.64 ± 0.12

LDA

Covariance

0.65 ± 0.11

0.62 ± 0.19

0.57 ± 0.15

0.63 ± 0.09

0.65 ± 0.09

WRST

0.63 ± 0.12

0.61 ± 0.18

0.53 ± 0.16

0.62 ± 0.08

0.64 ± 0.12

MRMR-mid

0.65 ± 0.13

0.66 ± 0.19

0.54 ± 0.15

0.61 ± 0.09

0.64 ± 0.12

MRMR-miq

0.64 ± 0.13

0.64 ± 0.21

0.55 ± 0.15

0.61 ± 0.09

0.66 ± 0.12

QDA

Covariance

0.62 ± 0.13

0.60 ± 0.18

0.54 ± 0.15

0.61 ± 0.09

0.64 ± 0.12

WRST

0.59 ± 0.13

0.59 ± 0.15

0.52 ± 0.16

0.62 ± 0.09

0.63 ± 0.11

MRMR-mid

0.61 ± 0.14

0.61 ± 0.17

0.51 ± 0.15

0.58 ± 0.11

0.62 ± 0.14

MRMR-miq

0.63 ± 0.12

0.62 ± 0.18

0.53 ± 0.16

0.60 ± 0.10

0.63 ± 0.12