Table 2. AUC (10-fold cross-validation).
Class. | Type | Random | t-test | Entropy | Bhatt. | Wilcoxon | SVM RFE | GFS | Lasso | Elastic Net |
NC | S | 0.62(0.17) | 0.66(0.14) | 0.58(0.15) | 0.60(0.15) | 0.62(0.15) | 0.62(0.15) | 0.58(0.15) | 0.63(0.15) | 0.63(0.15) |
E-M | 0.62(0.15) | 0.65(0.14) | 0.59(0.15) | 0.63(0.15) | 0.62(0.15) | 0.63(0.14) | 0.62(0.13) | 0.61(0.16) | 0.63(0.15) | |
E-E | 0.61(0.15) | 0.65(0.14) | 0.59(0.15) | 0.61(0.16) | 0.62(0.15) | 0.61(0.15) | 0.58(0.13) | 0.63(0.13) | 0.63(0.14) | |
E-S | 0.63(0.14) | 0.65(0.14) | 0.58(0.15) | 0.61(0.15) | 0.62(0.15) | 0.63(0.15) | 0.59(0.12) | 0.63(0.13) | 0.63(0.14) | |
KNN | S | 0.59(0.16) | 0.61(0.15) | 0.52(0.11) | 0.57(0.13) | 0.63(0.15) | 0.60(0.15) | 0.59(0.13) | 0.60(0.17) | 0.60(0.17) |
E-M | 0.61(0.14) | 0.62(0.15) | 0.57(0.15) | 0.60(0.15) | 0.64(0.16) | 0.62(0.15) | 0.61(0.12) | 0.61(0.15) | 0.60(0.12) | |
E-E | 0.55(0.13) | 0.63(0.15) | 0.53(0.10) | 0.54(0.10) | 0.63(0.16) | 0.60(0.17) | 0.54(0.16) | 0.61(0.14) | 0.60(0.17) | |
E-S | 0.60(0.13) | 0.63(0.15) | 0.54(0.11) | 0.54(0.12) | 0.62(0.16) | 0.58(0.14) | 0.55(0.14) | 0.62(0.14) | 0.60(0.14) | |
LDA | S | 0.54(0.12) | 0.56(0.12) | 0.51(0.14) | 0.55(0.13) | 0.52(0.12) | 0.56(0.12) | 0.50(0.13) | 0.58(0.14) | 0.57(0.14) |
E-M | 0.53(0.10) | 0.55(0.13) | 0.55(0.13) | 0.58(0.12) | 0.56(0.13) | 0.60(0.15) | 0.52(0.14) | 0.59(0.14) | 0.60(0.13) | |
E-E | 0.54(0.13) | 0.53(0.15) | 0.52(0.15) | 0.53(0.11) | 0.53(0.14) | 0.57(0.13) | 0.53(0.15) | 0.59(0.12) | 0.58(0.13) | |
E-S | 0.54(0.13) | 0.52(0.13) | 0.54(0.13) | 0.55(0.12) | 0.52(0.14) | 0.57(0.16) | 0.54(0.15) | 0.59(0.15) | 0.60(0.13) | |
NB | S | 0.57(0.14) | 0.60(0.13) | 0.58(0.11) | 0.58(0.14) | 0.57(0.13) | 0.56(0.14) | 0.54(0.11) | 0.59(0.15) | 0.59(0.15) |
E-M | 0.59(0.13) | 0.59(0.14) | 0.57(0.14) | 0.59(0.13) | 0.57(0.13) | 0.56(0.13) | 0.59(0.12) | 0.57(0.15) | 0.57(0.14) | |
E-E | 0.55(0.15) | 0.60(0.14) | 0.58(0.12) | 0.57(0.13) | 0.58(0.13) | 0.57(0.14) | 0.58(0.11) | 0.58(0.12) | 0.58(0.13) | |
E-S | 0.58(0.14) | 0.60(0.14) | 0.57(0.13) | 0.57(0.13) | 0.58(0.13) | 0.56(0.14) | 0.58(0.10) | 0.58(0.11) | 0.58(0.13) | |
SVM | S | 0.56(0.18) | 0.56(0.15) | 0.55(0.11) | 0.55(0.12) | 0.54(0.15) | 0.62(0.14) | 0.51(0.16) | 0.62(0.15) | 0.62(0.15) |
E-M | 0.51(0.15) | 0.55(0.14) | 0.59(0.16) | 0.60(0.13) | 0.56(0.13) | 0.62(0.15) | 0.55(0.16) | 0.61(0.16) | 0.61(0.16) | |
E-E | 0.54(0.16) | 0.54(0.15) | 0.54(0.13) | 0.54(0.12) | 0.55(0.15) | 0.61(0.17) | 0.56(0.17) | 0.63(0.13) | 0.62(0.16) | |
E-S | 0.54(0.17) | 0.55(0.18) | 0.56(0.12) | 0.56(0.12) | 0.54(0.14) | 0.61(0.16) | 0.55(0.17) | 0.63(0.14) | 0.62(0.16) |
AUC obtained for each combination of feature selection and classification method, in 10-fold cross validation and averaged over the datasets. Standard error is shown within parentheses. For each selection algorithm, we highlighted the setting in which it obtained the best performance. The Type column refers to the use of feature selection run a single time (S) or through ensemble feature selection, either with the mean (E-M), exponential (E-E) or stability selection (E-S) procedure to aggregate lists.