Table 2. Mean accuracy rate and area under the receiver operating curve (AUC) for the eight classifiers on VBM grey matter.
Algorithm | Mean accuracy | 5%limit | 95%limit | AUC | 5%limit | 95%limit | |
---|---|---|---|---|---|---|---|
Healthy vs. Schizophrenia | aridge | 0.756 | 0.701 | 0.806 | 0.836 | 0.788 | 0.885 |
blasso | 0.741 | 0.683 | 0.792 | 0.82 | 0.769 | 0.872 | |
celastic | 0.76 | 0.703 | 0.812 | 0.815 | 0.762 | 0.869 | |
dL0 norm | 0.752 | 0.712 | 0.795 | 0.835 | 0.785 | 0.884 | |
eSVC | 0.772 | 0.731 | 0.812 | 0.85 | 0.804 | 0.896 | |
fRDA | 0.745 | 0.683 | 0.804 | --- | --- | --- | |
gGPC | 0.756 | 0.699 | 0.805 | 0.828 | 0.778 | 0.878 | |
hRF | 0.752 | 0.69 | 0.805 | 0.837 | 0.788 | 0.885 | |
Healthy vs. Bipolar dis. | ridge | 0.623 | 0.586 | 0.664 | 0.686 | 0.621 | 0.75 |
lasso | 0.655 | 0.616 | 0.7 | 0.702 | 0.639 | 0.766 | |
elastic | 0.635 | 0.592 | 0.681 | 0.691 | 0.627 | 0.756 | |
L0 norm | 0.651 | 0.613 | 0.69 | 0.706 | 0.643 | 0.769 | |
SVC | 0.647 | 0.599 | 0.694 | 0.698 | 0.634 | 0.762 | |
RDA | 0.616 | 0.557 | 0.668 | --- | --- | --- | |
GPC | 0.608 | 0.565 | 0.658 | 0.671 | 0.605 | 0.737 | |
RF | 0.62 | 0.571 | 0.67 | 0.688 | 0.624 | 0.753 | |
Bipolar dis. vs. Schizophrenia | ridge | 0.66 | 0.605 | 0.716 | 0.692 | 0.627 | 0.756 |
lasso | 0.609 | 0.555 | 0.659 | 0.646 | 0.579 | 0.713 | |
elastic | 0.616 | 0.562 | 0.676 | 0.689 | 0.624 | 0.753 | |
L0 norm | 0.581 | 0.507 | 0.643 | 0.659 | 0.593 | 0.726 | |
SVC | 0.652 | 0.593 | 0.712 | 0.696 | 0.632 | 0.761 | |
RDA | 0.605 | 0.545 | 0.657 | --- | --- | --- | |
GPC | 0.621 | 0.583 | 0.661 | 0.696 | 0.632 | 0.76 | |
RF | 0.613 | 0.581 | 0.646 | 0.685 | 0.621 | 0.75 |
Lower and upper limits for the 95% confidence intervals generated by bootstrap are also reported for these two quantities.
aridge: Ridge regression,
blasso: Lasso regression,
celastic: Elastic net regularization,
dL0-norm: L0-norm regularization,
eSVC: Support vector classifier,
fRDA: Regularized discriminant analysis,
gGPC: Gaussian process classifier,
hRF: Random forest.