Tumour Relapse Prediction Using Multiparametric MR Data Recorded during Follow-Up of GBM Patients

Table 7

Weighted BER for supervised and semisupervised classifiers trained on complete and imputed data. We marked the best 6 classifiers by bold font.

Weighted BER	Complete features	Imputed features	Average
dLDA			0.194
SVM-lin			0.259
SVM-poly			0.310
SVM-rbf			0.507
SVM-mlp			0.244
Bayesian LSSVM			0.420
LSSVM-lin			0.366
LSSVM-poly	0.462	0.362	0.412
LSSVM-rbf			0.364
Random forests			0.221
AdaBoost			0.415
LogitBoost			0.242
GentleBoost			0.302
RobustBoost			0.237
LPBoost			0.381
TotalBoost			0.397
RUSBoost			0.295
Classification tree			0.307
3-NN (correlation)			0.392
Pattern net	0.449	0.288	0.366
Feed forward net	0.399	0.411	0.405
Cascade forward net	0.586	0.485	0.535
Fit net	0.535	0.350	0.443
LDS	0.442	0.534	0.488
SMIR	0.278	0.436	0.357
S4VM	0.456	0.473	0.465