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

Table 6

wBER comparison between our in-house method of imputing missing values and built-in imputation strategy of different supervised classifiers.

Weighted BER	Our method	Built-in method
Random forests	0.294	0.423
AdaBoost	0.324	0.333
LogitBoost	0.335	0.241
GentleBoost	0.308	0.245
RobustBoost	0.325	0.296
LPBoost	0.256	0.369
TotalBoost	0.289	0.323
RUSBoost	0.308	0.361
Decision tree	0.346	0.651