Table 5.

The characteristics of detection performance achieved by various ensembles of distance metric–based classifiers using the evaluation data set and the independent data set. Ten combinations with the highest performance as determined by the criterion C₂ are displayed.^a

Combination of distance metrics	Sensitivity (%)	Specificity (%)	Accuracy (%)	Balanced accuracy (%)	Error (%)	Precision (%)	C2
MAHb , MANc , CANd	85.71	72.73	75.86	79.22	24.14	50.00	1.649
CAN, EUCe, MAH, MAN, MINKf	85.71	68.18	72.41	76.95	27.59	46.15	1.627
EUC, MAH, CAN	85.71	68.18	72.41	76.95	27.59	46.15	1.627
MAH, MINK, CAN	85.71	68.18	72.41	76.95	27.59	46.15	1.627
EUC, MAH, MAN, CAN	85.71	68.18	72.41	76.95	27.59	46.15	1.627
EUC, MAH, MINK, CAN	85.71	68.18	72.41	76.95	27.59	46.15	1.627
MAH, MAN, MINK, CAN	85.71	68.18	72.41	76.95	27.59	46.15	1.627
MAH, MAN	71.43	86.36	82.76	78.90	17.24	62.50	1.503
EUC, MAH	71.43	81.82	79.31	76.62	20.69	55.56	1.481
MAH, MINK	71.43	81.82	79.31	76.62	20.69	55.56	1.481

^aThe best performing combination of the distance metrics is highlighted in bold.

^bMAH: Mahalanobis.

^cMAN: Manhattan.

^dCAN: Canberra.

^eEUC: Euclidean.

^fMINK: Minkowski.