Efficacious End User Measures—Part 1: Relative Class Size and End User Problem Domains

Table 11

These tables show the results for the compression-based classifier and the Bayesian classifier. The measures output on different scales and measure different characteristics; they cannot be directly compared. Because these are two different classifiers, the output ranges differ.

(a)

Compression classifier
rCS-sensitive measures			rCS-invariant measures
Measure	Value	Boundary	Measure	Value	Boundary
IC	0.528	0.800	Youden	0.791	0.200
TAR	0.930	0.800	DOR	6.78	0.200
MCC	0.786	0.600	ROC-AUC	0.851	NA
𝑇+𝐹-score	0.829	0.800

(b)

Bayesian classifier
rCS-sensitive measures			rCS-invariant measures
Measure	Value	Boundary	Measure	Value	Boundary
IC	0.068	188	Youden	0.057	−228
TAR	0.620	638	DOR	0.206	638
MCC	0.053	−228	ROC-AUC	0.505	NA
𝑇+𝐹-score	0.543	−387