Table 1.

Definition of performance metrics used in this paper.

Meaure	Formula
Sensitivity (Se)	$S_e=\frac{N_{TP}}{N_{FN}+N_{TP}}$
Specificity (Sp)	$S_p=\frac{N_{TN}}{N_{FP}+N_{TN}}$
False positive rate (FPr)	$F{P}_r=\frac{N_{FP}}{N_{FP}+N_{TN}}=1-S_p$
False Discovery Rate (FDr)	$F{D}_r=\frac{N_{FP}}{N_{FP}+N_{TP}}$
False positive proportion (FPp)	$F{P}_p=\frac{N_{FP}}{N_{FN}+N_{TP}}$
Matthew's correlation coefficient (MCC)	$\text{MCC=}\frac{{\text{N}}_{TP}\ast {\text{N}}_{TN}-{\text{N}}_{FP}\ast {\text{N}}_{FN}}{\sqrt{\begin{array}{l}\left({\text{N}}_{TP}+{\text{N}}_{FP}\right)\left({\text{N}}_{TP}+{\text{N}}_{FN}\right)\hfill \\ \left({\text{N}}_{TP}+{\text{N}}_{FP}\right)\left({\text{N}}_{TP}+{\text{N}}_{FN}\right)\hfill \end{array}}}$
F1 score	$F_1=\frac{2{\text{N}}_{TP}}{2{\text{N}}_{TP}+N_{FP}+N_{FN}}$
Cohen κ	$\begin{array}{l}\kappa =\frac{\text{X-}\left({\text{N}}_{TN}+{\text{N}}_{TP}\right)N_e}{X-N_e{}^2}\text{with}\\ X=\left({\text{N}}_{TP}+{\text{N}}_{FP}\right)\left({\text{N}}_{TP}+{\text{N}}_{FN}\right)\\ +\left({\text{N}}_{FP}+{\text{N}}_{TN}\right)\left({\text{N}}_{FN}+{\text{N}}_{TN}\right)\end{array}$