Table 2.

TOUCAN best-performing models per test set sliding windows and overlaps in A. niger

Sliding		Training		Post-
window	Overlap	set	Classifier	process	P	R	F-m
10 000	50%	50%–50%	mlp	merge3	1	0.871	0.931
10 000	50%	40%–60%	mlp	merge3	1	0.753	0.859
10 000	50%	30%–70%	mlp	merge2	1	0.706	0.828
10 000	50%	20%–80%	mlp	merge2	1	0.706	0.828
10 000	50%	10%–90%	mlp	merge3	1	0.647	0.786
10 000	50%	5%–95%	mlp	merge3	1	0.447	0.618
7000	50%	50%–50%	logit	merge3	0.929	0.765	0.839
7000	50%	40%–60%	logit	merge3	1	0.741	0.851
7000	50%	30%–70%	mlp	merge3	0.969	0.729	0.832
7000	50%	20%–80%	mlp	merge3	1	0.741	0.851
7000	50%	10%–90%	mlp	merge3	1	0.694	0.819
7000	50%	5%–95%	mlp	merge3	1	0.647	0.786
5000	50%	50%–50%	logit	merge3	0.817	0.788	0.802
5000	50%	40%–60%	logit	merge3	0.914	0.753	0.826
5000	50%	30%–70%	logit	merge3	0.953	0.718	0.819
5000	50%	20%–80%	logit	merge3	1	0.718	0.836
5000	50%	10%–90%	mlp	merge3	0.913	0.741	0.818
5000	50%	5%–95%	mlp	merge3	0.923	0.706	0.800
10 000	30%	50%–50%	mlp	merge3	1	0.847	0.917
10 000	30%	40%–60%	mlp	merge3	1	0.741	0.851
10 000	30%	30%–70%	mlp	merge2	1	0.694	0.819
10 000	30%	20%–80%	mlp	merge2	1	0.671	0.803
10 000	30%	10%–90%	mlp	merge3	1	0.6	0.750
10 000	30%	5%–95%	mlp	merge3	1	0.459	0.629
7000	30%	50%–50%	mlp	merge3	0.95	0.906	0.928
7000	30%	40%–60%	mlp	merge3	1	0.824	0.903
7000	30%	30%–70%	mlp	merge2	1	0.741	0.851
7000	30%	20%–80%	mlp	merge3	1	0.741	0.851
7000	30%	10%–90%	lsvc	merge3	1	0.553	0.712
7000	30%	5%–95%	mlp	merge3	1	0.635	0.777
5000	30%	50%–50%	logit	merge3	0.908	0.812	0.857
5000	30%	40%–60%	logit	merge3	0.985	0.788	0.876
5000	30%	30%–70%	logit	merge3	1	0.753	0.859
5000	30%	20%–80%	mlp	merge3	0.985	0.776	0.868
5000	30%	10%–90%	mlp	merge3	0.984	0.729	0.838
5000	30%	5%–95%	mlp	merge3	1	0.706	0.828