Table 5.

Result table with pretrained neural network on PI.

# layers	# hidden units	Sens.	Spec.	F1	Macro Ave.	Mic. Ave.
1	512	0.823	0.904	0.724	0.864	0.89
1	1024	0.83	0.911	0.738	0.87	0.896
1	1536	0.828	0.909	0.735	0.869	0.895
1	2048	0.818	0.913	0.735	0.865	0.896
2	512	0.836	0.909	0.739	0.872	0.896
2	1024	0.822	0.91	0.733	0.866	0.894
2	1536	0.824	0.912	0.736	0.868	0.896
2	2048	0.816	0.911	0.731	0.864	0.894
3	512	0.827	0.907	0.731	0.867	0.893
3	1024	0.835	0.904	0.73	0.869	0.892
3	1536	0.833	0.904	0.73	0.869	0.892
3	2048	0.831	0.906	0.732	0.869	0.893
4	512	0.813	0.91	0.727	0.862	0.893
4	1024	0.83	0.905	0.73	0.867	0.892
4	1536	0.829	0.909	0.736	0.869	0.895
4	2048	0.799	0.914	0.726	0.857	0.894
Baseline method		0.971	0.765	0.632	0.868	0.801