. Author manuscript; available in PMC: 2014 Jul 3.

Published in final edited form as: J Biomol Screen. 2009 Dec;14(10):1236–1244. doi: 10.1177/1087057109351312

Table I.

Classifier Validation Using PCA-based features

No. of Hidden Nodes used by the ANNs (First layer, Second layer)	No. of Training Samples Used (Active, Non-active)	Total No. of Compounds Classified as Active Compound (N_T)	Actual No. of Active Compounds Found in the Classified Set (N_A)	Actual No. of Non-active Compounds Found in the Classified Set (N_N)	Ratio of the Actual No. of Active and Non-active Compounds Identified (R_A/N)=N_A/N_N
(15,10)	(70, 300)	3672	196	3476	0.0564
(15,20)	(70, 370)	4242	212	4030	0.0526
(15,20)	(60, 300)	1706	129	1577	0.0818
(24,10)	(60, 390)	1407	117	1205	0.0971
(24,10)	(60,393)	1380	125	1255	0.0996
(24,10)	(60,395)	1391	85	1306	0.0651
(25,0)	(50, 100)	4194	221	3973	0.0556
(25,0)	(50, 200)	623	53	570	0.0930
(25,0)	(50, 300)	1614	107	1507	0.0710
(25,10)	(100, 600)	1503	117	1385	0.0845
(30,10)	(50, 200)	2277	132	2145	0.0615