Table 2.

Statistical information for the five most statistically significant kNN QSAR models based on three modeling sets.

Model	N-training	Pred-training	N-test	Pred-test	NNN
The best kNN classification model for 137 class 1 versus 93 class 2 compounds
1	173	0.84	55	0.73	1
2	147	0.86	74	0.70	1
3	193	0.83	37	0.73	1
4	165	0.86	59	0.70	1
5	173	0.81	55	0.75	1
The best kNN continuous model for 137 class 1 compounds
1	103	0.66	34	0.81	3
2	103	0.73	34	0.71	2
3	111	0.71	26	0.74	3
4	115	0.65	22	0.79	5
5	77	0.73	60	0.71	2
The best kNN continuous model for 93 class 2 compounds
1	80	0.61	13	0.84	2
2	77	0.67	16	0.77	1
3	80	0.69	13	0.74	1
4	80	0.65	13	0.76	2
5	79	0.63	14	0.78	2

Abbreviations: NNN; number of the nearest neighbors used for prediction; N-test, number of compounds in the test set; N-training, number of compounds in the training set; Pred-test, the overall predictivity of the test set (correct classification rate for classification models, R² for continuous models); Pred-training, the overall predictivity of the training set (correct classification rate for classification models, q² for continuous models).