Table 3:

Average accuracy (over 10 trial runs) of k-NN clustering algorithm with full linear and quadratic variables using leave-one-out (LOO) and 10-fold cross validation for tree and dual graphs.

	Average Accuracy (%)
	Linear variables		Quadratic variables
	LOO	10-fold	LOO	10-fold
Neighbors	Tree Graphs
1	60.82	60.51	75.63	76.08
3	62.66	62.78	77.28	76.71
5	58.73	59.43	78.61	78.61
7	59.94	60.32	80.25	80.32
9	58.73	59.43	81.08	80.70
11	58.86	58.73	79.68	80.32
13	58.23	59.62	79.24	78.23
15	59.68	59.62	77.66	77.47
17	60.51	60.95	76.52	76.39
19	60.70	59.56	75.70	76.27
Neighbors		Dual Graphs
1	64.24	63.86	78.81	78.81
3	65.55	65.59	81.10	81.06
5	66.74	66.53	80.17	80.17
7	67.84	67.67	79.62	79.24
9	68.64	68.01	79.41	78.81
11	68.86	68.35	78.98	78.94
13	68.35	68.35	78.31	78.22
15	68.60	69.19	77.75	77.50
17	68.52	68.31	77.71	77.25
19	68.69	68.39	76.91	76.57