Table 4.

Best Prediction Results of FDA, PLS, and the Four Kernel-Based Classifiers

		Dataset-1 results (%)			Dataset-2 results (%)
		Length of fragments			Length of fragments
Classifiers	Results (%)	192 bp	42 bp	30 bp	192 bp	42 bp	30 bp
FDA		96.28	82.99	82.82	96.36	80.18	82.83
		96.20	83.83	77.94	96.74	81.50	68.17
		96.24	83.41	80.38	96.55	80.84	75.50
PLS		98.40	86.59	82.91	98.85	88.44	84.59
		95.27	81.18	79.48	96.72	82.46	76.38
		96.83	83.89	81.19	97.78	85.45	80.48
KPLS		98.36	87.09	84.99	99.50	89.78	86.64
		97.98	86.08	81.67	98.52	87.30	84.59
		98.16	86.59	83.33	99.01	88.54	85.61
SVM		96.86	85.84	83.25	98.65	89.26	87.52
		97.16	85.78	81.94	98.63	86.27	80.48
		97.51	85.81	82.59	98.64	87.77	84.00
GP		97.88	86.58	83.74	98.89	88.75	80.48
		97.67	82.80	79.91	97.13	83.18	73.45
		97.78	84.69	81.82	98.01	85.96	76.97
KFDA		96.50	84.74	87.88	96.43	90.19	77.55
		96.32	86.62	76.64	99.22	83.80	79.60
		96.43	85.68	82.26	97.82	86.99	78.58

For comparison, the results obtained by Gao and Zhang (2004) with FDA and the best results of the four linear classifiers that were achieved by the PLS method are shown. The results of FDA calculated by Gao and Zhang (2004) are shown in italics. The average accuracies, which were the best ones among the algorithms evaluated here, are shown in boldface. FDA, Fisher discriminant analysis; PLS, partial least squares; KFDA, kernel Fisher discriminant analysis; KPLS, kernel partial least squares; SVM, support vector machine; GP, Gaussian process.