Feature selection and classification for microarray data analysis: Evolutionary methods for identifying predictive genes

Table 2 The average accuracy using out-of-sample prediction on the 34 leukemia test samples. The symbol (*) means that there is some perfect predictors found by the algorithm. The highest accuracy is written in bold.

Population size	Feature size	The accuracy of different rank methods on the Test data (out-of-sample) [%]
Population size	Feature size	R1	R2	R3	R4	R5	R6
10	30	97.35*	95.29	93.82	92.94*	93.53	94.12
	50	98.24*	95.59	94.71*	93.82	93.53	95.00
30	30	96.74*	92.65	94.41	95.00*	94.71	93.82
	50	97.06*	95.00	95.00	93.82	95.30	93.82
50	30	97.35*	93.82	94.71*	92.06*	94.12*	93.82*
	50	96.17	93.82	92.65	92.35	94.12	94.71

Abbreviations: R1. Information gain; R2. Twoing rule; R3. Gini index; R4. Sum minority; R5. Max minority; R6. Sum of variances.

ISSN: 1471-2105