Skip to main content
. 2010 Aug 17;26(20):2586–2593. doi: 10.1093/bioinformatics/btq472

Fig. 6.

Fig. 6.

Pairwise inter-study prediction accuracies of MBP and GBP. MBP used all genes in training data in model construction while GBP only used intersected genes across training and testing data. The evaluation was performed in three sets of studies (I: Beer, Bhat and Garber in the top three rows; II: Dhan and Luo in the fourth row; III: Welsh and Yu in the fifth row). LDA, KNN and SVM were evaluated and the top M = 10, 20, 30 features were used in the univariate feature selection (K = 100 and δ = 20).