Greedy feature selection for glycan chromatography data with the generalized Dirichlet distribution

BMC Bioinformatics

Table 9 Prostate cancer data classification performance (Gleason 5 vs. Gleason 7)

	Cross-validation error	Kappa	Sensitivity	AUC
GDFS	0.294	0.412	0.765	0.785
CFS	0.441	0.118	0.471	0.585
rpart	0.412	0.176	0.529	0.512

Evaluation of classification performance from feature selection in the prostate cancer dataset (Gleason 5 vs. Gleason 7 cases) for the proposed GDFS method (GDFS), correlation-based feature selection (CFS), and recursive partitioning (rpart). Reported are the cross-validation error (misclassification rate), Cohen’s kappa statistic, sensitivity, and AUC (corresponding to ROC curves in Figure 10) for the statistical classifications from each method.

ISSN: 1471-2105