Determining Cutoff Point of Ensemble Trees Based on Sample Size in Predicting Clinical Dose with DNA Microarray Data

Table 2

Average estimates of learning techniques with 100 repetitions.

Dataset	SVR (lin)	SVR (pol)	SVR bagging	RT	RT bagging	GBM
GSE1938		0.874	0.941↓	0.838	0.724↓	0.943↑
GSE12817		0.925	0.963↓	0.792	0.847↑	0.945↑
GSE3078		0.583	0.538↓	0.759	0.739↓	↑
GSE10748	0.367	0.289		0.254
GSE15457		0.407	0.769↓	0.720		↑
GSE8982		0.686	0.574↓	0.593	0.616↑	0.720↑
GSE14954		0.653	0.710↓	0.426	0.584↑	0.781↑
GSE9751		0.363	0.457↓	0.135
GSE7955		0.466	0.443↓	0.198	0.300↑	0.325↑
GSE2409		0.263	0.714↓	0.440	0.538↑	0.644↑
GSE9539		0.700	0.823↓	0.552	0.664↑	0.833↑

Bold indicates the best performance; : increase in ; : decrease in compared to single performance. and : significant at 0.001 and 0.05 level, respectively. SVR (lin), SVR (pol), and RT are single performances, while the others are ensemble performances of the methods.