Performance Prediction Challenge

SNB(CMA) + 10k F(3D) tv

Submitted by Marc Boulle

Same method as SNB(CMA) tv

Test of statistical and computational scalability

10 000 features constructed for each dataset
Each one is the sum of three randomly selected initial features

Dataset	Balanced Error			Test guess	Guess error	Test score	Area Under Curve
	Train	Valid	Test				Train	Valid	Test
ada	0.1408	0.1642	0.171	0.159	0.012	0.1829	0.934	0.9327	0.9159
gina	0.04	0.035	0.0826	0.091	0.0084	0.091	0.9923	0.99	0.9731
hiva	0.2405	0.2745	0.3102	0.298	0.0122	0.3203	0.8186	0.7695	0.7591
nova	0.0412	0.08	0.0835	0.099	0.0155	0.099	0.9925	0.9702	0.974
sylva	0.0034	0.0033	0.0061	0.006	0.0001	0.0062	0.9997	0.9996	0.9991
Overall	0.0932	0.1114	0.1307	0.1306	0.0096	0.1399	0.9474	0.9324	0.9242

This entry is a complete valid challenge entry.

WCCI 2006 Workshop on Model Selection Web Site Maintenance: MPS  (contact)