WCCI Performance Prediction Challenge

The challenge is over, but a new challenge is on-going using the same datasets, check it out!

KTA+CV+SVM (5)

Submitted by Tobias Glasmachers

A balanced version of the Kernel-Target-Alignment
(KTA) is used for kernel parameter adaptation, cross
validation (CV) for complexity control and a C-SVM
for prediction. Performance estimated by hand,
strongly guided by cross validation results and
J. Platt's probabilistic SVM output interpretation.

Dataset Balanced Error Test guess Guess error Test score Area Under Curve
Train Valid Test Train Valid Test
ada 0.1413 0.1365 0.1918 0.161 0.0308 0.2225 0.8618 0.8659 0.8082
gina 0 0 0.071 0.066 0.005 0.0759 1 1 0.929
hiva 0.2807 0.277 0.3901 0.286 0.1041 0.4941 0.7455 0.7622 0.6099
nova 0.0154 0.008 0.062 0.06 0.002 0.0632 0.9827 0.9846 0.938
sylva 0.0075 0.0065 0.0153 0.0192 0.0038 0.0192 0.9923 0.9925 0.9847
Overall 0.089 0.0856 0.146 0.1184 0.0291 0.175 0.9165 0.921 0.854

This entry is a complete valid challenge entry.