WCCI Performance Prediction Challenge

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

CLOP-models-only-5

Submitted by Juha Reunanen

Unfortunately, I just realized that my models for the previous submissions got overwritten when the algorithms were re-run (except for CLOP-models-only-2 which was computed on a different PC). This is because I only backed up the Results directory and not Models. So this is my final shot trying to get decent results (i.e., searching long enough) while having all the models available. Don't know if this has any significance though, as no bonus entries are available anymore.

Except for Hiva, this submission is identical to my submission -3. Of course, this means that the models for submissions -3 and -4 also happen to be available for datasets other than Hiva (which was for these submissions computed much earlier, already for submission -1). By the way, the model for Gina is 43 MB in size! Those for the other datasets are in the order of 1 MB, or less...

Dataset Balanced Error Test guess Guess error Test score Area Under Curve
Train Valid Test Train Valid Test
ada 0.1837 0.1881 0.2037 0.1884 0.0153 0.219 0.8954 0.8943 0.8825
gina 0 0 0.098 0.084 0.0141 0.1121 1 1 0.9631
hiva 0.2798 0.3317 0.3088 0.3172 0.0084 0.3148 0.8023 0.7191 0.7392
nova 0.005 0 0.0892 0.0917 0.0025 0.0907 1 1 0.9874
sylva 0.0084 0.0016 0.0341 0.032 0.0021 0.0357 0.9999 0.9999 0.9971
Overall 0.0954 0.1043 0.1468 0.1427 0.0085 0.1545 0.9395 0.9227 0.9138

This entry is a complete valid challenge entry.