Kaggle Seizures Competition

Changelog

14.09.16 18:59

  • Raw FFT results into features
  • Saving and loading npz properly
  • Some tests (unsuccessfull)

10.09.16 00:07

  • Refactoring of the notebook (deleted extra cells, added md-titles, cleared output)

09.09.16 23:32

  • mat_to_pd()
  • FFT histogram bins as features (PCA shows that it's a bad set of features + RF shows results ~0.5 roc_auc)
  • save_to_npz()
  • t-test showed that number of peaks in data does not matter