Home

Awesome

kaggle-tradeshift-text-classification

My solution for kaggle competition - Tradeshift Text Classification. Finished 71/375 (my first top 25%) with 0.0061

  1. run fast_solutionMod#14.py. Take only 4th iteration into final averaging. (public score around 0.0071)
  2. run tradeshift#5_all.py. I forgot how long but I think it takes few hours. (public score around 0.01)
  3. run tradeshift_h20_improv.R. It took 42h on my 16Gb 4-core machine. (public score around 0.0086)
  4. run tradeshift_sum.py. This file averages solutions from first three models (files). (private score around 0.0061)

2nd and 3rd model aren't tuned, just trained once (didn't have patience/time).