Second place submission to "TUT Copper Analysis Challenge" https://inclass.kaggle.com/c/copper-analysis
Please refer to the model description PDF file to get more information about the competition and the best model solution.
Data can be downloaded from: https://inclass.kaggle.com/c/copper-analysis/data
Example run for reproducing the best model solution:
python Copperium.py predict train/ train.csv test/ test.csv -C 0.00001 > predictions_with_prob.tsv
cut -f1,2 predictions_with_prob.tsv | sed 's/\t/,/' > predictions_no_prob.csv
python Copperium.py retrain train/ train.csv test/ test.csv predictions_with_prob.tsv -C 0.00001 -t 0.7 > predictions_with_prob_iter1.tsv
cut -f1,2 predictions_with_prob_iter1.tsv | sed 's/\t/,/' > predictions_iter1_no_prob.csv