daxiongshu/tradeshift-text-classification

This code generates the 1st place solution of Tradeshift Text Classification from our team "carl and snow"

https://www.kaggle.com/c/tradeshift-text-classification

It mainly includes two kinds of models:

1. two-stage models using Xgboost and sklearn.
2. online logistic regression.

Dependencies Python 2.7 pypy 2.4.0 Scikit learn-0.15.2 numpy 1.7.1 scipy 0.11.0 Xgboost 0.3

To generate a solution:

1. Set Up all the dependencies
2. change the data dir in run.py
3. change the xgboost wrapper path in ./src/xgb_classifier.py
4. python run.py

The best single solution: xgb-part1-d18-e0.09-min6-tree120-xgb_base.csv private LB 0.0044595

The best ensemble solution: best-solution.csv private LB 0.0043324 (1st place)