#!/usr/bin/python
import xgboost

dtrain = xgboost.DMatrix('mq2008.train')
dvalid = xgboost.DMatrix('mq2008.vali')
dtest = xgboost.DMatrix('mq2008.test')

params = {'objective': 'rank:ndcg', 'eta': 0.01, 'gamma': 1.0,
          'min_child_weight': 0.1, 'max_depth': 8, 'silent': 1, 'eval_metric':'ndcg'}
# num_boost_round=713 was chosen using early stopping on validation set mq2008.vali
xgb_model = xgboost.train(params, dtrain, num_boost_round=713,
                          evals=[(dtrain, 'train'), (dtest, 'test'), (dvalid, 'validation')],
                          verbose_eval=True)
xgb_model.save_model('mq2008.model')