def trainxgb(model_id,train_x,train_y,valid_x,valid_y,test_x):
train_x,train_y=shuffle(train_x,train_y)
random_state=random.randint(0, 1000000)
print('random state: {state}'.format(state=random_state))
xgboost = XGBoostClassifier(silent=0,
objective='multi:softprob',
eval_metric='mlogloss',
num_class=9,
nthread=4,
seed=random_state,
eta=random.uniform(0.01,0.1),
max_depth=random.randint(10,20),
max_delta_step=random.randint(1,10),
min_child_weight=random.randint(1,10),
subsample=random.uniform(0.0,1.0),
gamma=random.uniform(0.01,0.1),
colsample_bytree=random.uniform(0.0,1.0),
early_stopping_rounds=30,
num_round=1000
)
xgboost.fit(train_x, train_y)
valid_predictions = xgboost.predict_proba(valid_x)
score =log_loss(valid_y,valid_predictions)
print score
fName = open("report_xgb_"+str(model_id)+".txt", 'w')
print >> fName, "score:"
print >> fName, score
print >>fName, "model_id:"
print >>fName, model_id
print >> fName, xgboost
fName.close()
### building the classifiers
clfs = []
xgb1 = XGBoostClassifier(
alpha=0,
booster="gbtree",
colsample_bytree=0.391995799463,
early_stopping_rounds=30,
eta=0.0238249854939,
eval_metric="mlogloss",
gamma=0.0339065215885,
l=0,
lambda_bias=0,
max_delta_step=3,
max_depth=19,
min_child_weight=8,
nthread=4,
ntree_limit=0,
num_class=9,
num_round=2000,
objective="multi:softprob",
seed=463324,
silent=1,
subsample=0.732463140484,
use_buffer=True,
)
xgb1.fit(train_x, train_y)
print("xgb1 LogLoss {score}".format(score=log_loss(test_y, xgb1.predict_proba(test_x))))
clfs.append(xgb1)
