# In[2]:
device_brand = pd.read_csv('new_feature/device_brand.csv')
# label = sex+age, one-hot encode
# In[3]:
# trian data , test data
train_datapath = '../Demo/deviceid_train.tsv'
test_datapath = '../Demo/deviceid_test.tsv'
train_data, test_data = LoadData(train_datapath, test_datapath)
# In[4]:
train_data.drop(['sex', 'age'], axis=1, inplace=True)
test_data['label'] = 'unknow'
data = train_data.append(test_data)
# Merge device_brand
# In[5]:
data = data.merge(device_brand, on='device_id', how='left')
data.fillna('unknow', inplace=True)
# In[6]:
fold_names = list(range(n_folds))
fold_names.append('overall')
# Dataframe of validation scores
metrics = pd.DataFrame({
'fold': fold_names,
'train_logloss': train_logloss,
'valid_logloss': valid_logloss
})
return metrics, out_of_fold, test_predictions
# In[ ]:
useless_feature = FeatureSelect(train_data)
# In[ ]:
train = train_data.drop(useless_feature, axis=1)
test = test_data.drop(useless_feature, axis=1)
# In[ ]:
get_ipython().run_line_magic(
'time',
"metric, train_proba, test_proba = model(train, test, 'label', 22, 10)")
# In[ ]:
np.save('new_feature/lgbcnt_train.npy', train_proba)
np.save('new_feature/lgbcnt_test.npy', test_proba)