y_pred = np.zeros(X.shape[0], dtype=float)
for m in self.models:
y_pred += m.predict(X)
y_pred /= len(self.models)
return (y_pred >= 0.5).astype(int)
if __name__ == '__main__':
import sys
from os import path
sys.path.append(path.dirname(path.dirname(path.abspath(__file__))))
from kaggle_io.extract_inputs import extract_training_data, extract_testing_data
print 'Reading data...'
Id_train, X_train, y_train = extract_training_data(
'data/kaggle_train_tf_idf.csv')
Id_test, X_test = extract_testing_data('data/kaggle_test_tf_idf.csv')
print 'Reading models...'
H = []
H.append(joblib.load('rf1/1.pkl'))
H.append(joblib.load('rf1/2.pkl'))
H.append(joblib.load('rf1/3.pkl'))
H.append(joblib.load('ada1/1.pkl'))
H.append(joblib.load('ada1/2.pkl'))
print 'Hill climbing...'
ensemble = Ensemble()
ensemble.fit(X_train,
y_train,
H,
import sys
from os import path
sys.path.append(path.dirname(path.dirname(path.abspath(__file__))))
from kaggle_io.extract_inputs import extract_training_data
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import AdaBoostClassifier
from sklearn.externals import joblib
from CvModel import CvModel
Id, X, y = extract_training_data('data/kaggle_train_tf_idf.csv')
n_folds = 5
scaler = StandardScaler().fit(X)
ada = AdaBoostClassifier()
print 'Training AdaBoost with n_estimators=10'
ada.set_params(n_estimators=10)
cv_ada = CvModel(n_folds, scaler, ada)
cv_ada.fit(X, y)
joblib.dump(cv_ada, 'ada1/1.pkl')
print 'Training AdaBoost with n_estimators=50'
ada.set_params(n_estimators=50)
cv_ada = CvModel(n_folds, scaler, ada)
cv_ada.fit(X, y)
joblib.dump(cv_ada, 'ada1/2.pkl')