Add ensemble of rf and gbm for otto-product-classification

otto-group-product-classification/code/ensemble.py

@@ -0,0 +1,63 @@
+import pandas as pd
+import numpy as np
+from sklearn import ensemble, feature_extraction, preprocessing, cross_validation, metrics
+
+def validate():
+	X_train, X_test, y_train, y_test = cross_validation.train_test_split(train, labels, test_size=0.2)
+	# Train a random forest classifier
+	# 0.5428 with (500, 25, 2)
+	n_estimators = 500
+	max_features = 25
+	min_samples_split = 2
+	for max_features in [10,15,20,25,30]:
+		rf_clf = ensemble.RandomForestClassifier(n_jobs=-1, n_estimators=n_estimators, max_features = max_features, min_samples_split = min_samples_split, verbose=1)
+		rf_clf.fit(X_train, y_train)
+		rf_preds = rf_clf.predict_proba(X_test)
+		print metrics.log_loss(y_test, rf_preds)
+
+	# # Train a gradient boosting classifier
+	# n_estimators = 50
+	# max_depth = 6
+	# for max_depth in [6,7,8,9,10]:
+	# 	gbm_clf = ensemble.GradientBoostingClassifier(n_estimators=n_estimators, max_depth = max_depth, verbose=1)
+	# 	gbm_clf.fit(X_train, y_train)
+	# 	gbm_preds = gbm_clf.predict_proba(X_test)
+	# 	print metrics.log_loss(y_test, gbm_preds)
+
+	# preds = (rf_preds+gbm_preds)/2
+	# print metrics.log_loss(y_test, preds)
+
+
+def predict():
+	# Train a random forest classifier
+	rf_clf = ensemble.RandomForestClassifier(n_jobs=-1, n_estimators=1000, max_features = 25, min_samples_split=2, verbose=1)
+	rf_clf.fit(train, labels)
+	rf_preds = rf_clf.predict_proba(test)
+
+	# Train a gradient boosting classifier
+	gbm_clf = ensemble.GradientBoostingClassifier(n_estimators=60, max_depth = 6, verbose=1)
+	gbm_clf.fit(train, labels)
+	gbm_preds = gbm_clf.predict_proba(test)
+
+	preds = (rf_preds*0.8+gbm_preds*0.2)
+	# create submission file
+	preds = pd.DataFrame(preds, index=sample.id.values, columns=sample.columns[1:])
+	preds.to_csv('benchmark.csv', index_label='id')
+
+if __name__ == '__main__':
+	# import data
+	train = pd.read_csv('../data/train.csv')
+	test = pd.read_csv('../data/test.csv')
+	sample = pd.read_csv('../data/sampleSubmission.csv')
+
+	# drop ids and get labels
+	labels = train.target.values
+	train = train.drop('id', axis=1)
+	train = train.drop('target', axis=1)
+	test = test.drop('id', axis=1)
+
+	# encode labels 
+	lbl_enc = preprocessing.LabelEncoder()
+	labels = lbl_enc.fit_transform(labels)
+
+	predict()