-
Notifications
You must be signed in to change notification settings - Fork 0
/
blending.py
200 lines (153 loc) · 6.28 KB
/
blending.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
import numpy as np
import pathlib
from datetime import datetime
# classifiers
from sklearn.ensemble import AdaBoostClassifier, GradientBoostingClassifier, ExtraTreesClassifier, RandomForestClassifier, RandomForestRegressor
from sklearn.linear_model import LogisticRegression, PassiveAggressiveClassifier, RidgeClassifier, SGDClassifier
from sklearn.linear_model import LogisticRegressionCV, RidgeClassifierCV
from sklearn.neighbors import KNeighborsClassifier
from sklearn.neural_network import BernoulliRBM, MLPClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.gaussian_process import GaussianProcessClassifier
from sklearn.gaussian_process.kernels import RBF
from sklearn.model_selection import GridSearchCV, ShuffleSplit
from sklearn.preprocessing import Normalizer
from sklearn.pipeline import Pipeline
# import matplotlib.pyplot as plt
from sklearn.ensemble import VotingClassifier
from sklearn.model_selection import cross_val_score
def load_data(filename, train=True):
"""
Function loads data stored in the file filename and returns it as a numpy ndarray.
Inputs:
filename: given as a string
(optional) train: used to determine whether this is the training or test set
Outputs:
Data contained in the file, returned as a numpy ndarray
"""
X = []
y = []
with open(filename) as f:
for line in f:
if (train):
# remove \n, split on space, separate into label and weights
X.append(line.strip().split(' ')[1:])
y.append(line.strip().split(' ')[0])
else:
X.append(line.strip().split(' '))
# convert to np, cast to int, and remove the headers
X = np.asarray(X[1:]).astype(int)
if (train):
y = np.asarray(y[1:]).astype(int)
return X, y
def split_data(x_train, y_train):
'''
Function for cross validiation.
Inputs:
x_train: training data points
y_train: training labels
Outputs:
trainX: randomized 4/5 of given data points
trainY: corresponding labels
testX: randomized 1/5 of given data points
testY: corresponding lables
'''
dataSplit = ShuffleSplit(n_splits = 1, test_size = 0.2)
for train, test in dataSplit.split(x_train, y_train):
return x_train[train], y_train[train], x_train[test], y_train[test]
def normalization(X_train, X_test):
'''
Function to normalize training and test data
Inputs:
X_train: training set data points
X_test: test set data points
Outputs:
train_norm: normalized training set data points
test_norm: normalized test set data points
'''
normalizer = Normalizer().fit(X_train)
train_norm = normalizer.transform(X_train)
test_norm = normalizer.transform(X_test)
return (train_norm, test_norm)
def make_predictions(clf, X, y, test):
'''
Function to train and test our classifier
Inputs:
clf: classifier
X: data points
y: labels
test: test set
Outputs:
predictions: predictions from running the clf on the test set
'''
clf.fit(X, y)
predictions = clf.predict(test)
predictions = predictions.astype(int)
return predictions
def percentError(yPred, yTrue):
'''
Calculates the percent error between two given label sets
Inputs:
yPred: predicted labels
yTrue: actual labels
Outputs:
error: float of the number of mismatches divided by total length
'''
return 1.0-np.sum(np.equal(yPred, yTrue))/len(yTrue)
def save_data(data, filename="%s.txt" % datetime.today().strftime("%X").replace(":", "")):
'''
Function to save the predictions by the classifier
Inputs: predictions, (optional) filename
If filename isn't specified, then it just uses the current time
Outputs: Does not return anything
Writes the submisssion to a textfile that should have the same format as the sample_submission.txt
'''
# Creates a new submissions folder if one doesn't exist
pathlib.Path('submissions').mkdir(parents=True, exist_ok=True)
with open("submissions\\%s" % filename, "w") as f:
f.write("Id,Prediction\n")
for Id, prediction in enumerate(data, 1):
string = str(Id) + ',' + str(prediction) + '\n'
f.write(string)
def main():
# attempt at blending?
# load the data
X_train, y_train = load_data("training_data.txt")
X_test, _ = load_data("test_data.txt", False)
# normalize training and test data
X_train_n, X_test_n = normalization(X_train, X_test)
# split the data in to training and testing so we can test ourselves
trainX, trainY, testX, testY = split_data(X_train_n, y_train)
# PUT THE THINGS WE WANT TO BLEND HERE.
logclf = LogisticRegression(C=2.7825594)
SVCclf = SVC(gamma=1, C=2, probability = True)
mlpclf = MLPClassifier(activation = 'logistic', hidden_layer_sizes=(50, 50, 50))
# gnbclf = GaussianNB()
etclf = ExtraTreesClassifier(n_estimators=750, n_jobs=-1, verbose=1)
adaclf = AdaBoostClassifier(base_estimator=etclf, n_estimators=40)
# can add weights to this
votingclf = VotingClassifier(estimators=[('ada', adaclf),
('svc', SVCclf),
('mlp', mlpclf),
# ('nb', gnbclf),
('log', logclf)
], voting='soft', n_jobs=-1)
# print('Fitting to training data...')
# voting = make_predictions(votingclf, trainX, trainY, testX)
# print("Voting error:", percentError(voting, testY))
# 0.13975
print('Fitting to testing data...')
try:
votingsubmission = make_predictions(votingclf, X_train_n, y_train, X_test_n)
except Exception as e:
print ("Exception: \n\n", e)
save_data(votingsubmission, "votingsubmission.txt")
print('All done! \n')
# for clf, label in zip([test1, test2, test3, blend],
# ['Logistic Regression', 'LogReg', 'MLPClass', 'Ensemble']):
# # scores = cross_val_score(clf, X_train_n, y_train, cv=5, scoring = 'accuracy')
# print("Accuracy: %0.8f (+/- %0.8f) [%s]" % (scores.mean(), scores.std(), label))
if __name__ == '__main__':
main()