print 'Downsample favor: ' + str(downsample_rate_favor)
print 'Downsample none: ' + str(downsample_rate_none)
train_data = ptd.getTrainingData()
validate_data = ptd.getValidationData()
#test_data = ptd.getTestData()
sub_none = ptd.getDownsample2_0(train_data, "NONE", strength, downsample_rate_none)
sub_favor = ptd.getDownsample2_0(train_data, "FAVOR", strength, downsample_rate_favor)
against = train_data[train_data.Stance == "AGAINST"]
train_data = pd.concat([sub_favor, sub_none, against])
else:
print("using nothing")
train_data = ptd.getTrainingData()
validate_data = ptd.getValidationData()
test_data = ptd.getTestData()
if use_upsample:
print("using up sampling")
train_data = pd.concat([train_data, train_data[train_data.Stance == "AGAINST"]])
cv = StratifiedKFold(train_data.Stance, n_folds=10, shuffle=True, random_state=1)
# Select classifiers to use
classifiers = [
#LinearSVC(C=1.178),
SVC(C=6.9183097091893631, kernel='linear', shrinking=True)
#MultinomialNB(alpha=0.1, fit_prior=False)
#LogisticRegression()
]
from sklearn.metrics import classification_report
from sklearn.cross_validation import cross_val_predict, StratifiedKFold
from sklearn.metrics import fbeta_score
from sklearn.svm import LinearSVC, SVC
from sklearn.linear_model import LogisticRegression, SGDClassifier
from sklearn.naive_bayes import MultinomialNB, BernoulliNB
import pandas as pd
import sklearn
strength = 'soft'
# ***** LOAD DATA *****
train_data = ptd.getTrainingData()
validate_data = ptd.getValidationData()
test_data = ptd.getTestData()
cv = StratifiedKFold(train_data.Stance,
n_folds=10,
shuffle=True,
random_state=1)
# Select classifiers to use
classifiers = [
LinearSVC(C=2.3988329190194899, multi_class='crammer_singer'),
#SVC(C=5.2, kernel='linear')
#MultinomialNB(alpha=0.63, fit_prior=True)
#LogisticRegression(C=22.759, penalty='l2', solver='lbfgs')
#SGDClassifier(alpha=0.0001, loss='squared_hinge')
#BernoulliNB(alpha=0.1, fit_prior=True)
]
import json
#################
# Parameters #
#################
store_to_file = 0
################
# Load Data #
################
print("Loading data...")
train_data = pd.concat([ptd.getTrainingData(), ptd.getValidationData(), ptd.getTestData()])
unlabelled_data = ptd.getUnlabelledData()
#########################
# Train classifier #
#########################
print("Training classifier")
best_classifier = LinearSVC(C=1.178)
pipeline = Pipeline([('vect', CountVectorizer(decode_error='ignore',
analyzer='word',
ngram_range=(1, 2),
stop_words= None,