n_classes=3,
batch_size=10,
print_steps=20)
classifier = skflow.TensorFlowEstimator(model_fn=char_cnn_model,
n_classes=3,
steps=100,
optimizer='Adam',
learning_rate=0.01,
continue_training=True)
# Write results to file:
f = fileWriter.initFile("../TextFiles/FindingsAndResults/ex9/ex9")
# Continuously train for 1000 steps & predict on test set.
i = 0
print("Initiating training...")
fileWriter.writeTextToFile("Initiating training...", f)
while i < 11:
print(80 * '=')
fileWriter.writeTextToFile(80 * '=', f)
classifier.fit(X_train,
y_train,
val_monitor,
logdir='../TextFiles/logs/cnn_on_characters')
pred_stances = classifier.predict(X_val)
score = metrics.accuracy_score(y_val, pred_stances)
print("Accuracy: %f" % score)
fileWriter.writeTextToFile("Accuracy: %f" % score, f)
print(classification_report(y_val, pred_stances, digits=4))
logging.basicConfig(level=logging.INFO,
format='%(asctime)s %(levelname)s %(message)s')
###############################################################################
# Load some categories from the training set
# Uncomment the following to do the analysis on all the categories
#categories = None
data = pd.read_csv('../TextFiles/data/tcp_train.csv', sep='\t')
cv = StratifiedKFold(data.Stance, n_folds=10, shuffle=True, random_state=1)
print("%d documents" % len(data))
write.writeTextToFile("%d documents" % len(data),file)
print("%d categories" % 3)
write.writeTextToFile("%d categories" % 3,file)
print()
###############################################################################
# define a pipeline combining a text feature extractor with a simple
# classifier
pipeline = Pipeline([
('features', FeatureUnion([
('unigram_word', Pipeline([
('vect', CountVectorizer())
])),
('td-idf', Pipeline([
('vect', CountVectorizer()),
('tfidf', TfidfTransformer())
# Display progress logs on stdout
logging.basicConfig(level=logging.INFO,
format='%(asctime)s %(levelname)s %(message)s')
###############################################################################
# Load
strength = 'soft'
#data = pd.read_csv('../../TextFiles/data/tcp_train.csv', sep='\t')
data = ptd.getTrainingData()
data = data[data.Stance != 'NONE']
cv = StratifiedKFold(data.Stance, n_folds=10, shuffle=True, random_state=1)
print("%d training documents" % len(data.Abstract))
write.writeTextToFile("%d training documents" % len(data.Abstract), file)
print("%d categories" % 3)
write.writeTextToFile("%d categories" % 3, file)
print()
###############################################################################
# Classifiers
# MultinomialNB(), BernoulliNB(), SVM(), LinearSVM(), SGDClassifier(), LogisticRegression()
clf = MultinomialNB()
print("Using train, validation and test approach with clf {}".format(clf))
write.writeTextToFile(
"Using train, validation and test approach with clf {}".format(clf), file)
###############################################################################
# define a pipeline combining a text feature extractor with a simple classifier
format='%(asctime)s %(levelname)s %(message)s')
###############################################################################
# Load some categories from the training set
# Uncomment the following to do the analysis on all the categories
#categories = None
data = pd.read_csv(open('../../TextFiles/data/tcp_train.csv'),
sep='\t',
index_col=0)
cv = StratifiedKFold(data.Stance, n_folds=10, shuffle=True, random_state=1)
print("%d documents" % len(data))
write.writeTextToFile("%d documents" % len(data), file)
print("%d categories" % 3)
write.writeTextToFile("%d categories" % 3, file)
print()
###############################################################################
# define a pipeline combining a text feature extractor with a simple
# classifier
pipeline = Pipeline([
('vect', CountVectorizer()),
#('tfidf', TfidfTransformer()),
#('clf', MultinomialNB()),
#('clf', LinearSVC()),
('clf', LogisticRegression()),
#('clf', SVC()),
])
val_monitor = skflow.monitors.ValidationMonitor(X_val, y_val,
early_stopping_rounds=200,
n_classes=3,
batch_size=10,
print_steps=20)
classifier = skflow.TensorFlowEstimator(model_fn=char_cnn_model, n_classes=3,
steps=100, optimizer='Adam', learning_rate=0.01,
continue_training=True)
# Write results to file:
f = fileWriter.initFile("../TextFiles/FindingsAndResults/ex9/ex9")
# Continuously train for 1000 steps & predict on test set.
i = 0
print("Initiating training...")
fileWriter.writeTextToFile("Initiating training...", f)
while i<11:
print(80 * '=')
fileWriter.writeTextToFile(80 * '=' , f)
classifier.fit(X_train, y_train, val_monitor, logdir='../TextFiles/logs/cnn_on_characters')
pred_stances = classifier.predict(X_val)
score = metrics.accuracy_score(y_val, pred_stances)
print("Accuracy: %f" % score)
fileWriter.writeTextToFile("Accuracy: %f" % score, f)
print (classification_report(y_val, pred_stances, digits=4))
fileWriter.writeTextToFile(classification_report(y_val, pred_stances, digits=4), f)
macro_f = fbeta_score(y_val, pred_stances, 1.0,
###############################################################################
# Load
strength = 'soft'
#data = pd.read_csv('../../TextFiles/data/tcp_train.csv', sep='\t')
data = ptd.getTrainingData()
binaryStances = []
for endorse in data.Endorse.tolist():
binaryStances.append(ptd.getAbstractStanceVsNoStance(strength, endorse))
cv = StratifiedKFold(binaryStances, n_folds=10, shuffle=True, random_state=1)
print("%d training documents" % len(data.Abstract))
write.writeTextToFile("%d training documents" % len(data.Abstract),file)
print("%d categories" % 3)
write.writeTextToFile("%d categories" % 3,file)
print()
###############################################################################
# Classifiers
# MultinomialNB(), BernoulliNB(), SVM(), LinearSVM(), SGDClassifier(), LogisticRegression()
clf = MultinomialNB()
print("Using train, validation and test approach with clf {}".format(clf))
write.writeTextToFile("Using train, validation and test approach with clf {}".format(clf), file)
###############################################################################
# define a pipeline combining a text feature extractor with a simple classifier
pipeline = Pipeline([
