def naive_bayse_cross(train_x, train_y, validation, test, test_data):
print("training data...")
clf_pipe = make_pipeline(CountVectorizer(ngram_range=(1, 2)),
RandomUnderSampler(), MultinomialNB(alpha=0.01))
scores = cross_val_score(clf_pipe, train_x, train_y, cv=10)
print("Model is fitted!")
if validation:
print("scores: ", scores)
print("std of score: ", np.std(scores))
print("Accuracy: %0.2f (+/- %0.2f)" %
(scores.mean(), scores.std() * 2))
y_pred = cross_val_predict(clf_pipe, train_x, train_y, cv=5)
# Evaluation
# classification report
print("classification reports:",
classification_report(train_y, y_pred))
# confusion matrix
conf_mat = confusion_matrix(train_y, y_pred)
print(conf_mat)
plot_conf(conf_mat)
if test:
naive_bayes(test_data)
def test_custom_review(count_vec, train_vec, y_train_data):
print('\nTest a custom review message')
print('Enter review to be analysed: ', end=" ")
test = []
test_list = []
test.append(input())
test_review = pd.DataFrame(data={"id": 1, "review": test})
print("Cleaning the test review")
for i in range(0, len(test_review.review)):
test_list.append(clean_review(test_review.review[i]))
print("Vectorizing the test review")
test_review_vec = count_vec.transform(test_list)
print("Predicting")
pred_naive_bayes = naive_bayes(train_vec, test_review_vec, y_train_data)
if (pred_naive_bayes == 1):
print("The review is predicted positive")
else:
print("The review is predicted negative")
def __init__(self, config):
"""This class is a wrapper for a model which targets either the Purpose or Field category set.
"""
self.target_set = config['target_set']
self.model_config = config['model_config']
self.model_name = config['model']
self.model = None
# look for a non-default label set
self.labelset = None
if 'labels' in config.keys():
self.labelset = LabelTransformer(self.target_set, config['labels'])
else:
self.labelset = LabelTransformer(
self.target_set,
LabelTransformer.default_labels(self.target_set))
if self.target_set != 'purpose' and self.target_set != 'field':
raise ValueError('Unknown target_set configuration value: %s \n' %
config['target_set'])
# sort out which model the configuration contains
if self.model_name == 'LogisticRegression':
self.model = LogisticRegression()
elif self.model_name == 'DecisionTree':
self.model = DecisionTreeClassifier()
elif self.model_name == 'SVC':
self.model = SVC()
elif self.model_name == 'Naivebayes':
self.model = naive_bayes()
elif self.model_name == 'RandomForest':
self.model = RandomForestClassifier()
else:
sys.exit('Invalid config, model given is unknown: %s' % self.model)
self.model.set_params(**self.model_config)
return
def train_model(classifier, train_path, test_path, type_classification, train=True, validation=True, test=True,
cross_validation=False):
# collect train data
print("reading train set...")
if type_classification == "T":
# read titles and their label
train_x, train_y = collect_titles(train_path)
elif type_classification == "TB":
# read whole document
train_x, train_y = collect_documents(train_path)
elif type_classification == "TBW":
# weighted title and body
train_x, train_y = collect_weighted_doc(train_path)
else:
print("wrong argument")
# if test:
print("loading test data...")
test_data, reference = collect_test_documents(test_path)
# split data
if not cross_validation:
print("spliting the train set...")
train_data, validate_data, train_target, validate_target = train_test_split(train_x, train_y, test_size=0.4,
random_state=0)
# Naive bayes classifier
if classifier == "NB":
# train data set
if train:
print("training data...")
naive_bayes_train(train_data, train_target)
# validate validation set
if validation:
print("evaluating data...")
naive_bayes_evaluate(validate_data, validate_target)
# test data
if test:
print("testing data...")
naive_bayes(test_data, reference)
print("results are written in: \Results\Prediction.xlsx")
# SVM classifier
if classifier == "SVM":
# train data set
if train:
print("training data...")
svm_train(train_data, train_target)
# validate validation set
if validation:
print("evaluating data...")
svm_evaluate(validate_data, validate_target)
# test data
if test:
print("testing data...")
svm(test_data, reference)
print("results are written in: \Results\Prediction.xlsx")
# Logistic regression
if classifier == "LR":
# train data set
if train:
print("training data...")
train_logistic_regression(train_data, train_target)
# validate validation set
if validation:
print("evaluating data...")
validate_logistic_regression(validate_data, validate_target)
# test data
if test:
print("testing data...")
logistic_regression(test_data, reference)
print("results are written in: \Results\Prediction.xlsx")
# using cross validation
else:
if classifier == "NB":
naive_bayse_cross(train_x, train_y, validation, test, test_data)
if classifier == "SVM":
SVM_train_cross(train_x, train_y, validation, test, test_data)
print "SVM, N-Gram Vectors: ", accuracy
def random_forrest():
#RF on Count Vectors
accuracy = train_model(ensemble.RandomForestClassifier(), xtrain_count,
train_y, xvalid_count)
print "RF, Count Vectors: ", accuracy
# RF on Word Level TF IDF Vectors
accuracy = train_model(ensemble.RandomForestClassifier(), xtrain_tfidf,
train_y, xvalid_tfidf)
print "RF, WordLevel TF-IDF: ", accuracy
while choice != "q":
if choice == "1":
naive_bayes()
elif choice == "2":
linear_classifier()
elif choice == "3":
svm()
elif choice == "4":
random_forrest()
else:
print("Invalid choice, please choose again")
print("\n")
choice = getChoice()
delimiter="\t",
quoting=0)
y_train_data = train_data.sentiment
#Vectorization - TFIDF
print("Using TFIDF ")
train_vect, test_vec, count_vec = tfidf_vectorizer(train_list, test_list,
train_data, test_data)
#Dimensionality Reduction
train_vec, test_vec = dimensionality_reduction(train_vect, test_vec,
y_train_data)
#Prediction
pred_naive_bayes = naive_bayes(train_vec, test_vec, y_train_data)
pred_random_forest = random_forest(train_vec, test_vec, y_train_data)
pred_linear_svc = linear_svc(train_vec, test_vec, y_train_data)
pred_logistic = logistic_regression(train_vec, test_vec, y_train_data)
#Writing output of classifier with highest accuracy(Linear SVC)to csv
output = pd.DataFrame(
data={
"id": test_data.id,
"review": test_data.review,
"sentiment": pred_linear_svc
})
output.to_csv("tfidf_svc.csv", index=False)
print("Using pre-trained word2vec model")
train_list = []