def preprocess(corpus_folder, mode, settings, out_file, verbose_level):
preprocessor.format_corpus(corpus_folder)
preprocessor.clean_directory(corpus_folder + "_partitioned")
filter_words = True
if settings['filter_words'] == 0:
filter_words = False
preprocessor.split_files(settings['num_lines_split'],
settings['sliding_window_size'], filter_words,
corpus_folder, corpus_folder + "_partitioned")
all_files = load_files(corpus_folder + "_partitioned")
if verbose_level > 1:
print(
"mode : {} filter: {} window size: {} num_lines_split: {}".format(
mode, filter_words, settings['sliding_window_size'],
settings['num_lines_split']))
if mode == SVM:
run_svm(all_files, settings['svm']['num_runs'], out_file,
verbose_level)
elif mode == KMEANS:
param_dict = {}
param_dict['n_init'] = settings['kmeans']['n_init']
param_dict['max_iter'] = settings['kmeans']['max_iter']
param_dict['tol'] = settings['kmeans']['tol']
param_dict['num_runs'] = settings['kmeans']['num_runs']
param_dict['k'] = settings['kmeans']['k']
run_cluster(all_files, param_dict, out_file, verbose_level)
def main(args):
'''
flowers recognition gogogo!
'''
if args.method == 'vgg':
print('Using vgg network for flowers recognition')
vgg.run_vgg(args.lr, args.epochs, args.batch_size, args.reg)
if args.method == 'fc':
print('Using fully connected network for flowers recognition')
fc.run_fc(args.lr, args.epochs, args.batch_size, args.reg)
if args.method == 'resnet34':
print('Using deep residual network(34-layers) for flowers recognition')
resnet.run_resnet(args.lr, args.epochs, args.batch_size, args.reg)
if args.method == 'resnet50':
print(
'Using deep residual network(50-layers) pretrained for flowers recognition'
)
res.run_resnet50(args.batch_size, args.epochs, args.lr)
if args.method == 'svm':
print('Using Support Vector Machine for flowers recognition')
svm.run_svm()
if args.method == 'knn':
print('Using K nearest neighbors for flowers recognition')
knn.run_knn()
def main():
model = LogisticRegressionMulticlass.LogisticRegression()
model.start_logistic_regression(learning_rate=0.7,
num_epoch=50,
theta=1,
show_graph=False)
svm.run_svm(kernal=svm.linear, train_size=20000)
RandomForest.run_random_forest()
nn = NeuralNetwork.NeuralNetwork()
nn.start_neural_network(num_epoch=100,
learning_rate=0.015,
show_graph=False)
combine_model.combine_model(combine_model.MNIST)
combine_model.combine_model(combine_model.USPS)
def cross_validate(train_inputs, train_labels, identity):
kfold = cross_validation.StratifiedKFold(
identity, n_folds=3, shuffle=True, random_state=5)
errors = []
for train_i, valid_i in kfold:
train_x = train_inputs[train_i]
valid_x = train_inputs[valid_i]
train_y = train_labels[train_i]
valid_y = train_labels[valid_i]
error = svm.run_svm(train_x, train_y, valid_x, valid_y)
errors.append(error)
return errors
def test_images(labels, images, classifier):
global label_dict
sum = 0
for idx, image in enumerate(images):
img = cv2.imread(image)
correct_label = labels[idx]
rectified_img, boxes = get_candidates(img)
# Draw boxes and labels around ROI
img_clone = rectified_img.copy()
answers = []
for box in boxes:
(x, y, w, h) = box
cv2.rectangle(img_clone, (x, y), (x + w, y + h), (0, 255, 0), 1)
# Run classifier
h = hog(rectified_img, box)
output = 0
if classifier == "svm-linear" or classifier == "svm-rbf":
output = int(run_svm(h)[1][0][0])
elif classifier == "rf":
output = int(run_rf(h)[1][0][0])
elif classifier == "mlp":
output = int(run_mlp(h))
else:
ValueError("Wrong classifier")
exit(1)
output_str = label_dict[output]
answers.append(output)
cv2.putText(img_clone, output_str, (x, y - 3), 3, 1, (0, 255, 0),
2, cv2.LINE_AA)
Image.fromarray(img_clone).show()
if correct_label in answers:
sum += 1
# Reverse Rectification with labels
# img_clone = reverse_rectification(img_clone)
# Image.fromarray(img_clone).show()
# Evaluation
print("Test Accuracy: {:6}".format(sum / len(labels)))
def test_sliding_window():
# list of testing set accuracies
test_error_list = []
overlap_list = []
overlap = 20
while overlap >= 0:
print("overlap num : {}".format(overlap))
preprocessor.format_corpus("sermons")
preprocessor.clean_directory("sermons_partitioned")
preprocessor.split_files(38, overlap, True, "sermons", "sermons_partitioned")
all_files = load_files("./sermons_partitioned/")
test_error_list.append(100 - run_svm(all_files, 4, None, 3))
overlap_list.append(overlap)
overlap -= 2
print(test_error_list)
plot_sliding_window(test_error_list, overlap_list)
def test_split_num():
# list of testing set accuracies
test_error_list = []
split_num_list = []
split_num = 38
while split_num >= 4:
print("split num : {}".format(split_num))
preprocessor.format_corpus("sermons")
preprocessor.clean_directory("sermons_partitioned")
preprocessor.split_files(split_num, 0, True, "sermons", "sermons_partitioned")
all_files = load_files("./sermons_partitioned/")
test_error_list.append(100 - run_svm(all_files, 4, None, 3))
split_num_list.append(split_num)
split_num -= 2
print(test_error_list)
plot_split_num(test_error_list, split_num_list)
print('# Classifying Online Review Sentiment with Machine Learning #')
print('# #')
print('#################################################################')
print()
dataset = SentimentCorpus()
nb = MultinomialNaiveBayes()
params = nb.train(dataset.train_X, dataset.train_y)
predict_train = nb.test(dataset.train_X, params)
eval_train = nb.evaluate(predict_train, dataset.train_y)
predict_test = nb.test(dataset.test_X, params)
eval_test = nb.evaluate(predict_test, dataset.test_y)
print("\n=======================================================\n")
print("+++ Naive Bayes +++")
print ("Accuracy on training data = %f \n Accuracy on testing data = %f" % (eval_train, eval_test))
print("Confusion Matrix:")
print(confusion_matrix(dataset.test_y,predict_test))
print(classification_report(dataset.test_y,predict_test))
print("=======================================================\n")
print("+++ Support Vector Machine +++")
svm.run_svm(dataset.train_X, dataset.train_y, dataset.test_X, dataset.test_y)
print("=======================================================\n")
print("+++ Neural Network +++")
nn.run_nn(dataset.train_X, dataset.train_y, dataset.test_X, dataset.test_y)
print("=======================================================")
#######################
if chosen_classifier == "knn":
# Enable to run knn classifier
print("Running knn classifier")
# knn.run_knn(train_images, train_labels, valid_images, valid_labels)
# knn.run_knn(lbp_train_images, train_labels, lbp_val_images, valid_labels)
# knn.run_knn(exp_train_images, exp_train_labels, exp_val_images, exp_val_labels)
knn.run_knn(gc_train, exp_train_labels, gc_val, exp_val_labels)
elif chosen_classifier == "svm":
# svm classifier
# print("Running svm classifier")
# svm.run_svm(exp_train_images, exp_train_labels, exp_val_images, exp_val_labels)
print("Running svm classifier on gamma corrected images")
svm.run_svm(gc_train, exp_train_labels, gc_val, exp_val_labels)
elif chosen_classifier == "mog":
# mog classifier
print("Running mog classifier")
mog.run_mog(gc_train, exp_train_labels, gc_val, exp_val_labels)
elif chosen_classifier == "dt":
print("Running decision tree classifier")
dt.decision_tree(gc_train, exp_train_labels, gc_val, exp_val_labels)
elif chosen_classifier == "ensemble":
print("Running ensemble method")
ensemble.run_ensemble(exp_train_images, exp_train_labels, exp_val_images, exp_val_labels)
else:
if __name__ == '__main__':
x_train, x_test, y_train, y_test = get_data(True)
print(
"\n-------------------------------------\nAccuracies with top 5 features:\n-------------------------------------"
)
run_decision_tree(x_train, x_test, y_train, y_test)
run_k_nearest_neighbour(x_train, x_test, y_train, y_test)
run_logistic_regression(x_train, x_test, y_train, y_test)
run_naive_bayes(x_train, x_test, y_train, y_test)
run_neural_network(x_train, x_test, y_train, y_test)
run_perceptron(x_train, x_test, y_train, y_test)
run_random_forest(x_train, x_test, y_train, y_test)
run_svm(x_train, x_test, y_train, y_test)
run_xg_boost(x_train, x_test, y_train, y_test)
print(
"\n-------------------------------------\nAccuracy with Voting in top 5 features:\n-------------------------------------"
)
run_voting(x_train, x_test, y_train, y_test)
x_train, x_test, y_train, y_test = get_data()
print(
"\n-------------------------------------\nAccuracies with all 22 features:\n-------------------------------------"
)
run_decision_tree(x_train, x_test, y_train, y_test)
run_k_nearest_neighbour(x_train, x_test, y_train, y_test)
run_logistic_regression(x_train, x_test, y_train, y_test)
#logging.info('This is an info log')
#logging.warning('This is a warning log')
#logging.error('This is an error log')
#*******************************************************************************
# DATA EXTRACTION
TRAIN_DATA = data_extractor.get_data(defines.DATA_TRAIN_CSV_FILE)
TEST_DATA = data_extractor.get_data(defines.DATA_TEST_CSV_FILE)
#*******************************************************************************
# STOP WORDS FILTER
#logging.info('Prepare Data')
#nb_lib.nb_lib_prepare(TRAIN_DATA)
#*******************************************************************************
# FEATURE SELECTION
logging.info('Feature selection')
#feature_select.get_selected_features(TRAIN_DATA)
#*******************************************************************************
logging.info('SVM')
svm.run_svm(TRAIN_DATA, TEST_DATA)
logging.info ('Cross Validation')
#CrossVal.run_crossval(TRAIN_DATA)
print ("done!")
sys.exit()
def execute(topic1, topic2, test, dump_files):
if dump_files == "True":
print_bold("\n" + "Downloading the datasets ..." + "\n")
create_cleaned_files(topic1, topic2, test)
print_bold("Dumps TFIDF features ..." + "\n")
# category is used to specify the unique Id of the dumped model
category = topic1 + "-" + topic2
dump_tfidf(category)
print("=========================================================")
print_bold("Start Running bayes model to establish a baseline")
print("=========================================================")
print_bold("\n" + "Run Bayes model ..." + "\n")
pred_train_bayes, pred_test_bayes = run_bayes(category)
print("=========================================================")
print_bold("Improvement of the baseline")
print("=========================================================")
print_bold("Run Cnn model ..." + "\n")
pred_train_cnn, pred_test_cnn = run_cnn()
print(
"--------------------------------------------------------------------------"
)
print_bold("Run Fasttext model ..." + "\n")
pred_train_fasttext, pred_test_fasttext = run_fasttext()
print(
"--------------------------------------------------------------------------"
)
print_bold("Run SVM model ..." + "\n")
pred_train_svm, pred_test_svm, y_train = run_svm(category)
print(
"--------------------------------------------------------------------------"
)
print_bold("Run Logistic Regression model ..." + "\n")
pred_train_logreg, pred_test_logreg, y_test = run_logreg(category)
print(
"--------------------------------------------------------------------------"
)
print_bold("Starting Ensemble Method")
# using train+val for training the ensemble (training on more dataset == stronger results)
train = np.column_stack((pred_train_svm, pred_train_logreg, pred_train_cnn,
pred_train_fasttext))
test = np.column_stack(
(pred_test_svm, pred_test_logreg, pred_test_cnn, pred_test_fasttext))
model = xgb().fit(train, y_train)
print(
"--------------------------------------------------------------------------"
)
print_bold("Final results on the test set : ")
print(classification_report(y_test, model.predict(test)))
