def post(self, id):
parser = reqparse.RequestParser()
parser.add_argument('file',
type=FileStorage,
required=True,
location='files')
args = parser.parse_args()
repository_dao = RepositoryDao(self.db_session())
repository = repository_dao.retrieve(id=id)
args['repository'] = repository
args['storage_id'] = generate_string()
args['storage_path'] = os.path.join(current_app.root_path,
self.config()['UPLOAD_DIR'],
args['storage_id'])
args['file'].save(args['storage_path'])
args['name'] = args['file'].filename
args['extension'] = '.'.join(args['name'].split('.')[1:])
args['content_type'] = 'application/octet-stream'
args['media_link'] = args['storage_path']
args['size'] = 0
# Remove 'file' element in the arguments because the File constructor
# does not expect it
del args['file']
f_dao = FileDao(self.db_session())
f = f_dao.create(**args)
return f.to_dict(), 201
def post(self):
parser = reqparse.RequestParser()
parser.add_argument('file',
type=FileStorage,
required=True,
location='files')
args = parser.parse_args()
args['storage_id'] = generate_string()
args['storage_path'] = os.path.join(current_app.root_path,
self.config()['UPLOAD_DIR'],
args['storage_id'])
args['file'].save(args['storage_path'])
args['name'] = args['file'].filename
args['extension'] = '.'.join(args['name'].split('.')[1:])
args['content_type'] = 'application/octet-stream'
args['media_link'] = args['storage_path']
args['size'] = 0
del args['file']
f_dao = FileDao(self.db_session())
f = f_dao.create(**args)
html = '<h3>Thanks for uploading your file!</h3>'
html += '<p>You can get it here: <a href="/files/{}/content">download</a></p>'.format(
f.id)
html += '<br>'
html += '<p>Next step is to create a new classifier by clicking the button below.</p>'
html += '<form method="post" action="/classifiers">'
html += ' <input type="submit" value="Create classifier">'
html += '</form>'
return self.output_html(html, 201)
def post(self, id):
parser = reqparse.RequestParser()
parser.add_argument('file',
type=FileStorage,
required=True,
location='files')
args = parser.parse_args()
args['storage_id'] = generate_string()
args['storage_path'] = os.path.join(current_app.root_path,
self.config()['UPLOAD_DIR'],
args['storage_id'])
args['file'].save(args['storage_path'])
args['name'] = args['file'].filename
args['extension'] = '.'.join(args['name'].split('.')[1:])
args['content_type'] = 'application/octet-stream'
args['media_link'] = args['storage_path']
args['size'] = 0
del args['file']
f_dao = FileDao(self.db_session())
f = f_dao.create(**args)
session_dao = SessionDao(self.db_session())
session = session_dao.retrieve(id=id)
print('Running prediction...')
# Load classifier object file and run the prediction
# Run the prediction.
# Delete CSV file
html = '<h3>Congratulations!</h3>'
html += '<p>You have successfully completed your prediction!</p>'
html += '<p>Here are the results:</p>'
return self.output_html(html, 201)
def post(self, id):
parser = reqparse.RequestParser()
parser.add_argument('file',
type=FileStorage,
required=True,
location='files')
parser.add_argument('R', type=str, location='form')
parser.add_argument('target_column', type=str, location='form')
parser.add_argument('exclude_columns', type=str, location='form')
parser.add_argument('nr_iters', type=str, location='form')
parser.add_argument('nr_folds', type=str, location='form')
args = parser.parse_args()
R = args['R']
target_column = args['target_column']
exclude_columns = args['exclude_columns'].split(',')
nr_iters = int(args['nr_iters'])
nr_folds = int(args['nr_folds'])
args['storage_id'] = generate_string()
args['storage_path'] = os.path.join(current_app.root_path,
self.config()['UPLOAD_DIR'],
args['storage_id'])
args['file'].save(args['storage_path'])
args['name'] = args['file'].filename
args['extension'] = '.'.join(args['name'].split('.')[1:])
args['content_type'] = 'application/octet-stream'
args['media_link'] = args['storage_path']
args['size'] = 0
del args['file']
del args['R']
del args['target_column']
del args['exclude_columns']
del args['nr_iters']
del args['nr_folds']
f_dao = FileDao(self.db_session())
f = f_dao.create(**args)
classifier_dao = ClassifierDao(self.db_session())
classifier = classifier_dao.retrieve(id=id)
print('Training classifier {} on file {}'.format(
classifier.name, f.name))
if R == 'true':
print('R scripting is not implemented yet...')
# After classifier training finishes, create a session that captures the results
print('Calculating classifier performance...')
scores = 0
features = pd.read_csv(f.storage_path)
x, y = get_xy(features,
target_column=target_column,
exclude_columns=exclude_columns)
for i in range(nr_iters):
for train, test in StratifiedKFold(n_splits=nr_folds).split(x, y):
_, score = score_svm(x, y, train, test)
scores += score
avg_score = scores / (nr_iters * nr_folds)
path = f.storage_path + '.classifier'
print('Building optimized classifier and storing in {}'.format(path))
classifier_model = train_svm(x, y)
joblib.dump(classifier_model, path)
session_dao = SessionDao(self.db_session())
session = session_dao.create(
**{
'classifier': classifier,
'training_file_path': f.storage_path,
'classifier_file_path': path,
})
html = '<h3>Congratulations!</h3>'
html += '<p>You have successfully trained your classifier! It has an average '
html += 'classification accuracy of {} after {} iterations and {} folds.</p>'.format(
avg_score, nr_iters, nr_folds)
html += '<p>The next step is to use your classifier for predictions. Again, upload a<br>'
html += 'CSV file with the cases you want to predict.</p>'
html += '<form method="post" enctype="multipart/form-data" action="/sessions/{}/predictions">'.format(
session.id)
html += ' <input type="file" name="file">'
html += ' <input type="submit" value="Upload predictions">'
html += '</form>'
return self.output_html(html, 201)
def post(self, id):
"""
Creates a new training session for classifier {id}. The classifier uses the
uploaded file for training.
:param id:
:return:
"""
parser = reqparse.RequestParser()
parser.add_argument('file', type=FileStorage, required=True, location='files')
parser.add_argument('R', type=str, location='form')
parser.add_argument('target_column', type=str, location='form')
parser.add_argument('exclude_columns', type=str, location='form')
parser.add_argument('nr_iters', type=str, location='form')
parser.add_argument('nr_folds', type=str, location='form')
parser.add_argument('subject_id', type=str, required=True, location='form')
args = parser.parse_args()
subject_id = args['subject_id']
R = args['R']
target_column = args['target_column']
exclude_columns = args['exclude_columns'].split(',')
nr_iters = int(args['nr_iters'])
nr_folds = int(args['nr_folds'])
args['storage_id'] = generate_string()
args['storage_path'] = os.path.join(current_app.root_path, self.config()['UPLOAD_DIR'], args['storage_id'])
args['file'].save(args['storage_path'])
args['name'] = args['file'].filename
args['extension'] = '.'.join(args['name'].split('.')[1:])
args['content_type'] = 'application/octet-stream'
args['media_link'] = args['storage_path']
args['size'] = 0
del args['subject_id']
del args['file']
del args['R']
del args['target_column']
del args['exclude_columns']
del args['nr_iters']
del args['nr_folds']
f_dao = FileDao(self.db_session())
f = f_dao.create(**args)
classifier_dao = ClassifierDao(self.db_session())
classifier = classifier_dao.retrieve(id=id)
print('Training classifier {} on file {}'.format(classifier.name, f.name))
if R == 'true':
os.system('Rscript ./R/svm.R 1')
# After classifier training finishes, create a session that captures the results
print('Calculating classifier performance...')
features = pd.read_csv(f.storage_path, index_col=subject_id)
x, y = get_xy(features, target_column=target_column)
scores = 0
# Run performance evaluation on classifier
for i in range(nr_iters):
for train, test in StratifiedKFold(n_splits=nr_folds).split(x, y):
_, score = score_svm(x, y, train, test)
scores += score
avg_score = scores / (nr_iters * nr_folds)
path = f.storage_path + '.classifier'
print('Building optimized classifier and storing in {}'.format(path))
classifier_model = train_svm(x, y)
joblib.dump(classifier_model, path)
session_dao = SessionDao(self.db_session())
session = session_dao.create(**{
'classifier': classifier,
'training_file_path': f.storage_path,
'classifier_file_path': path,
})
html = '<h3>Step 3 - View training session</h3>'
html += '<p>You have successfully trained your classifier. Click the link below<br>'
html += 'To view the training session and upload a file with cases to predict.</p>'
html += '<a href="/sessions/{}">Session {}</a>'.format(session.id, session.id)
return self.output_html(html, 201)