def process_xval():
"""Cross validation on training data"""
mysql_db.connect()
try:
logger.debug("Getting training data")
features, labels = get_training()
logger.debug("Retrieved training data")
except:
mysql_db.close()
raise
mysql_db.close()
logger.debug("Calculating k-fold indices")
skf = StratifiedKFold(labels, n_folds=N_FOLD)
conf_mtx = None
logger.debug("Starting cross validation")
for train_idx, test_idx in skf:
X_train, X_test = features[train_idx], features[test_idx]
y_train, y_test = labels[train_idx], labels[test_idx]
logger.debug("scaling")
scaler = StandardScaler(copy=False)
scaler.fit_transform(X_train)
scaler.transform(X_test)
logger.debug("Running classifier")
predictions = Classify.runClassifiers(X_train, y_train, X_test)
if conf_mtx == None:
conf_mtx = confusion_matrix(y_test, predictions)
else:
conf_mtx += confusion_matrix(y_test, predictions)
print classification_report(y_test, predictions, digits=2)
print conf_mtx
def main():
"""Main"""
description = "Extract features and save to Nialls's db"
parser = argparse.ArgumentParser(description=description)
setup_arguments(parser)
setup_options(parser)
args = parser.parse_args()
error = validate_arguments(args)
if error != None:
parser.error(error)
ch = logging.StreamHandler()
if args.debug:
logger.setLevel(logging.DEBUG)
ch.setLevel(logging.DEBUG)
else:
logger.setLevel(logging.WARN)
ch.setLevel(logging.WARN)
logger.addHandler(ch)
mysql_db.connect()
try:
features_to_db(mysql_db, args.training_dir, args.test_dir,
args.label_file)
except:
mysql_db.close()
raise
mysql_db.close()
def process_normal(output_file):
"""Process test data instead of crossval data"""
mysql_db.connect()
try:
logger.debug("Getting training data")
train_features, train_labels = get_training()
logger.debug("Getting test data")
test_features, test_names = get_test()
except:
mysql_db.close()
raise
mysql_db.close()
logging.debug("scaling")
scaler = StandardScaler(copy=False)
scaler.fit_transform(train_features)
scaler.transform(test_features)
logger.debug("running classifier")
predictions = Classify.runClassifiers(train_features, train_labels,
test_features)
logger.debug("writing predictions")
write_predictions(output_file, test_names, predictions)
from feature import Feature, mysql_db
import numpy as np
import sys
#connect to Nialls's through ssh tunnel
#At start, connect (MAKE SURE YOU HAVE UPDATED feature.py WITH DB
# SETTINGS)
mysql_db.connect()
#Get training
query = Feature.select().where(Feature.label.is_null(False))
gray_hists = [] # See feature.py for fields in table
red_hists = []
labels = []
# naive().iterator() is nicer for large numbers of rows (which we have)
for feature in query.naive().iterator():
gray_hists.append(feature.gray_hist)
red_hists.append(feature.red_hist)
labels.append(feature.label)
gray_train = np.vstack(gray_hists)
red_train = np.vstack(red_hists)
gray_and_red_train = np.hstack([gray_train, red_train])
#Fit using features
# model.fit(gray_train, np.array(labels))
#Get test
query = Feature.select().where(Feature.label.is_null(True)).order_by(