def main():
classifier = svm.SVC()
transformed_train_data = [
create_feature_vector(coordinates) for coordinates in train_data
]
classifier.fit(transformed_train_data, train_labels)
test_data, test_labels = apa.read_data_files(train_files)
transformed_test_data = [
create_feature_vector(coordinates) for coordinates in test_data
]
predicted = classifier.predict(transformed_test_data)
conf_mat = np.zeros((10, 10))
success, error = 0, 0
for test_label, predicted_label in zip(test_labels, predicted):
conf_mat[test_label, predicted_label] += 1
if test_label == predicted_label:
success += 1
else:
error += 1
precision = success / float(success + error) * 100
print("precision: %s " % precision)
print_conf_mat(conf_mat)
def main():
classifier = svm.SVC()
transformed_train_data = [create_feature_vector(coordinates) for coordinates in train_data]
classifier.fit(transformed_train_data, train_labels)
test_data, test_labels = apa.read_data_files(train_files)
transformed_test_data = [create_feature_vector(coordinates) for coordinates in test_data]
predicted = classifier.predict(transformed_test_data)
conf_mat = np.zeros((10, 10))
success, error = 0, 0
for test_label, predicted_label in zip(test_labels, predicted):
conf_mat[test_label, predicted_label] += 1
if test_label == predicted_label:
success += 1
else:
error += 1
precision = success / float(success + error) * 100
print("precision: %s " % precision)
print_conf_mat(conf_mat)
test_nums = ['15', '19', '20']
apa_train_dirs = ["%s/apa%s" % (apa_data_path, i) for i in train_nums]
apa_test_dirs = ["%s/apa%s" % (apa_data_path, i) for i in test_nums]
def get_apa_files(apa_dirs):
files = []
for apa_dir in apa_dirs:
files += map(lambda s: os.path.join(apa_dir, s), os.listdir(apa_dir))
return files
train_files = get_apa_files(apa_train_dirs)
test_files = get_apa_files(apa_test_dirs)
train_data, train_labels = apa.read_data_files(train_files)
def create_feature_vector(coordinates):
X, Y = [], []
for x, y in coordinates:
X.append(x)
Y.append(y)
total_distance = 0
for pair in zip(coordinates[0:len(coordinates)-2], coordinates[1::len(coordinates)-1]):
a, b = pair
distance = np.linalg.norm(a-b)
total_distance += distance
var_x = np.var(X)
apa_train_dirs = ["%s/apa%s" % (apa_data_path, i) for i in train_nums]
apa_test_dirs = ["%s/apa%s" % (apa_data_path, i) for i in test_nums]
def get_apa_files(apa_dirs):
files = []
for apa_dir in apa_dirs:
files += map(lambda s: os.path.join(apa_dir, s), os.listdir(apa_dir))
return files
train_files = get_apa_files(apa_train_dirs)
test_files = get_apa_files(apa_test_dirs)
train_data, train_labels = apa.read_data_files(train_files)
def create_feature_vector(coordinates):
X, Y = [], []
for x, y in coordinates:
X.append(x)
Y.append(y)
total_distance = 0
for pair in zip(coordinates[0:len(coordinates) - 2],
coordinates[1::len(coordinates) - 1]):
a, b = pair
distance = np.linalg.norm(a - b)
total_distance += distance