for c in range(len(overall_list_title)):
overall_sheet.write(0, c, str(overall_list_title[c]))
dataset_list_title = ['activities'] + per_class_performance_index
# Go through all bosch datasets
datasets = ['b1']
for datafile in datasets:
feature_filename = 'feature_' + datafile + '.pkl'
# Looking for processed feature data
if os.path.exists(feature_filename):
feature_file = open(feature_filename, mode='r')
feature_dict = pickle.load(feature_file)
feature = AlFeature()
feature.load_from_dict(feature_dict)
else:
feature = load_casas_from_file(datafile,
normalize=True,
per_sensor=True,
ignore_other=False)
feature_file = open(feature_filename, mode='w')
pickle.dump(feature.export_to_dict(), feature_file, protocol=-1)
feature_file.close()
num_samples = feature.x.shape[0]
train_index = []
test_index = []
x_tensor = theano.shared(np.asarray(feature.x,
dtype=theano.config.floatX),
borrow=True)
y_tensor = T.cast(theano.shared(feature.y, borrow=True), 'int32')
week_array = get_boundary(feature, period='week')
# Number of perceptrons in hidden layer
hidden_layer_list = [[200, 200, 200]]
for hidden_layer in hidden_layer_list:
assert (type(model) == StackedDenoisingAutoencoder)
x_tensor = theano.shared(np.asarray(x, dtype=theano.config.floatX),
borrow=True)
result = model.classify(x_tensor)
predicted_y = result[0]
confusion_matrix = get_confusion_matrix(num_classes=num_classes,
label=y,
predicted=predicted_y)
return confusion_matrix
if __name__ == '__main__':
# Set current directory to local directory
os.chdir(os.path.dirname(os.path.realpath(__file__)))
# Go through all bosch datasets
datasets = ['b1']
for datafile in datasets:
feature_filename = 'feature_' + datafile + '.pkl'
# Looking for processed feature data
if os.path.exists(feature_filename):
feature_file = open(feature_filename, mode='r')
feature_dict = pickle.load(feature_file)
feature = AlFeature()
feature.load_from_dict(feature_dict)
else:
feature = load_casas_from_file(datafile, datafile + '.translate')
feature_file = open(feature_filename, mode='w')
pickle.dump(feature.export_to_dict(), feature_file, protocol=-1)
feature_file.close()
run_test(feature)
for c in range(len(overall_list_title)):
overall_sheet.write(0, c, str(overall_list_title[c]))
dataset_list_title = ['activities'] + performance_index
# Go through all bosch datasets
datasets = ['b1', 'b2', 'b3']
for datafile in datasets:
feature_filename = 'feature_' + datafile + '.pkl'
# Looking for processed feature data
if os.path.exists(feature_filename):
feature_file = open(feature_filename, mode='r')
feature_dict = pickle.load(feature_file)
feature = AlFeature()
feature.load_from_dict(feature_dict)
else:
feature = load_casas_from_file(datafile,
datafile + '.translate',
normalize=False,
per_sensor=False)
feature_file = open(feature_filename, mode='w')
pickle.dump(feature.export_to_dict(), feature_file, protocol=-1)
feature_file.close()
num_samples = feature.x.shape[0]
train_index = []
test_index = []
# for j in range(num_samples):
# if j % 3 == 0:
# test_index.append(j)
# else:
# train_index.append(j)
num_test = num_samples / 3
test_index = range(num_samples - num_test, num_samples)
train_index = range(num_samples - num_test)
from actlearn.utils.event_bar_plot import event_bar_plot
if __name__ == '__main__':
# Set current directory to local directory
os.chdir(os.path.dirname(os.path.realpath(__file__)))
# Go through all bosch datasets
datasets = ['b1']
for datafile in datasets:
feature_filename = 'feature_' + datafile + '.pkl'
# Looking for processed feature data
if os.path.exists(feature_filename):
feature_file = open(feature_filename, mode='r')
feature_dict = pickle.load(feature_file)
feature = AlFeature()
feature.load_from_dict(feature_dict)
else:
feature = load_casas_from_file(datafile,
datafile + '.translate',
dataset_dir='../../datasets/bosch/')
feature_file = open(feature_filename, mode='w')
pickle.dump(feature.export_to_dict(), feature_file, protocol=-1)
feature_file.close()
# feature.save_data_as_xls('tmp.xls', 0)
# event_bar_plot(feature.time[0:10000], feature.y[0:10000], feature.num_enabled_activities,
# classified=feature.y[1:10001], ignore_activity=feature.activity_list['Other_Activity']['index'])
event_bar_plot(
feature.time[0:100000],
feature.y[0:100000],
feature.num_enabled_activities,
ignore_activity=feature.activity_list['Other_Activity']['index'])
overall_list_row = 0
for c in range(len(overall_list_title)):
overall_sheet.write(0, c, str(overall_list_title[c]))
dataset_list_title = ['activities'] + per_class_performance_index
# Go through all bosch datasets
datasets = ['b1']
for datafile in datasets:
feature_filename = 'feature_' + datafile + '.pkl'
# Looking for processed feature data
if os.path.exists(feature_filename):
feature_file = open(feature_filename, mode='r')
feature_dict = pickle.load(feature_file)
feature = AlFeature()
feature.load_from_dict(feature_dict)
else:
feature = load_casas_from_file(datafile, normalize=False, per_sensor=False)
feature_file = open(feature_filename, mode='w')
pickle.dump(feature.export_to_dict(), feature_file, protocol=-1)
feature_file.close()
num_samples = feature.x.shape[0]
train_index = []
test_index = []
week_array = get_boundary(feature, period='week')
learning_result_fname = 'dt_learning_' + datafile + '.pkl'
learning_result = AlResult(result_name='%s decision tree' % datafile, data_fname=datafile, mode='by_week')
if os.path.exists(learning_result_fname):
learning_result.load_from_file(learning_result_fname)
for week_id in range(len(week_array) - 1):
train_index = range(0, week_array[week_id])
test_index = range(week_array[week_id], week_array[week_id + 1])
decision_tree = DecisionTree(feature.x.shape[1], feature.num_enabled_activities)