def main():
print 'now entered main'
parser = argparse.ArgumentParser()
parser.add_argument('-i', '-I', help='input CSV file', required=True)
parser.add_argument('-o', '-O', help='output path', required=True)
parser.add_argument('-c', '-C', help='cluster data path', required=False)
args = parser.parse_args()
input_file = args.i
output_path = args.o
cluster_data_path = args.c
cluster_service = gps_service()
print 'arguments assigned variables'
data = pr.getAllData(input_file)
print 'keeping only app init and live listening'
data_to_use = pr.filtersurveydata(data, SurveyConstants.CONDITION_ID, ['1', '2', '3', '4'])
cluster_data_files = glob(cluster_data_path + '*.data')
app_init_data = pr.filtersurveydata(data_to_use, SurveyConstants.USER_INITIATED, ['false'])
listening_data = pr.filtersurveydata(data_to_use, SurveyConstants.LISTENING, ['true'])
data = app_init_data + listening_data
print 'done'
per_participant_data = pr.getPerParticipantData(data)
print 'per participant data extracted'
participant_list = per_participant_data.keys()
print participant_list
min_data_sample_no = 5
final_result = {}
cluster_results = {}
conf_mat_results = {}
for pid in participant_list:
print '\n\npid: ' + pid
if len(per_participant_data[pid]) < min_data_sample_no:
print '# of samples < min_data_sample_no (' + str(min_data_sample_no) + '), skipping pid'
continue
if cluster_data_path + pid + '_all_data.data' not in cluster_data_files:
print 'could not find data file for pid: ', pid, ', skipping'
continue
final_result[pid] = {}
cluster_results[pid] = {}
cluster_service.clean_house()
cluster_service.set_pid(pid)
cluster_service.set_participant_data(per_participant_data[pid])
with open(cluster_data_path+pid+'_all_data.data', 'rb') as f:
data_dict = pickle.load(f)
cluster_boundaries = data_dict['boundary']
cluster_labels = data_dict['label']
missing_gps = 0
for data_sample in per_participant_data[pid]:
n_pid = data_sample[SurveyConstants.PATIENT_ID]
cid = data_sample[SurveyConstants.CONDITION_ID]
sid = data_sample[SurveyConstants.SESSION_ID]
if '' == data_sample[SurveyConstants.GPS_PATH]:
#print 'empty gps file path, skipping \n', data_sample
missing_gps += 1
continue
gps_coords_clean = pr.getcleangpsdata(data_sample[SurveyConstants.GPS_PATH], remove_duplicates=True,
pid=n_pid, cid=cid, sid=sid)
if gps_coords_clean is None:
print 'no GPS data for ', n_pid, cid, sid, ', skipping'
continue
travel_result = cluster_service.find_travelling(gps_coords_clean)
final_result[pid][(n_pid, cid, sid)] = \
[(LocationContext.LOCATION_CONTEXT_VALUES[data_sample[SurveyConstants.LOCATION_CONTEXT]],
len(gps_coords_clean))]
cluster_results[pid][(n_pid, cid, sid)] = [
(LocationContext.LOCATION_CONTEXT_VALUES[data_sample[SurveyConstants.LOCATION_CONTEXT]],
gps_coords_clean)]
if travel_result[0]:
if not 0 == len(travel_result[1]):
final_result[pid][(n_pid, cid, sid)].append(('Travel', len(travel_result[1])))
cluster_results[pid][(n_pid, cid, sid)].append(('Travel', (travel_result[1])))
if not 0 == len(travel_result[2]):
cluster_decisions = gps.check_polygon_memberships(cluster_boundaries, travel_result[2])
cluster_vals = [sum(x) for x in cluster_decisions]
for idx in range(len(cluster_vals)):
if not 0 == cluster_vals[idx]:
final_result[pid][(n_pid, cid, sid)].append((cluster_labels[idx], cluster_vals[idx]))
cluster_results[pid][(n_pid, cid, sid)].append((cluster_labels[idx], travel_result[2],
cluster_decisions[idx]))
print 'confusion matrix: '
conf_mat = {}
for main_tuple in final_result[pid]:
count_data = final_result[pid][main_tuple]
actual_label = count_data[0][0]
for count_idx in range(1, len(count_data)):
predicted_label = count_data[count_idx][0]
if (actual_label, predicted_label) not in conf_mat:
conf_mat[(actual_label, predicted_label)] = 0
conf_mat[(actual_label, predicted_label)] += count_data[count_idx][1]
print conf_mat
conf_mat_results[pid] = conf_mat
print 'Missing GPS: ', missing_gps
print 'Writing results to file'
bD.write_variable(final_result, 'count_result.res', output_path)
bD.write_variable(cluster_results, 'cluster_results.res', output_path)
bD.write_variable(conf_mat_results, 'conf_mat_results.res', output_path)
print 'done'
print 'TADAA!!'
def get_travelling_and_stationary_clusters(self, eps_list=range(20, 51, 10), min_sample_list=[3, 5, 7]):
'''
for each data point within the participant data, distinguish between the travelling, and non-travelling data.
Once all the travelling clusters, and non-travelling points have been extracted perform the DBSCAN clustering
on the non-travelling points to obtain the stationary clusters, and noise markers.
:return:
'''
for data_sample in self.__participant_data:
try:
self.__internal_location_info[(data_sample[SurveyConstants.PATIENT_ID],
data_sample[SurveyConstants.CONDITION_ID],
data_sample[SurveyConstants.SESSION_ID])] = \
LocationContext.LOCATION_CONTEXT_VALUES[data_sample[SurveyConstants.LOCATION_CONTEXT]]
gps_coords_clean = pr.getcleangpsdata(data_sample[34], remove_duplicates=True,
pid=data_sample[0], cid=data_sample[1], sid=data_sample[2])
if not gps_coords_clean:
continue
except IOError:
self.__error_files += 1
continue
# TODO: the speed limit has to be decided, are people walking also considered travelling?
travel_result = self.find_travelling(gps_coords_clean)
if travel_result[0]:
# travel_clusters.append(gps_coords_clean)
if not 0 == len(travel_result[1]):
self.__travel_clusters.append(travel_result[1])
if not 0 == len(travel_result[2]):
self.__stationary_points += travel_result[2]
else:
self.__stationary_points += gps_coords_clean
'''
since all the stationary points are being collected for a given participants, the hull intersection functions
never get called.
'''
print 'collected all points, clustering, eps_list:', eps_list, ', min_sample_list:', min_sample_list
sc_nz = clusters.getdbscanclusters(self.__stationary_points, eps_list, min_sample_list)
print 'done'
if sc_nz is not None:
if not ([] == sc_nz['sc']):
self.__stationary_clusters = sc_nz['sc']
if not ([] == sc_nz['nz']):
self.__noise_markers = sc_nz['nz']
print 'stationary clusters: ' + str(len(self.__stationary_clusters)) + ', travel clusters: ' + str(
len(self.__travel_clusters))
x = 0
for cluster_points in self.__stationary_clusters:
x += 1
try:
boundary_points = gps.getconvexhull(cluster_points)
except:
print 'Error getting the convex hull of the cluster. cluster #', x
self.__stationary_cluster_label.append("Error,C-"+str(x))
continue
self.__stationary_cluster_boundaries.append(boundary_points)
cluster_point_types = []
for cluster_point in cluster_points:
cluster_point_types.append(self.__internal_location_info[(cluster_point[-3],
cluster_point[-2],
cluster_point[-1])])
label_counts = collections.Counter(cluster_point_types)
most_common_label = label_counts.most_common(1)
self.__stationary_cluster_label.append(most_common_label[0][0])
return self.__travel_clusters, self.__stationary_clusters, self.__stationary_cluster_boundaries, \
self.__stationary_cluster_label, self.__noise_markers, self.__error_files, self.__stationary_points