def main(*argv):
start_time = datetime.datetime.now()
print("--------------------------------------")
print("STARTTIME:", (datetime.datetime.now()))
print("--------------------------------------")
# set parameters
global CLUSTER_NUM
global MAX_KTH
global GPS_WEIGHT
global FILTER_TIME_S
global FILTER_TIME_E
global OUTPUT_MAP
global OUTPUT_PATTERN
if len(argv) > 0:
CLUSTER_NUM = argv[0]
MAX_KTH = argv[1]
GPS_WEIGHT = argv[2]
FILTER_TIME_S = argv[3]
FILTER_TIME_E = argv[4]
LOCATION_TOPIC = "./data/LocationTopic/LocationTopic_OCT_c35.txt"
else:
LOCATION_TOPIC = "./data/LocationTopic/LocationTopic_NOVDEC_c35.txt"
OUTPUT_MAP = OUTPUT_MAP + str(CLUSTER_NUM) + "k" + str(MAX_KTH) + "w" + str(GPS_WEIGHT)
OUTPUT_PATTERN = OUTPUT_PATTERN + str(CLUSTER_NUM) + "k" + str(MAX_KTH)
# Getting data
users, locations = locationclustering.main(FILTER_TIME_S, FILTER_TIME_E)
location_id, doc_topic = ccluster.open_doc_topic(LOCATION_TOPIC)
locations = ccluster.fit_locations_membership(locations, numpy.transpose(doc_topic), location_id, "semantic_mem")
print(" users # :", len(users))
# Getting sequences of posts & locations
#sequences = ctrajectory.split_trajectory([a_user.posts for a_user in users.values() if len(a_user.posts) != 0], SPLIT_DAY)
sequences = ctrajectory.split_trajectory_byday([a_user.posts for a_user in users.values() if len(a_user.posts) != 0])
sequences = ctrajectory.remove_adjacent_location(sequences)
#sequences = get_specific(sequences)
sequences = ctrajectory.remove_short(sequences)
print(" remain users #:", len(set([x[0].uid for x in sequences])))
location_sequences, longest_len = ctrajectory.convertto_location_sequences(sequences, locations)
spatial_array = ctrajectory.get_vector_array(location_sequences, longest_len)
semantic_array = ctrajectory.get_vector_array(location_sequences, longest_len, "semantic_mem")
u, u0, d, jm, p, fpc, center, membership = cfuzzy.sequences_clustering_i("Location", spatial_array, CLUSTER_NUM, MAX_KTH, semantic_array, GPS_WEIGHT, e = ERROR, algorithm="2WeightedDistance")
#sequences, location_sequences, membership = filter_sequence(sequences, location_sequences, u, membership)
#ouput_pattern(sequences, location_sequences, u, membership)
#output_each_pattern(sequences, location_sequences, u, membership, 30)
#ctrajectory.output_clusters(sequences, membership, u, OUTPUT_PATTERN)
#output_cluster_detail(sequences, location_sequences, u, membership, 39, file = OUTPUT_ANALYSIS)
print("--------------------------------------")
print("ENDTIME:", (datetime.datetime.now()), ", SPEND:", datetime.datetime.now() - start_time)
print("--------------------------------------")
return location_sequences, spatial_array, semantic_array, u
def main():
print("--------------------------------------")
print("STARTTIME:", (datetime.datetime.now()))
print("--------------------------------------")
# Getting data
users, locations = locationclustering.main()
location_id, doc_topic = ccluster.open_doc_topic(LOCATION_TOPIC)
locations = ccluster.fit_locations_membership(locations, numpy.transpose(doc_topic), location_id, "semantic_mem")
semantic_cluster = numpy.argmax(doc_topic, axis=1)
locations = ccluster.fit_locations_cluster(locations, semantic_cluster, location_id, "semantic_cluster")
# Getting sequences cluster
sequences = ctrajectory.split_trajectory(
[a_user.posts for a_user in users.values() if len(a_user.posts) != 0], SPLIT_DAY
)
cluster_sequences = ctrajectory.get_cluster_sequence(sequences, locations)
semantic_sequences = ctrajectory.get_cluster_sequence(sequences, locations, "semantic_cluster")
location_sequences = ctrajectory.convertto_location_sequences(sequences, locations)
print("Filtering short trajectories...")
fail_indices = []
for i, s in enumerate(sequences):
if len(s) <= 2:
fail_indices.append(i)
print(" will delete #:", len(fail_indices))
sequences = numpy.delete(numpy.array(sequences), fail_indices)
cluster_sequences = numpy.delete(numpy.array(cluster_sequences), fail_indices)
semantic_sequences = numpy.delete(numpy.array(semantic_sequences), fail_indices)
location_sequences = numpy.delete(numpy.array(location_sequences), fail_indices)
print(" remain sequences #:", len(sequences))
u, u0, d, jm, p, fpc, membership, distance = cfuzzy.sequences_clustering(
"Cluster", cluster_sequences, CLUSTER_NUM, MAX_KTH, semantic_sequences, e=ERROR, algorithm="2Distance"
)
print("Start Outputting...")
for c in range(CLUSTER_NUM):
this_cluster_indices = [i for i, x in enumerate(membership) if x == c]
print(c, " >> this cluster #:", len(this_cluster_indices))
if len(this_cluster_indices) is not 0:
top_10_u = sorted(u[c, this_cluster_indices], reverse=True)
if len(top_10_u) >= MAX_KTH:
top_10_u = top_10_u[MAX_KTH - 1]
else:
top_10_u = top_10_u[-1]
top_10_indices = [i for i, x in enumerate(u[c, this_cluster_indices]) if x >= top_10_u]
# top_10_indices = sorted(range(len(u[c, this_cluster_indices])), key=lambda x: u[c, this_cluster_indices][x], reverse=True)[0:10]
print(" top_10:", top_10_u, ">", top_10_indices)
print(u[c, this_cluster_indices][top_10_indices])
points_sequences = numpy.array(location_sequences)[this_cluster_indices][top_10_indices]
color = sorted(range(len(points_sequences)), key=lambda x: top_10_indices[x])
cpygmaps.output_patterns_l(
points_sequences, color, len(points_sequences), OUTPUT_MAP + "_" + str(c) + ".html"
)
print("--------------------------------------")
print("ENDTIME:", (datetime.datetime.now()))
print("--------------------------------------")
def main(*argv):
print("--------------------------------------")
print("STARTTIME:", (datetime.datetime.now()))
print("--------------------------------------")
# set UNIXTIME
global FILTER_TIME_S
global FILTER_TIME_E
if len(argv) > 0:
FILTER_TIME_S = argv[0]
FILTER_TIME_E = argv[1]
# getting data
locations = clocation.open_locations()
users = cuser.open_users_posts_afile(USER_POSTS_FILE)
# preporcessing. remove unqualified users
removes = filter_users_timeperiod(users, FILTER_TIME_S, FILTER_TIME_E)
sequences = ctrajectory.split_trajectory_byday([a_user.posts for a_user in users.values() if len(a_user.posts) != 0])
sequences = ctrajectory.remove_adjacent_location(sequences)
sequences = ctrajectory.remove_short(sequences)
removes = list(set(removes) | (set(users.keys()) - set([x[0].uid for x in sequences])))
for key in removes:
del users[key]
print(" remain users #:", len(users.keys()))
locations = clocation.fit_users_to_location(locations, users, "uid")
set_location_user_count(locations)
coordinate = numpy.array([(float(x.lat), float(x.lng)) for x in locations.values()])
location_frequency = numpy.array([x.usercount for x in locations.values()])
# clustering locations
cntr, u, u0, d, jm, p, fpc, membership = cfuzzy.cmeans_location(coordinate.T, CLUSTER_NUM, MAX_KTH, location_frequency, Y, e=ERROR, algorithm="kthCluster_LocationFrequency")
locations = ccluster.fit_locations_membership(locations, u, locations.keys())
locations = ccluster.fit_locations_cluster(locations, membership, locations.keys())
# output result
"""
cpygmaps.output_clusters(\
[(float(x.lat), float(x.lng), str(x.cluster) + " >> " + x.lname + "(" + x.lid + ")>>u:" + str(u[x.cluster, i])) \
for i, x in enumerate(locations.values())], membership, CLUSTER_NUM, OUTPUT_MAP)
output_representatives(coordinate, u, MAX_KTH)
"""
#ccluster.output_location_cluster(locations.values(), "cluster", OUTPUT_CLUSTER)
print("--------------------------------------")
print("ENDTIME:", (datetime.datetime.now()))
print("--------------------------------------")
return users, locations
def main():
print("--------------------------------------")
print("STARTTIME:", (datetime.datetime.now()))
print("--------------------------------------")
# Getting data
users, locations = locationclustering.main()
location_id, doc_topic = ccluster.open_doc_topic(LOCATION_TOPIC)
locations = ccluster.fit_locations_membership(locations,
numpy.transpose(doc_topic),
location_id, "semantic_mem")
# Getting sequences cluster
sequences = ctrajectory.split_trajectory(
[a_user.posts for a_user in users.values() if len(a_user.posts) != 0],
SPLIT_DAY)
vector_sequences = ctrajectory.get_vector_sequence(sequences, locations)
semantic_sequences = ctrajectory.get_vector_sequence(
sequences, locations, "semantic_mem")
location_sequences = ctrajectory.convertto_location_sequences(
sequences, locations)
print("Filtering short trajectories...")
fail_indices = []
for i, s in enumerate(sequences):
if len(s) <= 2:
fail_indices.append(i)
print(" will delete #:", len(fail_indices))
sequences = numpy.delete(numpy.array(sequences), fail_indices)
vector_sequences = numpy.delete(numpy.array(vector_sequences),
fail_indices)
semantic_sequences = numpy.delete(numpy.array(semantic_sequences),
fail_indices)
location_sequences = numpy.delete(numpy.array(location_sequences),
fail_indices)
print(" remain sequences #:", len(sequences))
u, u0, d, jm, p, fpc, membership, distance = cfuzzy.sequences_clustering(
"Location",
vector_sequences,
CLUSTER_NUM,
MAX_KTH,
semantic_sequences,
e=ERROR,
algorithm="2Distance")
print("Start Outputting...")
for c in range(CLUSTER_NUM):
this_cluster_indices = [i for i, x in enumerate(membership) if x == c]
print(c, " >> this cluster #:", len(this_cluster_indices))
if len(this_cluster_indices) is not 0:
top_10_u = sorted(u[c, this_cluster_indices], reverse=True)
if len(top_10_u) >= MAX_KTH:
top_10_u = top_10_u[MAX_KTH - 1]
else:
top_10_u = top_10_u[-1]
top_10_indices = [
i for i, x in enumerate(u[c, this_cluster_indices])
if x >= top_10_u
]
#top_10_indices = sorted(range(len(u[c, this_cluster_indices])), key=lambda x: u[c, this_cluster_indices][x], reverse=True)[0:10]
print(" top_10:", top_10_u, ">", top_10_indices)
print(u[c, this_cluster_indices][top_10_indices])
points_sequences = numpy.array(
location_sequences)[this_cluster_indices][top_10_indices]
color = sorted(range(len(points_sequences)),
key=lambda x: top_10_indices[x])
cpygmaps.output_patterns_l(points_sequences, color,
len(points_sequences),
OUTPUT_MAP + "_" + str(c) + ".html")
print("--------------------------------------")
print("ENDTIME:", (datetime.datetime.now()))
print("--------------------------------------")