def clustering_oka(nec_set, k=25):
"""
Group record according to NCP. OKA: one time pass k-means
"""
can_clusters = [cluster for cluster in nec_set if len(cluster) >= k]
nec_set = [cluster for cluster in nec_set if len(cluster) < k]
remain = sum([len(t) for t in nec_set])
clusters = []
# randomly choose seed and find k-1 nearest records to form cluster with size k
seed_index = random.sample(range(len(nec_set)), remain / k)
for index in seed_index:
can_clusters.append(nec_set[index])
nec_set = [t for i, t in enumerate(nec_set[:]) if i not in set(seed_index)]
while len(nec_set) > 0:
nec = nec_set.pop()
index = find_best_cluster_iloss(nec, can_clusters)
can_clusters[index].merge_cluster(nec)
residual = []
less_clusters = []
for cluster in can_clusters:
if len(cluster) < k:
less_clusters.append(cluster)
else:
if len(cluster) > k:
adjust_cluster(cluster, residual, k)
clusters.append(cluster)
while len(residual) > 0:
record = residual.pop()
record_key = qid_to_key(record[:QI_LEN])
if len(less_clusters) > 0:
index = find_best_cluster_iloss(record, less_clusters)
less_clusters[index].add_record(record)
residual_handle(residual, record_key, less_clusters[index])
if len(less_clusters[index]) >= k:
clusters.append(less_clusters.pop(index))
else:
index = find_best_cluster_iloss(record, clusters)
clusters[index].add_record(record)
residual_handle(residual, record_key, clusters[index])
# sometimes residual records cannot satisfy less_clusters
# so we need to handle these clusters
if len(less_clusters) > 0:
for cluster in less_clusters:
residual.extend(cluster.member)
while len(residual) > 0:
record = residual.pop()
record_key = qid_to_key(record[:QI_LEN])
index = find_best_cluster_iloss(record, clusters)
clusters[index].add_record(record)
residual_handle(residual, record_key, clusters[index])
return clusters
def clustering_oka(nec_set, k=25):
"""
Group record according to NCP. OKA: one time pass k-means
"""
can_clusters = [cluster for cluster in nec_set if len(cluster) >= k]
nec_set = [cluster for cluster in nec_set if len(cluster) < k]
remain = sum([len(t) for t in nec_set])
clusters = []
# randomly choose seed and find k-1 nearest records to form cluster with size k
seed_index = random.sample(range(len(nec_set)), remain / k)
for index in seed_index:
can_clusters.append(nec_set[index])
nec_set = [t for i, t in enumerate(nec_set[:]) if i not in set(seed_index)]
while len(nec_set) > 0:
nec = nec_set.pop()
index = find_best_cluster_iloss(nec, can_clusters)
can_clusters[index].merge_cluster(nec)
residual = []
less_clusters = []
for cluster in can_clusters:
if len(cluster) < k:
less_clusters.append(cluster)
else:
if len(cluster) > k:
adjust_cluster(cluster, residual, k)
clusters.append(cluster)
while len(residual) > 0:
record = residual.pop()
record_key = qid_to_key(record[:QI_LEN])
if len(less_clusters) > 0:
index = find_best_cluster_iloss(record, less_clusters)
less_clusters[index].add_record(record)
residual_handle(residual, record_key, less_clusters[index])
if len(less_clusters[index]) >= k:
clusters.append(less_clusters.pop(index))
else:
index = find_best_cluster_iloss(record, clusters)
clusters[index].add_record(record)
residual_handle(residual, record_key, clusters[index])
# sometimes residual records cannot satisfy less_clusters
# so we need to handle these clusters
if len(less_clusters) > 0:
for cluster in less_clusters:
residual.extend(cluster.member)
while len(residual) > 0:
record = residual.pop()
record_key = qid_to_key(record[:QI_LEN])
index = find_best_cluster_iloss(record, clusters)
clusters[index].add_record(record)
residual_handle(residual, record_key, clusters[index])
return clusters
def NCP(record):
"""Compute NCP (Normalized Certainty Penalty)
when generate record to gen_result.
"""
ncp = 0.0
# exclude SA values(last one type [])
list_key = qid_to_key(record)
try:
return NCP_CACHE[list_key]
except KeyError:
pass
for i in range(QI_LEN):
# if leaf_num of numerator is 1, then NCP is 0
width = 0.0
if IS_CAT[i] is False:
try:
float(record[i])
except ValueError:
temp = record[i].split(',')
width = float(temp[1]) - float(temp[0])
else:
width = len(ATT_TREES[i][record[i]]) * 1.0
width /= QI_RANGE[i]
ncp += width
NCP_CACHE[list_key] = ncp
return ncp
def NCP(record):
"""Compute NCP (Normalized Certainty Penalty)
when generate record to gen_result.
"""
ncp = 0.0
# exclude SA values(last one type [])
list_key = qid_to_key(record)
try:
return NCP_CACHE[list_key]
except KeyError:
pass
for i in range(QI_LEN):
# if leaf_num of numerator is 1, then NCP is 0
width = 0.0
if IS_CAT[i] is False:
try:
float(record[i])
except ValueError:
temp = record[i].split(',')
width = float(temp[1]) - float(temp[0])
else:
width = len(ATT_TREES[i][record[i]]) * 1.0
width /= QI_RANGE[i]
ncp += width
NCP_CACHE[list_key] = ncp
return ncp
def residual_handle(residual, record_key, cluster):
while True:
try:
same_record = residual[-1]
except IndexError:
break
if record_key == qid_to_key(same_record[:QI_LEN]):
cluster.add_record(residual.pop(-1))
else:
break
def residual_handle(residual, record_key, cluster):
while True:
try:
same_record = residual[-1]
except IndexError:
break
if record_key == qid_to_key(same_record[:QI_LEN]):
cluster.add_record(residual.pop(-1))
else:
break
def create_nec(data):
"""
create NEC from dateset using dict
:param data: dataset
:return: NEC in dict format: key is str, value is Cluster
"""
nec_dict = dict()
for record in data:
key = qid_to_key(record[:QI_LEN])
try:
nec_dict[key].add_same_record(record)
except KeyError:
nec_dict[key] = Cluster([record], record)
return nec_dict
def create_nec(data):
"""
create NEC from dateset using dict
:param data: dataset
:return: NEC in dict format: key is str, value is Cluster
"""
nec_dict = dict()
for record in data:
key = qid_to_key(record[:QI_LEN])
try:
nec_dict[key].add_same_record(record)
except KeyError:
nec_dict[key] = Cluster([record], record)
return nec_dict
