def kmeans(task):
connection.close()
start = datetime.datetime.now()
# import the file to the local SequenceDb
sequences = Sequence.objects.all()
profiles = []
seqOut = []
for sequence in sequences:
profiles.append(np.array(sequence.dna))
seqOut.append(sequence)
results, error, nfound = kcluster(
profiles,
nclusters = task.params['ncluster'],
mask=None,
weight=None,
transpose=0,
npass=task.params['npass'],
method=task.params['method'],
dist=task.params['distance'],
initialid=None)
# save the results
connection.close()
stop = datetime.datetime.now()
task.running = False
task.duration = stop - start
task.save()
service = ResultService()
service.buildResult(task, results, seqOut)
def dbscan(task):
connection.close()
start = datetime.datetime.now()
# import the file to the local SequenceDb
sequences = Sequence.objects.all()
profiles = []
seqOut = []
for sequence in sequences:
profiles.append(np.array(sequence.dna))
seqOut.append(sequence)
algorithm = DBSCAN(eps=task.params['eps'], min_samples=task.params['min_samples'], metric=task.params['metric'])
results = algorithm.fit(profiles)
"""
rewrite the -1 labeled sequences to new clusters with size 1
to make it compairable with kmeans results
"""
# get the last cluster index
idx = np.amax(results.labels_) + 1
labels = []
for label in results.labels_:
if label == -1:
labels.append(idx)
idx += 1
else:
labels.append(label)
labels = np.array(labels)
# save the results
connection.close()
stop = datetime.datetime.now()
task.duration = stop - start
task.save()
service = ResultService()
service.buildResult(task, labels, seqOut)
