def __init__(self, time, categories, dimensions, values, data):
try:
global conn
conn = psycopg2.connect(dbname='postgres',
user='******',
host='localhost',
password='******')
except psycopg2.DatabaseError as ex:
print(ex)
sys.exit(1)
try:
global engine
engine = create_engine(
'postgresql://*****:*****@localhost:5432/postgres',
echo=True)
except Exception as ex:
print(ex)
sys.exit(1)
self.cursor = conn.cursor()
self.data = data
self.categories = categories
#for testing
# self.values = values
# self.dimensions = dimensions+time
#for testing
#org begin
global conn
reduced_dimensions, reduced_values = Clustering.Cluster(
dimensions, values, self.data, conn)
print(reduced_dimensions)
print(reduced_values)
self.dimensions = reduced_dimensions + time
self.values = reduced_values
#org end
PatternStore.create_table_object(self.data)
self.formDatacube()
def kmeans(examples: list, numClusters: int, verbose: bool = False) -> list:
"""
Generate numClusters Clusters using greedy algorithm.
Args
----
examples (list): List of sample points
numClusters (int): Number of clusters to generate
verbose (bool, optional): Print statements. Defaults to False.
Returns
-------
list: Current set of defined clusters.
Algorithm
---------
Randomly choose numExamples examples as initial centroids
while true
- Assign each member of passed in examples to the cluster whose
centroid is closest
- Compute updated clusters centroid with the addition of new members
- If new cluster centroids are not different then they were in the
previous iteration of the while loop
- return the current set of clusters
"""
# Generate numClusters centroids and add them to separate cluster list
# elements
initialCentroids = random.sample(examples, numClusters)
clusters = []
for centroid in initialCentroids:
clusters.append(Clustering.Cluster([centroid]))
#
# Iterate until centroids do not change (converge)
#
converged = False
numIterations = 0
while not converged:
numIterations += 1
# Create a list containing numClusters distinct empty lists
newClusters = []
for i in range(numClusters):
newClusters.append([])
#
# Associate each example with the cluster with the closest centroid
#
for example in examples:
# Find the centroid among the centroids of the cluster list
# that is closest to the example starting with first centroid
# group and compare other centroids
currentSmallestDistance = \
example.distance(clusters[0].getCentroid())
index = 0
for i in range(1, numClusters):
# Comment out one of the two folllowing lines
# if power paramter is 1, calculate Manhattan distance
# if power paramter is 2, calculate Euclidean distance
# power = 1
power = 2
distance = example.distance(clusters[i].getCentroid(), power)
if distance < currentSmallestDistance:
currentSmallestDistance = distance
index = i
# Add example element to the cluster list of examples, where the
# distance between the element and the cluster list's
# centroid is minimum.
newClusters[index].append(example)
for cluster in newClusters: # Check to see if any empty clusters
if len(cluster) == 0:
raise ValueError('Empty cluster')
# Update each cluster; determine if cluster centroid has changed
converged = True
for i in range(numClusters):
if clusters[i].update(newClusters[i]) > 0.0:
converged = False
if verbose:
print('Iteration num:', numIterations)
for cluster in clusters:
print(cluster)
print("\n")
return clusters
