class DBscanUnitTest(object):
"""
DBscanUnitTest class provides unit test method for DBscan class
"""
test_array = [[0, 100], [0, 200], [0, 275], [100, 150], [200, 100], [250, 200], [0, 300],[675, 700], [675, 710], [675, 720]]
proper_cluster = [-1, 1, 1, -1, -1, -1, 1, 2, 2, 2]
dbscan = 0
def __init__(self):
self.dbscan = DBscan()
# start_test function start unit test for DBscan class. Return False if failed elsewhere return True
def start_test(self):
point_array = []
for p in self.test_array:
pt = Point(p[0], p[1])
point_array.append(pt)
self.dbscan.start(point_array, 100, 3)
i = 0
for p in point_array:
if p.get_clusterID() is not self.proper_cluster[i]:
print("Clustering algorithm failed")
return False
i += 1
return True
def start_test(self):
db = DBscan()
table = [100, 1000, 3000, 5000, 10000]
for i in range(0, len(table)):
number_of_centers = randint(2, 5)
range1 = randint(1, 10)
scikit_average = 0.0
our_average = 0.0
random_generator = RandomDataGenerator(number_of_samples=table[i],
number_of_centers=number_of_centers, range=range1)
for _ in range(0, 3):
EPS = uniform(0.09, range1/10)
MIN_SAMPLES = randint(5, 10)
#################################################################################
# Generate sample data
#################################################################################
X, labels = random_generator.generate_data()
xx = X.tolist()
point_array = []
for x in xx:
pt = Point(x[0], x[1])
point_array.append(pt)
#################################################################################
# sklearn
#################################################################################
start = timeit.default_timer()
DBSCAN(eps=EPS, min_samples=MIN_SAMPLES).fit(X)
end = timeit.default_timer()
scikit_average += end - start
start =timeit.default_timer()
clusters = db.start(points=point_array, eps=EPS, minPts=MIN_SAMPLES)
end = timeit.default_timer()
our_average += end - start
print("Number of samples: %d" % table[i])
print("Scikit average time of computing = %0.5f" % (scikit_average / 3))
print("Our version average time of computing = %0.5f" % (our_average / 3))
print("Our algorithm is %0.2f slower than algorithm from scikit" % (our_average/scikit_average))
return True
#################################################################################
# Compute DBSCAN
#################################################################################
##xx = X.tolist()
##print( "\nxx= \n", xx)
xx, test_tab = getTextData()
#xx = StandardScaler().fit_transform(xxx).tolist()
#print( xx )
point_array = []
for x in xx:
pt = Point(x[0], x[1])
point_array.append(pt)
# db = DBscan(point_array=parr, start_point_index= 0, cluster_map=clustmap, epsilon=EPS, min_neighbour=MIN_SAMPLES)
db = DBscan()
clusters = db.start(points=point_array, eps=EPS, minPts=MIN_SAMPLES)
#################################################################################
# Plot result
#################################################################################
plt.subplot(212)
colors = plt.cm.Spectral(np.linspace(0, 1, len(clusters)))
for p in point_array:
if p.get_clusterID() == Point.NOISE:
col = 'k'
elif p.get_clusterID == Point.UNCLASSIFIED:
col = 'g'
else:
col = colors[p.get_clusterID() - 1]