def get_parameters(self):
"""
This function creates some parameters to be used with DBScan.
@return: A tuple with the generated parameters and an empty list corresponding to the clusterings.
"""
run_parameters = []
# (minpts, eps tuples)
if "max" in self.parameters["clustering"]["algorithms"]["dbscan"]:
max_eps_tries = self.parameters["clustering"]["algorithms"]["dbscan"]["max"]
else:
max_eps_tries = 10
num_elements = self.distance_matrix.row_length
klist = k_scale_gen(math.log(num_elements))
buffer = numpy.empty(num_elements)
kdist_matrix = k_dist(klist, buffer, self.distance_matrix)
dbscan_param_pairs = dbscanTools.dbscan_param_space_search(self.parameters["clustering"]["evaluation"]["maximum_noise"],
max_eps_tries,
num_elements,
klist,
kdist_matrix) +\
zhou_adaptative_determination(kdist_matrix, self.distance_matrix)
for (minpts, eps) in dbscan_param_pairs:
run_parameter = ParametersGenerator.get_base_parameters()
run_parameter["minpts"] = minpts
run_parameter["eps"] = eps
run_parameters.append(run_parameter)
return run_parameters, []
def test_zhou_adaptative_determination(self):
distances = CondensedMatrix([ 0., 0., 2., 2.,
0., 2., 2.,
2., 2.,
0.])
buffer = numpy.empty(distances.row_length)
k_dist_matrix = k_dist(numpy.array([2,4]), buffer, distances)
zhou_params = zhou_adaptative_determination(k_dist_matrix, distances)
self.assertItemsEqual(zhou_params,[(1.0, 0.8), (4.0, 2.0)])
def test_k_dist(self):
distances = CondensedMatrix([
17.46, 9.21, 4.47, 3.16, 4.47, 5.65, 12.36, 5.83, 9.43, 12.52,
15.65, 5., 8.06, 11.18, 13.15, 3.16, 6.35, 8.24, 3.16, 5.10, 2.
])
buffer = numpy.empty(distances.row_length)
result = k_dist(numpy.array([2, 4]), buffer, distances)
# [[ 4.47 5.65]
# [ 9.43 12.52]
# [ 8.06 11.18]
# [ 4.47 5.83]
# [ 3.16 5.1 ]
# [ 3.16 6.35]
# [ 5.1 8.24]]
expected = [
sorted([4.47, 9.43, 8.06, 4.47, 3.16, 3.16, 5.1]),
sorted([5.65, 12.52, 11.18, 5.83, 5.1, 6.35, 8.24])
]
numpy.testing.assert_array_almost_equal(result, expected, decimal=2)
def test_k_dist(self):
distances = CondensedMatrix([17.46, 9.21, 4.47, 3.16, 4.47, 5.65,
12.36, 5.83, 9.43, 12.52, 15.65,
5., 8.06, 11.18, 13.15,
3.16, 6.35, 8.24,
3.16, 5.10,
2. ])
buffer = numpy.empty(distances.row_length)
result = k_dist(numpy.array([2,4]), buffer, distances)
# [[ 4.47 5.65]
# [ 9.43 12.52]
# [ 8.06 11.18]
# [ 4.47 5.83]
# [ 3.16 5.1 ]
# [ 3.16 6.35]
# [ 5.1 8.24]]
expected = [sorted([ 4.47, 9.43, 8.06, 4.47, 3.16, 3.16, 5.1]),
sorted([ 5.65, 12.52, 11.18, 5.83, 5.1, 6.35, 8.24])]
numpy.testing.assert_array_almost_equal(result,expected,decimal = 2)
def test_zhou_adaptative_determination(self):
distances = CondensedMatrix([0., 0., 2., 2., 0., 2., 2., 2., 2., 0.])
buffer = numpy.empty(distances.row_length)
k_dist_matrix = k_dist(numpy.array([2, 4]), buffer, distances)
zhou_params = zhou_adaptative_determination(k_dist_matrix, distances)
self.assertItemsEqual(zhou_params, [(1.0, 0.8), (4.0, 2.0)])