def process(self):
"""!
@brief Performs cluster analysis in line with rules of OPTICS algorithm.
@remark Results of clustering can be obtained using corresponding gets methods.
@see get_clusters()
@see get_noise()
@see get_ordering()
"""
if (self.__ccore is True):
(self.__clusters, self.__noise, self.__ordering,
self.__eps) = wrapper.optics(self.__sample_pointer, self.__eps,
self.__minpts,
self.__amount_clusters)
else:
self.__kdtree = kdtree(self.__sample_pointer,
range(len(self.__sample_pointer)))
self.__allocate_clusters()
if ((self.__amount_clusters is not None)
and (self.__amount_clusters != len(self.get_clusters()))):
analyser = ordering_analyser(self.get_ordering())
radius, _ = analyser.calculate_connvectivity_radius(
self.__amount_clusters)
if (radius is not None):
self.__eps = radius
self.__allocate_clusters()
def __process_by_ccore(self):
"""!
@brief Performs cluster analysis using CCORE (C/C++ part of pyclustering library).
"""
(self.__clusters, self.__noise, self.__ordering, self.__eps,
objects_indexes, objects_core_distances, objects_reachability_distances) = \
wrapper.optics(self.__sample_pointer, self.__eps, self.__minpts, self.__amount_clusters, self.__data_type)
self.__optics_objects = []
for i in range(len(objects_indexes)):
if objects_core_distances[i] < 0.0:
objects_core_distances[i] = None
if objects_reachability_distances[i] < 0.0:
objects_reachability_distances[i] = None
optics_object = optics_descriptor(objects_indexes[i], objects_core_distances[i], objects_reachability_distances[i])
optics_object.processed = True
self.__optics_objects.append(optics_object)
def fit(self, X):
"""Cluster analysis in line with rules of OPTICS algorithm.
Results of clustering can be obtained using corresponding gets methods.
"""
if not self.eps > 0.0:
raise ValueError("eps must be positive.")
if self.ccore:
# use c implementation by pyclustering
import pyclustering.core.optics_wrapper as wrapper
(self.clusters_, self.noise_, self.ordering_, self.eps) = \
wrapper.optics(X, self.eps, self.minpts,
self.n_clusters)
else:
# Performs cluster allocation and builds ordering diagram based on
# reachability-distances.
self.allocate_clusters(X)
if self.n_clusters is not None and self.n_clusters != len(
self.clusters_):
radius = calculate_connectivity_radius(
self.get_ordering(), self.n_clusters)
if radius is not None:
self.eps = radius
self.allocate_clusters(X)
labels = np.empty(X.shape[0], dtype=int)
for i, cluster in enumerate(self.clusters_):
cluster = np.array(cluster, dtype=int)
labels[cluster] = i
# noise do not belong to any cluster, so they belong to their own
len_clusters = len(self.clusters_)
for i, cluster in enumerate(self.noise_):
labels[cluster] = -1 #i + len_clusters
self.labels_ = labels
return self
def __process_by_ccore(self):
"""!
@brief Performs cluster analysis using CCORE (C/C++ part of pyclustering library).
"""
(self.__clusters, self.__noise, self.__ordering, self.__eps,
objects_indexes, objects_core_distances, objects_reachability_distances) = \
wrapper.optics(self.__sample_pointer, self.__eps, self.__minpts, self.__amount_clusters, self.__data_type)
self.__optics_objects = []
for i in range(len(objects_indexes)):
if objects_core_distances[i] < 0.0:
objects_core_distances[i] = None
if objects_reachability_distances[i] < 0.0:
objects_reachability_distances[i] = None
optics_object = optics_descriptor(objects_indexes[i], objects_core_distances[i], objects_reachability_distances[i])
optics_object.processed = True
self.__optics_objects.append(optics_object)