def data_clustering(data, distance=Orange.distance.Euclidean, linkage=AVERAGE):
"""
Return the hierarchical clustering of the data set's rows.
:param Orange.data.Table data: Data set to cluster.
:param Orange.distance.Distance distance: A distance measure.
:param str linkage:
"""
matrix = distance(data)
return dist_matrix_clustering(matrix, linkage=linkage)
def compute_distances(self):
if self.data is None or len(self.data) == 0 \
or self.reference is None or len(self.reference) == 0:
self.distances = None
return
distance = METRICS[self.distance_index][1]
n_ref = len(self.reference)
all_data = Table.concatenate([self.reference, self.data], 0)
pp_all_data = Impute()(RemoveNaNColumns()(all_data))
pp_reference, pp_data = pp_all_data[:n_ref], pp_all_data[n_ref:]
self.distances = distance(pp_data, pp_reference).min(axis=1)
def feature_clustering(data, distance=Orange.distance.PearsonR,
linkage=AVERAGE):
"""
Return the hierarchical clustering of the data set's columns.
:param Orange.data.Table data: Data set to cluster.
:param Orange.distance.Distance distance: A distance measure.
:param str linkage:
"""
matrix = distance(data, axis=1)
return dist_matrix_clustering(matrix, linkage=linkage)
def data_clustering(data, distance=Orange.distance.Euclidean,
linkage=AVERAGE):
"""
Return the hierarchical clustering of the data set's rows.
:param Orange.data.Table data: Data set to cluster.
:param Orange.distance.Distance distance: A distance measure.
:param str linkage:
"""
matrix = distance(data)
return dist_matrix_clustering(matrix, linkage=linkage)
def feature_clustering(data, distance=Orange.distance.PearsonR,
linkage=AVERAGE):
"""
Return the hierarchical clustering of the data set's columns.
:param Orange.data.Table data: Data set to cluster.
:param Orange.distance.Distance distance: A distance measure.
:param str linkage:
"""
matrix = distance(data, axis=0)
return dist_matrix_clustering(matrix, linkage=linkage)
def compute_distances(self):
self.Error.diff_domains.clear()
if self.data is None or len(self.data) == 0 \
or self.reference is None or len(self.reference) == 0:
self.distances = None
return
if self.reference.domain != self.data.domain:
self.Error.diff_domains()
self.distances = None
return
distance = METRICS[self.distance_index][1]
n_ref = len(self.reference)
all_data = Table.concatenate([self.reference, self.data], 0)
pp_all_data = Impute()(RemoveNaNColumns()(all_data))
pp_reference, pp_data = pp_all_data[:n_ref], pp_all_data[n_ref:]
self.distances = distance(pp_data, pp_reference).min(axis=1)
def apply(self):
if self.data is None or self.reference is None:
self.send("Neighbors", None)
return
distance = METRICS[self.distance_index][1]
n_data, n_ref = len(self.data), len(self.reference)
all_data = Table.concatenate([self.reference, self.data], 0)
pp_all_data = Impute()(RemoveNaNColumns()(all_data))
pp_data, pp_reference = pp_all_data[n_ref:], pp_all_data[:n_ref]
dist = distance(np.vstack((pp_data, pp_reference)))[:n_data, n_data:]
data = self._add_similarity(self.data, dist)
sorted_indices = list(np.argsort(dist.flatten()))[::-1]
indices = []
while len(sorted_indices) > 0 and len(indices) < self.n_neighbors:
index = int(sorted_indices.pop() / len(self.reference))
if (self.data[index] not in self.reference or
not self.exclude_reference) and index not in indices:
indices.append(index)
neighbors = data[indices]
neighbors.attributes = self.data.attributes
self.send("Neighbors", neighbors)
def apply(self):
if self.data is None or self.reference is None:
self.send("Neighbors", None)
return
distance = METRICS[self.distance_index][1]
n_data, n_ref = len(self.data), len(self.reference)
all_data = Table.concatenate([self.reference, self.data], 0)
pp_all_data = Impute()(RemoveNaNColumns()(all_data))
pp_data, pp_reference = pp_all_data[n_ref:], pp_all_data[:n_ref]
dist = distance(np.vstack((pp_data, pp_reference)))[:n_data, n_data:]
data = self._add_similarity(self.data, dist)
sorted_indices = list(np.argsort(dist.flatten()))[::-1]
indices = []
while len(sorted_indices) > 0 and len(indices) < self.n_neighbors:
index = int(sorted_indices.pop() / len(self.reference))
if (self.data[index] not in self.reference
or not self.exclude_reference) and index not in indices:
indices.append(index)
neighbours = data[indices]
neighbours.attributes = self.data.attributes
self.send("Neighbors", neighbours)
def compute_distances(self):
self.Error.diff_domains.clear()
if self.data is None or len(self.data) == 0 \
or self.reference is None or len(self.reference) == 0:
self.distances = None
return
if set(self.reference.domain.attributes) != \
set(self.data.domain.attributes):
self.Error.diff_domains()
self.distances = None
return
distance = METRICS[self.distance_index][1]
n_ref = len(self.reference)
# comparing only attributes, no metas and class-vars
new_domain = Domain(self.data.domain.attributes)
reference = self.reference.transform(new_domain)
data = self.data.transform(new_domain)
all_data = Table.concatenate([reference, data], 0)
pp_all_data = Impute()(RemoveNaNColumns()(all_data))
pp_reference, pp_data = pp_all_data[:n_ref], pp_all_data[n_ref:]
self.distances = distance(pp_data, pp_reference).min(axis=1)