|
1
|
|
|
"""!
|
|
2
|
|
|
|
|
3
|
|
|
@brief Cluster analysis algorithm: K-Medians
|
|
4
|
|
|
@details Implementation based on paper @cite book::algorithms_for_clustering_data.
|
|
5
|
|
|
|
|
6
|
|
|
@authors Andrei Novikov ([email protected])
|
|
7
|
|
|
@date 2014-2018
|
|
8
|
|
|
@copyright GNU Public License
|
|
9
|
|
|
|
|
10
|
|
|
@cond GNU_PUBLIC_LICENSE
|
|
11
|
|
|
PyClustering is free software: you can redistribute it and/or modify
|
|
12
|
|
|
it under the terms of the GNU General Public License as published by
|
|
13
|
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
14
|
|
|
(at your option) any later version.
|
|
15
|
|
|
|
|
16
|
|
|
PyClustering is distributed in the hope that it will be useful,
|
|
17
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
18
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
19
|
|
|
GNU General Public License for more details.
|
|
20
|
|
|
|
|
21
|
|
|
You should have received a copy of the GNU General Public License
|
|
22
|
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
23
|
|
|
@endcond
|
|
24
|
|
|
|
|
25
|
|
|
"""
|
|
26
|
|
|
|
|
27
|
|
|
|
|
28
|
|
|
import math
|
|
29
|
|
|
|
|
30
|
|
|
from pyclustering.cluster.encoder import type_encoding
|
|
31
|
|
|
|
|
32
|
|
|
from pyclustering.utils.metric import distance_metric, type_metric
|
|
33
|
|
|
|
|
34
|
|
|
import pyclustering.core.kmedians_wrapper as wrapper
|
|
35
|
|
|
|
|
36
|
|
|
from pyclustering.core.wrapper import ccore_library
|
|
37
|
|
|
from pyclustering.core.metric_wrapper import metric_wrapper
|
|
38
|
|
|
|
|
39
|
|
|
|
|
40
|
|
|
class kmedians:
|
|
41
|
|
|
"""!
|
|
42
|
|
|
@brief Class represents clustering algorithm K-Medians.
|
|
43
|
|
|
@details The algorithm is less sensitive to outliers than K-Means. Medians are calculated instead of centroids.
|
|
44
|
|
|
|
|
45
|
|
|
CCORE option can be used to use the pyclustering core - C/C++ shared library for processing that significantly increases performance.
|
|
46
|
|
|
|
|
47
|
|
|
Example:
|
|
48
|
|
|
@code
|
|
49
|
|
|
# load list of points for cluster analysis
|
|
50
|
|
|
sample = read_sample(path);
|
|
51
|
|
|
|
|
52
|
|
|
# create instance of K-Medians algorithm
|
|
53
|
|
|
kmedians_instance = kmedians(sample, [ [0.0, 0.1], [2.5, 2.6] ]);
|
|
54
|
|
|
|
|
55
|
|
|
# run cluster analysis and obtain results
|
|
56
|
|
|
kmedians_instance.process();
|
|
57
|
|
|
kmedians_instance.get_clusters();
|
|
58
|
|
|
@endcode
|
|
59
|
|
|
|
|
60
|
|
|
"""
|
|
61
|
|
|
|
|
62
|
|
View Code Duplication |
def __init__(self, data, initial_centers, tolerance=0.001, ccore=True, **kwargs):
|
|
|
|
|
|
|
63
|
|
|
"""!
|
|
64
|
|
|
@brief Constructor of clustering algorithm K-Medians.
|
|
65
|
|
|
|
|
66
|
|
|
@param[in] data (list): Input data that is presented as list of points (objects), each point should be represented by list or tuple.
|
|
67
|
|
|
@param[in] initial_centers (list): Initial coordinates of medians of clusters that are represented by list: [center1, center2, ...].
|
|
68
|
|
|
@param[in] tolerance (double): Stop condition: if maximum value of change of centers of clusters is less than tolerance than algorithm will stop processing
|
|
69
|
|
|
@param[in] ccore (bool): Defines should be CCORE library (C++ pyclustering library) used instead of Python code or not.
|
|
70
|
|
|
@param[in] **kwargs: Arbitrary keyword arguments (available arguments: 'metric').
|
|
71
|
|
|
|
|
72
|
|
|
<b>Keyword Args:</b><br>
|
|
73
|
|
|
- metric (distance_metric): Metric that is used for distance calculation between two points.
|
|
74
|
|
|
|
|
75
|
|
|
"""
|
|
76
|
|
|
self.__pointer_data = data
|
|
77
|
|
|
self.__clusters = []
|
|
78
|
|
|
self.__medians = initial_centers[:]
|
|
79
|
|
|
self.__tolerance = tolerance
|
|
80
|
|
|
|
|
81
|
|
|
self.__metric = kwargs.get('metric', distance_metric(type_metric.EUCLIDEAN_SQUARE))
|
|
82
|
|
|
if self.__metric is None:
|
|
83
|
|
|
self.__metric = distance_metric(type_metric.EUCLIDEAN_SQUARE)
|
|
84
|
|
|
|
|
85
|
|
|
self.__ccore = ccore and self.__metric.get_type() != type_metric.USER_DEFINED
|
|
86
|
|
|
if self.__ccore:
|
|
87
|
|
|
self.__ccore = ccore_library.workable()
|
|
88
|
|
|
|
|
89
|
|
|
|
|
90
|
|
View Code Duplication |
def process(self):
|
|
|
|
|
|
|
91
|
|
|
"""!
|
|
92
|
|
|
@brief Performs cluster analysis in line with rules of K-Medians algorithm.
|
|
93
|
|
|
|
|
94
|
|
|
@remark Results of clustering can be obtained using corresponding get methods.
|
|
95
|
|
|
|
|
96
|
|
|
@see get_clusters()
|
|
97
|
|
|
@see get_medians()
|
|
98
|
|
|
|
|
99
|
|
|
"""
|
|
100
|
|
|
|
|
101
|
|
|
if self.__ccore is True:
|
|
102
|
|
|
ccore_metric = metric_wrapper.create_instance(self.__metric)
|
|
103
|
|
|
|
|
104
|
|
|
self.__clusters = wrapper.kmedians(self.__pointer_data, self.__medians, self.__tolerance, ccore_metric.get_pointer())
|
|
105
|
|
|
self.__medians = self.__update_medians()
|
|
106
|
|
|
|
|
107
|
|
|
else:
|
|
108
|
|
|
changes = float('inf')
|
|
109
|
|
|
|
|
110
|
|
|
# Check for dimension
|
|
111
|
|
|
if len(self.__pointer_data[0]) != len(self.__medians[0]):
|
|
112
|
|
|
raise NameError('Dimension of the input data and dimension of the initial medians must be equal.')
|
|
113
|
|
|
|
|
114
|
|
|
while changes > self.__tolerance:
|
|
115
|
|
|
self.__clusters = self.__update_clusters()
|
|
116
|
|
|
updated_centers = self.__update_medians()
|
|
117
|
|
|
|
|
118
|
|
|
changes = max([self.__metric(self.__medians[index], updated_centers[index]) for index in range(len(updated_centers))])
|
|
119
|
|
|
|
|
120
|
|
|
self.__medians = updated_centers
|
|
121
|
|
|
|
|
122
|
|
|
|
|
123
|
|
|
def get_clusters(self):
|
|
124
|
|
|
"""!
|
|
125
|
|
|
@brief Returns list of allocated clusters, each cluster contains indexes of objects in list of data.
|
|
126
|
|
|
|
|
127
|
|
|
@see process()
|
|
128
|
|
|
@see get_medians()
|
|
129
|
|
|
|
|
130
|
|
|
"""
|
|
131
|
|
|
|
|
132
|
|
|
return self.__clusters
|
|
133
|
|
|
|
|
134
|
|
|
|
|
135
|
|
|
def get_medians(self):
|
|
136
|
|
|
"""!
|
|
137
|
|
|
@brief Returns list of centers of allocated clusters.
|
|
138
|
|
|
|
|
139
|
|
|
@see process()
|
|
140
|
|
|
@see get_clusters()
|
|
141
|
|
|
|
|
142
|
|
|
"""
|
|
143
|
|
|
|
|
144
|
|
|
return self.__medians
|
|
145
|
|
|
|
|
146
|
|
|
|
|
147
|
|
|
def get_cluster_encoding(self):
|
|
148
|
|
|
"""!
|
|
149
|
|
|
@brief Returns clustering result representation type that indicate how clusters are encoded.
|
|
150
|
|
|
|
|
151
|
|
|
@return (type_encoding) Clustering result representation.
|
|
152
|
|
|
|
|
153
|
|
|
@see get_clusters()
|
|
154
|
|
|
|
|
155
|
|
|
"""
|
|
156
|
|
|
|
|
157
|
|
|
return type_encoding.CLUSTER_INDEX_LIST_SEPARATION
|
|
158
|
|
|
|
|
159
|
|
|
|
|
160
|
|
View Code Duplication |
def __update_clusters(self):
|
|
|
|
|
|
|
161
|
|
|
"""!
|
|
162
|
|
|
@brief Calculate Manhattan distance to each point from the each cluster.
|
|
163
|
|
|
@details Nearest points are captured by according clusters and as a result clusters are updated.
|
|
164
|
|
|
|
|
165
|
|
|
@return (list) updated clusters as list of clusters where each cluster contains indexes of objects from data.
|
|
166
|
|
|
|
|
167
|
|
|
"""
|
|
168
|
|
|
|
|
169
|
|
|
clusters = [[] for i in range(len(self.__medians))]
|
|
170
|
|
|
for index_point in range(len(self.__pointer_data)):
|
|
171
|
|
|
index_optim = -1
|
|
172
|
|
|
dist_optim = 0.0
|
|
173
|
|
|
|
|
174
|
|
|
for index in range(len(self.__medians)):
|
|
175
|
|
|
dist = self.__metric(self.__pointer_data[index_point], self.__medians[index])
|
|
176
|
|
|
|
|
177
|
|
|
if (dist < dist_optim) or (index is 0):
|
|
178
|
|
|
index_optim = index
|
|
179
|
|
|
dist_optim = dist
|
|
180
|
|
|
|
|
181
|
|
|
clusters[index_optim].append(index_point)
|
|
182
|
|
|
|
|
183
|
|
|
# If cluster is not able to capture object it should be removed
|
|
184
|
|
|
clusters = [cluster for cluster in clusters if len(cluster) > 0]
|
|
185
|
|
|
|
|
186
|
|
|
return clusters
|
|
187
|
|
|
|
|
188
|
|
|
|
|
189
|
|
|
def __update_medians(self):
|
|
190
|
|
|
"""!
|
|
191
|
|
|
@brief Calculate medians of clusters in line with contained objects.
|
|
192
|
|
|
|
|
193
|
|
|
@return (list) list of medians for current number of clusters.
|
|
194
|
|
|
|
|
195
|
|
|
"""
|
|
196
|
|
|
|
|
197
|
|
|
medians = [[] for i in range(len(self.__clusters))]
|
|
198
|
|
|
|
|
199
|
|
|
for index in range(len(self.__clusters)):
|
|
200
|
|
|
medians[index] = [ 0.0 for i in range(len(self.__pointer_data[0]))]
|
|
201
|
|
|
length_cluster = len(self.__clusters[index])
|
|
202
|
|
|
|
|
203
|
|
|
for index_dimension in range(len(self.__pointer_data[0])):
|
|
204
|
|
|
sorted_cluster = sorted(self.__clusters[index], key=lambda x: self.__pointer_data[x][index_dimension])
|
|
205
|
|
|
|
|
206
|
|
|
relative_index_median = int(math.floor((length_cluster - 1) / 2))
|
|
207
|
|
|
index_median = sorted_cluster[relative_index_median]
|
|
208
|
|
|
|
|
209
|
|
|
if (length_cluster % 2) == 0:
|
|
210
|
|
|
index_median_second = sorted_cluster[relative_index_median + 1]
|
|
211
|
|
|
medians[index][index_dimension] = (self.__pointer_data[index_median][index_dimension] + self.__pointer_data[index_median_second][index_dimension]) / 2.0
|
|
212
|
|
|
|
|
213
|
|
|
else:
|
|
214
|
|
|
medians[index][index_dimension] = self.__pointer_data[index_median][index_dimension]
|
|
215
|
|
|
|
|
216
|
|
|
return medians
|
|
217
|
|
|
|
annoviko /
pyclustering
Loading history...