|
1
|
|
|
"""!
|
|
2
|
|
|
|
|
3
|
|
|
@brief Cluster analysis algorithm: TTSAS (Two-Threshold Sequential Algorithmic Scheme).
|
|
4
|
|
|
@details Implementation based on book:
|
|
5
|
|
|
- Theodoridis, Koutroumbas, Konstantinos. Elsevier Academic Press - Pattern Recognition - 2nd Edition. 2003.
|
|
6
|
|
|
|
|
7
|
|
|
@authors Andrei Novikov ([email protected])
|
|
8
|
|
|
@date 2014-2018
|
|
9
|
|
|
@copyright GNU Public License
|
|
10
|
|
|
|
|
11
|
|
|
@cond GNU_PUBLIC_LICENSE
|
|
12
|
|
|
PyClustering is free software: you can redistribute it and/or modify
|
|
13
|
|
|
it under the terms of the GNU General Public License as published by
|
|
14
|
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
15
|
|
|
(at your option) any later version.
|
|
16
|
|
|
|
|
17
|
|
|
PyClustering is distributed in the hope that it will be useful,
|
|
18
|
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
19
|
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
20
|
|
|
GNU General Public License for more details.
|
|
21
|
|
|
|
|
22
|
|
|
You should have received a copy of the GNU General Public License
|
|
23
|
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
24
|
|
|
@endcond
|
|
25
|
|
|
|
|
26
|
|
|
"""
|
|
27
|
|
|
|
|
28
|
|
|
|
|
29
|
|
|
from pyclustering.core.ttsas_wrapper import ttsas as ttsas_wrapper;
|
|
30
|
|
|
from pyclustering.core.metric_wrapper import metric_wrapper;
|
|
31
|
|
|
|
|
32
|
|
|
from pyclustering.cluster.bsas import bsas;
|
|
33
|
|
|
|
|
34
|
|
|
|
|
35
|
|
|
class ttsas(bsas):
|
|
36
|
|
|
"""!
|
|
37
|
|
|
@brief Class represents TTSAS (Two-Threshold Sequential Algorithmic Scheme).
|
|
38
|
|
|
@details Clustering results of BSAS and MBSAS are strongly dependent on the order in which the points in data.
|
|
39
|
|
|
TTSAS helps to overcome this shortcoming by using two threshold parameters. The first - if the distance
|
|
40
|
|
|
to the nearest cluster is less than the first threshold then point is assigned to the cluster. The
|
|
41
|
|
|
second - if distance to the nearest cluster is greater than the second threshold then new cluster is
|
|
42
|
|
|
allocated.
|
|
43
|
|
|
|
|
44
|
|
|
Code example of TTSAS usage:
|
|
45
|
|
|
@code
|
|
46
|
|
|
# Read data sample from 'Simple03.data'.
|
|
47
|
|
|
sample = read_sample(SIMPLE_SAMPLES.SAMPLE_SIMPLE3);
|
|
48
|
|
|
|
|
49
|
|
|
# Prepare algorithm's parameters.
|
|
50
|
|
|
threshold1 = 1.0;
|
|
51
|
|
|
threshold2 = 2.0;
|
|
52
|
|
|
|
|
53
|
|
|
# Create instance of TTSAS algorithm.
|
|
54
|
|
|
ttsas_instance = ttsas(sample, max_clusters, threshold);
|
|
55
|
|
|
ttsas_instance.process();
|
|
56
|
|
|
|
|
57
|
|
|
# Get clustering results.
|
|
58
|
|
|
clusters = ttsas_instance.get_clusters();
|
|
59
|
|
|
representatives = ttsas_instance.get_representatives();
|
|
60
|
|
|
|
|
61
|
|
|
# Display results using BSAS visualizer.
|
|
62
|
|
|
bsas_visualizer.show_clusters(sample, clusters, representatives);
|
|
63
|
|
|
@endcode
|
|
64
|
|
|
|
|
65
|
|
|
@see pyclustering.cluster.bsas, pyclustering.cluster.mbsas
|
|
66
|
|
|
|
|
67
|
|
|
"""
|
|
68
|
|
|
|
|
69
|
|
|
def __init__(self, data, threshold1, threshold2, ccore, **kwargs):
|
|
70
|
|
|
"""!
|
|
71
|
|
|
@brief Creates TTSAS algorithm.
|
|
72
|
|
|
|
|
73
|
|
|
@param[in] data (list): Input data that is presented as list of points (objects), each point should be represented by list or tuple.
|
|
74
|
|
|
@param[in] threshold1: Dissimilarity level (distance) between point and its closest cluster, if the distance is
|
|
75
|
|
|
less than 'threshold1' value then point is assigned to the cluster.
|
|
76
|
|
|
@param[in] threshold2: Dissimilarity level (distance) between point and its closest cluster, if the distance is
|
|
77
|
|
|
greater than 'threshold2' value then point is considered as a new cluster.
|
|
78
|
|
|
@param[in] ccore (bool): If True than DLL CCORE (C++ solution) will be used for solving.
|
|
79
|
|
|
@param[in] **kwargs: Arbitrary keyword arguments (available arguments: 'metric').
|
|
80
|
|
|
|
|
81
|
|
|
Keyword Args:
|
|
82
|
|
|
metric (distance_metric): Metric that is used for distance calculation between two points.
|
|
83
|
|
|
|
|
84
|
|
|
"""
|
|
85
|
|
|
|
|
86
|
|
|
self._threshold2 = threshold2;
|
|
87
|
|
|
self._amount_skipped_objects = len(data);
|
|
88
|
|
|
self._skipped_objects = [ True ] * len(data);
|
|
89
|
|
|
|
|
90
|
|
|
super().__init__(data, len(data), threshold1, ccore, **kwargs);
|
|
91
|
|
|
|
|
92
|
|
|
|
|
93
|
|
|
def process(self):
|
|
94
|
|
|
"""!
|
|
95
|
|
|
@brief Performs cluster analysis in line with rules of BSAS algorithm.
|
|
96
|
|
|
|
|
97
|
|
|
@remark Results of clustering can be obtained using corresponding get methods.
|
|
98
|
|
|
|
|
99
|
|
|
@see get_clusters()
|
|
100
|
|
|
@see get_representatives()
|
|
101
|
|
|
|
|
102
|
|
|
"""
|
|
103
|
|
|
|
|
104
|
|
|
if self._ccore is True:
|
|
105
|
|
|
self.__process_by_ccore();
|
|
106
|
|
|
else:
|
|
107
|
|
|
self.__prcess_by_python();
|
|
108
|
|
|
|
|
109
|
|
|
|
|
110
|
|
|
def __process_by_ccore(self):
|
|
111
|
|
|
ccore_metric = metric_wrapper.create_instance(self._metric);
|
|
112
|
|
|
self._clusters, self._representatives = ttsas_wrapper(self._data, self._threshold, self._threshold2, ccore_metric.get_pointer());
|
|
113
|
|
|
|
|
114
|
|
|
|
|
115
|
|
|
def __prcess_by_python(self):
|
|
116
|
|
|
changes = 0;
|
|
117
|
|
|
while self._amount_skipped_objects != 0:
|
|
118
|
|
|
previous_amount = self._amount_skipped_objects;
|
|
119
|
|
|
self.__process_objects(changes);
|
|
120
|
|
|
|
|
121
|
|
|
changes = previous_amount - self._amount_skipped_objects;
|
|
122
|
|
|
|
|
123
|
|
|
|
|
124
|
|
|
def __process_objects(self, changes):
|
|
125
|
|
|
index_point = self._skipped_objects.index(True);
|
|
126
|
|
|
|
|
127
|
|
|
if changes == 0:
|
|
128
|
|
|
self.__allocate_cluster(index_point, self._data[index_point]);
|
|
129
|
|
|
index_point += 1;
|
|
130
|
|
|
|
|
131
|
|
|
for i in range(index_point, len(self._data)):
|
|
132
|
|
|
if self._skipped_objects[i] is True:
|
|
133
|
|
|
self.__process_skipped_object(i);
|
|
134
|
|
|
|
|
135
|
|
|
|
|
136
|
|
|
def __process_skipped_object(self, index_point):
|
|
137
|
|
|
point = self._data[index_point];
|
|
138
|
|
|
|
|
139
|
|
|
index_cluster, distance = self._find_nearest_cluster(point);
|
|
140
|
|
|
|
|
141
|
|
|
if distance <= self._threshold:
|
|
142
|
|
|
self.__append_to_cluster(index_cluster, index_point, point);
|
|
143
|
|
|
elif distance > self._threshold2:
|
|
144
|
|
|
self.__allocate_cluster(index_point, point);
|
|
145
|
|
|
|
|
146
|
|
|
|
|
147
|
|
|
def __append_to_cluster(self, index_cluster, index_point, point):
|
|
148
|
|
|
self._clusters[index_cluster].append(index_point);
|
|
149
|
|
|
self._update_representative(index_cluster, point);
|
|
150
|
|
|
|
|
151
|
|
|
self._amount_skipped_objects -= 1;
|
|
152
|
|
|
self._skipped_objects[index_point] = False;
|
|
153
|
|
|
|
|
154
|
|
|
|
|
155
|
|
|
def __allocate_cluster(self, index_point, point):
|
|
156
|
|
|
self._clusters.append( [index_point] );
|
|
157
|
|
|
self._representatives.append(point);
|
|
158
|
|
|
|
|
159
|
|
|
self._amount_skipped_objects -= 1;
|
|
160
|
|
|
self._skipped_objects[index_point] = False;
|
|
161
|
|
|
|
annoviko /
pyclustering
