Information Technology Reference
In-Depth Information
Tabl e 6 . 18
The dissimilarity matrix for Fig. 6.29 dataset.
i s123456789012345
1
4.08 4.63 4.48 4.19 5.69
2
4.52
4.55 4.93 4.89
4.87
3
4.40
4.59 4.66 4.64 4.41
4
4.65 4.59
4.39 4.40 4.72
5
4.48 4.72 4.98 4.41
4.59
6
4.52 4.35
4.60 4.45 4.35
7
4.17 4.78
4.53
4.20 4.51
8
4.57
4.65 4.51
4.55 4.49
9
4.63
4.29 4.89 4.43
4.28
10
5.57
4.81 4.26 3.97 4.10
11
4.84
4.23 4.53 4.24 4.28
12
4.78
4.34
4.43 4.33 4.59
13
4.84
4.48 4.26
4.17 4.40
14
4.88
4.38 4.36 4.40
4.38
15
5.33
4.05 4.33 4.33 4.02
Tabl e 6 . 19
The entropic proximity matrix for Fig. 6.29 dataset.
Points L1 L2 L3 L4 L5
1
25436
2
13754
3
27465
4
56327
5
41623
6
94837
7
38964
8
07936
9
06847
10
89763
11
15 12 14 13
5
12
14 11 13 15
5
13
14 12 15 11
5
14
12 11 15 13
5
15
14 11 12 13
5
joining each cluster with the ones having at least
k
connections with it, or b),
joining each cluster with the one having the highest number of connections
with it, not less than a predefined
k
. In the experiments reported in [204],
this second rule proved to be more reliable, and the resulting clusters were
usually "better" than using the first rule. In our simple example we chose
to join the clusters with the maximum number of connections not less than
2 (
k
≥
2). In the second step 2 clusters are formed by joining clusters 1, 2
and 3 having at least 2 edges connecting them. Note that, apparently, there
is only one connection between each pair of clusters but, as one can confirm
in the EPM, these are double connections and therefore count as two.
