Information Technology Reference
In-Depth Information
(a) dataset13;
nc
=4;
σ
=0
.
065
(b) dataset22;
nc
=3;
σ
=0
.
071
(c) spiral;
nc
=2;
σ
=0
.
0272
(d) spiral;
nc
=2;
σ
=0
.
0275
Fig. 6.34 Some clustering solutions given by Spectral-Ng. Each label shows: the
dataset name; the pre-established number of clusters and the
σ
value.
Tabl e 6 . 21 Results and parameters used in the comparison of LEGClust and
Spectral clustering in experiments with DHN, 20NewsGroups and NCI Microarray
datasets.
LEGClust
Spectral-Ng
Spectral-Shi
Mk ARI nc
σ
ARI
nc
σ
ARI
DHN
30 10 0.628
10 10 0.287
10 12 0.573
30 8 0.608
30 12 0.574
20NewsGroups
20
3
0.289
20 12 0.479
20 20 0.006
20
2
0.287
NCI Microarray
4
2
0.148
3
80 0.177
3
10 0.138
6
3
0.148
10
3
0.148
In Table 6.22 we show an example of a confusion matrix obtained with
LEGClust for an experiment with the DHN dataset.
In the experiments with the 20NewsGroups dataset, a random sub-sample
of 1000 elements from the original dataset was used. This dataset is a 20 class
text classification set obtained from 20 different news groups. The dataset
was prepared by stemming words according to the algorithm described in