Database Reference
In-Depth Information
Clusters
Feature
Importance
1
cluster
Label
Size
cluster-1
cluster-2
cluster-3
cluster-4
cluster-5
cluster-6
Active users
Voice users
SMSers
Roamers
Tech users
Basic users
24.5%
(3817)
7.6%
(1179)
8.2%
(1273)
34.7%
(5405)
1.6%
(250)
23.5%
(3660)
Features
Component 1-SMS
usage
0.52
Component 1-SMS
usage
-0.15
Component 1-SMS
usage
1.83
Component 1-SMS
usage
0.03
Component 1-SMS
usage
0.24
Component 1-SMS
usage
-0.43
Component 2-Voice
usage
0.77
Component 2-Voice
usage
1.71
Component 2-Voice
usage
-0.31
Component 2-Voice
usage
-0.20
Component 2-Voice
usage
-0.07
Component 2-Voice
usage
-0.24
Component 3 -
Roaming usage
5.58
Component 4-MMS
usage
3.25
Component 3 -
Roaming usage
-0.20
Component 3 -
Roaming usage
-0.25
Component 3 -
Roaming usage
2.48
Component 3 -
Roaming usage
-0.13
Component 3 -
Roaming usage
-0.16
Component 4-MMS
usage
-0.11
Component 4- MMS
usage
-0.26
Component 4-MMS
usage
-0.13
Component 4-MMS
usage
2.44
Component 4-MMS
usage
-0.12
Component 5-
Internet usage
2.34
Component 5-
Internet usage
0.00
Component 5-
Internet usage
-0.03
Component 5-
Internet usage
-0.03
Component 5-
Internet usage
0.20
Component 5-
Internet usage
-0.03
Figure 3.15
IBM SPSS Modeler's representation of the cluster centers.
Cluster Comparison
Cluster-2
Cluster-3
-0.18
Component 1-SMS usage
-0.64
0.95
-0.06
Component 2-Voice usage
-0.570.76
Figure 3.16
Cluster comparison with boxplots.
The background plot is a boxplot that summarizes the entire population. The
vertical line inside the box represents the population median on the respective
clustering field. The median is the 50th percentile, the value which separates
the population into two sets of equal size. The width of the box is equal to the
interquartile range, the difference between the 25th and the 75th percentiles, and
indicates the degree of dispersion in the data. Thus the box represents the range
Search WWH ::
Custom Search