Database Reference
In-Depth Information
Clusters
Feature
Importance
1
Cluster
cluster-1
cluster-2
cluster-3
cluster-4
cluster-5
cluster-6
cluster-7
cluster-8
Label
Description
Features
Component_1 -
Fashion
−
0.16
Component_1 -
Fashion
−
Component_1 -
Fashion
−
0.45
Component_1 -
Fashion
−
Component_1 -
Fashion
−
Component_1 -
Fashion
1.77
Component_1 -
Fashion
−
Component_1 -
Fashion
−
0.67
0.56
0.60
1.16
0.74
Component_2 -
Daily needs
0.02
Component_2 -
Daily needs
−
Component_2 -
Daily needs
2.76
Component_2 -
Daily needs
−
Component_2 -
Daily needs
−
Component_2 -
Daily needs
−
Component_2 -
Daily needs
−
Component_2 -
Daily needs
−
0.54
0.68
0.42
0.26
1.03
0.70
Component_3 -
Appliances
−
0.05
Component_3 -
Appliances
0.36
Component_3 -
Appliances
0.17
Component_3 -
Appliances
1.44
Component_3 -
Appliances
0.30
Component_3 -
Appliances
0.08
Component_3 -
Appliances
1.03
Component_3 -
Appliances
−
3.36
Component_4 -
Telcos
−
0.02
Component_4 -
Telcos
−
2.96
Component_4 -
Telcos
0.23
Component_4 -
Telcos
0.90
Component_4 -
Telcos
0.44
Component_4 -
Telcos
0.14
Component_4 -
Telcos
1.53
Component_4 -
Telcos
0.58
Component_5 -
Health
−
0.04
Component_5 -
Health
0.37
Component_5 -
Health
0.16
Component_5 -
Health
1.10
Component_5 -
Health
−
2.12
Component_5 -
Health
0.07
Component_5 -
Health
0.55
Component_5 -
Health
0.47
Component_6 -
Home/Garden
−
0.01
Component_6 -
Home/Garden
−
0.40
Component_6 -
Home/Garden
−
0.18
Component_6 -
Home/Garden
−
0.08
Component_6 -
Home/Garden
2.01
Component_6 -
Home/Garden
−
0.13
Component_6 -
Home/Garden
−
1.47
Component_6 -
Home/Garden
−
0.31
Component_7 -
Tr a v e l
−
0.06
Component_7 -
Tr a v e l
−
0.20
Component_7 -
Tr a v e l
−
0.19
Component_7 -
Tr a v e l
1.22
Component_7 -
Tr a v e l
−
0.12
Component_7 -
Tr a v e l
−
0.07
Component_7 -
Tr a v e l
0.05
Component_7 -
Tr a v e l
0.01
Figure 6.4
The table of cluster centers.
The cluster centroids and these tables convey similar results. Most of the
clusters seem to be related to specific merchant categories and purchasing habits.
More specifically, cluster 1 contains customers with average spending in many
different merchant categories. These customers have the most diverse purchasing
habits. They present moderate usage and tend to buy a variety of different
products with their cards. On the other hand, the rest of the clusters have
more partial preferences and seem more tightly connected with the buying
of specific products. Cluster 2 is associated with fees for telecommunication
services and cluster 3 is characterized by increased spending at supermarkets and
gas stations. Moreover, cluster 4 shows increased usage for traveling expenses.
Spending at home/garden stores and for health purposes appears to be a defining
characteristic of cluster 5. Customers assigned to cluster 6 tend to buy clothing
and accessories. Cluster 7 relates to spending for cultural/educational reasons
and entertainment, whereas customers in cluster 8 use their cards mainly for the
purchase of appliances.
It is always a good idea to depict results in a graphical way so that the profiles
of each cluster are clearly illustrated. The bubble plot in Figure 6.5 is based on
aggregated data. Each ''dot'' represents a cluster. The
X
-axis co-ordinate of each
dot or cluster denotes its mean percentage of purchases at apparel stores, the
Search WWH ::
Custom Search