Information Technology Reference
In-Depth Information
does not move around throughout the lifetime of the network, fewer resources need to be dedicated into
identifying each single mobile node and its current location in relation to the time series.
The motivation of the proposed method is that matrix, due to the fixed structure could fit into the
mobile environment perfectly by using static nodes. By using static nodes, the behaviour of the mobile
users can still be analysed by using location dependent method and merely identifying the mobile users
without any tracking system. This allows great performance improvements and cost savings and easy
deployment of mobile user data mining systems.
The matrix pattern based mobile user data mining methodology can be divided into two categories,
single layer matrix and multi-layer matrix.
A
single layer matrix
uses a two dimensional, single layer matrix for mobile user data mining.
Example :
In a shopping centre, retailers of fashion wish to find out how their stores are frequented by mobile
users. Fashion is the single layer domain. Consider that there are Retailers
A
,
B
, and
C
in the shopping
centre and participate in this mining exercise. The mining exercise is based on the frequency of contact
that a mobile user has on a particular point in the matrix. A possible outcome would be:
Relationship 1:
A
-
B
= 70%
Relationship 2:
B
-
C
= 80%
Relationship 3:
A
-
C
= 30%
The above relationships indicate that mobile users frequent both
A
and
B
at the same time, or
B
and
C
at the same time, but not
A
and
C
at the same time. This suggests that
A
may be a total replacement
of
C
as mobile users may see no point to visit
C
.
A
multi-layer matrix
uses a three-dimensional matrix for mobile user data mining and is an exten-
sion of the single layer matrix to accommodate for robustness and flexibility.
Example :
In a shopping centre, retailers of fashion and retailers of leisure wish to find out how their stores are
frequented by mobile users. Knowing the reason why could improve their business performance. Fashion
and leisure are the two domains in this context. Here, a multi-layer matrix pattern is used.
The mining exercise first performs the single layer operation of fashion and leisure. After this has
been done, fashion and leisure domain are analysed together. Consider that there are two retailers of
fashion and two retailers of leisure namely
F
1,
F
2,
L
1,
L
2 respectively.
Relationship 1:
F
1,
L
1 = 70%
Relationship 2:
F
2,
L
1 = 90%
Relationship 3:
F
1,
L
2 = 30%
The aobve relationships indicate that there is a strong relationship among
F
1 and
L
1 and among
F
2 and
L
1 but not among
F
1 and
L
2. This suggests that there consists among
F
1,
L
1 or
F
2,
L
1 a good
combination of service for mobile users which attracted them to frequent them together one after the
Search WWH ::
Custom Search