Information Technology Reference
In-Depth Information
Figure 20. Frequency pattern
0.9
0.9
A
B
A
B
0.5
0.5
0.8
0.8
C
D
C
D
0.7
0.7
series allows better control for the decision maker. The use of frequency in calculating the confidence
of group characteristics of mobile users allows better representation of group formations of mobile users
regardless of their physical location. In this way, a family group that spread throughout different parts
of the world but remain constant in touch with each other will be detected as a group.
Figure 20 illustrates the frequency pattern with
A
,
B
,
C
, and
D
represented as mobile nodes. It can be
observed that each mobile node has a specific relationship confidence calculated based on the frequency
of communication between them. Each arrow represents the logical relationship (Goh & Taniar, 2004a)
among two mobile nodes.
A
,
B
,
C
, and
D
are a set of mobile nodes in the mobile environment that have
been determined by the algorithm as logically close to each other. The pre-specified criteria (Goh &
Taniar, 2004a) allow different emphasis on different parts of the time series. The diagram on the left
in Figure 20 shows the calculated confidences between nodes. When the confidence threshold of
0.6
is
set, those relationships which have less confidence are discarded and the right diagram represents the
final outcome of frequency pattern that is a list of group of mobile users that are frequently staying in
touch with each other (Goh & Taniar, 2004a).
Parallel Pattern
Figure 21 shows the fundamental concept of parallel pattern, which is to find the similarities of arrows,
that move in similar directions. The goal of parallel pattern is to find out the similarities in decisions,
such as similarities in decisions to move or similarities of decision to change taste from one starting
point to another among many mobile users. The result of this exercise is a better understanding of the
behaviour of mobile users.
Our related work aims to address different parts of the nature of the problem faced in finding useful
knowledge from mobile users. Frequency pattern addresses the issue of using frequency rather than
physical distance in order to determine relative closeness. Parallel pattern on the other hand, addresses
the interesting issue of movement decisions of mobile users and is a method, which proactively seeks
and determines the behaviour of mobile users. All previously proposed methods are resource consum-
ing as there is a requirement to constantly identify the mobile nodes as they move from one location to
another. Matrix pattern aims to solve this problem by eliminating the need to precisely identifying the
coordinates of the mobile nodes by using a simplified matrix method to identify and mark them, and
perform mining on the matrix itself for the rest of the mining process.
Search WWH ::
Custom Search