Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 251

If parallel pattern is used, parallel pattern, being a method that it finds the similarities of movements
by detecting the similarity of change of states among mobile users. It aims to find the trend of movements
among mobile users from one state to another. For physical parallel pattern, the frequency of occur-
rence of mobile nodes moving from B to A will then reveal the knowledge. Although both methods can
reveal the same knowledge, matrix pattern has the advantage of being able to fine tune the system, as
it utilizes the resolution method. The main disadvantage of physical parallel pattern is that, it must use
a fixed location identifier for mobile user data mining. This fixed location identifier is often the static
node itself, or the wireless access point. However, it cannot fine tune the system, by detecting patterns
that occur within the wireless access point area itself.
Therefore, the main advantage of matrix pattern is the ability to use the existing infrastructure, and
provide fine tuned data mining system, by providing coordinates based reference points for identifiers in
mobile user data mining. This extends identifications beyond the restrictions of the internal boundaries
of the wireless access points. For algorithm aspects, both methods use similar steps in arriving at the
same knowledge, that is, by using frequency as the primary unit of measurement to generate knowledge
by means of percentage of confidence.

Performance of Multi-Layered Matrix

Figure 17 highlights the comparisons for net location identifiers between the three configurations in the
multi-layer matrix pattern data mining exercise. It has been found that the increase in the resolution

Figure 17. Net location identifiers for XYZ, Z = 2Y, Z = 3Y datasets

100000

80000

60000

40000

20000

0

No of X Coordinates

X=Y=Z Z=2Y Z=3Y

X=Y=Z, 27000,
17%

Z=3Y, 81000,
50%

Z=2Y, 54000,
33%