Mining Matrix Pattern from Mobile Users - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 257

Comparison with Previous r elated Works
Figures 22 and 23 show the comparison between the capabilities of frequency pattern and matrix pattern
(both single layer and multi-layer). Figure 22 shows that matrix pattern has the advantage over other
methods in terms of having minimal overhead requirements. Figure 23 shows the comparison among the
different possible paths for mobile user data mining that are using single layer matrix pattern, multiple
layer matrix pattern, or group pattern. It shows that, in order to fully utilize the resources available, it
must be decided with caution whether to use a single layer matrix or multi-layer matrix. Single layer
matrix is sufficient for decisions that involve only one domain of activity, such as shoe producers. How-
ever, if a decision needs to be made that involves multiple domains, such as the relationship among shoe
makers and socks makers, then the multi-layer matrix pattern would be suitable to perform the job.
conclusion and f uture work
In conclusion, it is found that matrix pattern is able to provide mobile user data mining in a cost efficient
way, by changing the fundamental method by which knowledge is mined. By doing so, the knowledge
found previously will be mobile user dependent, but for matrix pattern it will be location dependent.
The fact that the knowledge is location dependent does not necessarily mean that the knowledge is only
related to the particular physical location. As physical locations can be named with logical themes, the
patterns essentially signify the relationships among the logical themes.
Future work in the mobile user data mining area includes the development of cost models to be used
for the evaluation of each mobile user data mining model. Different cost models have to be developed
in order to cater to different organisational view of costing, including economic costing, accounting
costing, and technical costing. Furthermore, a behavioural model of mobile users is also necessary for
the better implementation of algorithms in order to seek knowledge about behaviour of mobile users.
r eferences
Agrawal, R., & Srikant, R. (1994). Fast algorithms for mining association rules. In Proceedings of the
20th International Conference on Very Large Data Bases (pp. 487-499).
Agrawal, R., & Srikant, R. (1995). Mining sequential patterns. In Proceedings of the 11th International
Conference on Data Engineering (pp. 3-14).
Ashcraft, C., & Liu, J. (1998). Robust ordering of sparse matrices using Multisection. SIAM Journal on
Matrix Analysis and Applications, 19 (3), 816-832.
Barbar'a, D., Chen, P., & Nazeri, Z. (2004). Self-similar mining of time association rules. In 8th Pacific
Asia Conference on Knowledge Discovery and Data Mining, (Lecture Notes in Artificial Intelligence,
3056, pp. 86-95).
Chen, S. Y. & Liu, X. (2005). Data mining from 1994 to 2004: An application-oriented review. Inter-
national Journal of Business Intelligence and Data Mining, 1 (1), 4-21.