Abstract:
This paper describes a distributed system of intelligent
agents, MARA (Mining Association Rule Agents),
for performing association rules mining over databases on
behalf of a community of users. MARA can not only identify
association rules in databases, but also refine the mined
rules by the information sharing among users with similar
interests automatically. MARA provides agents which
can collect outer information from such as Web, journals,
and news medium. And then the collecting information is
also used to refine the mined association rules for the purpose
of mining very small databases. For example, nuclear
power plants and earthquake bureaus have only some very
small databases. Apparently, the data in small databases
may not be large enough to form any meaningful patterns.
We explore a functional method for these databases. We
evaluate the proposed techniques in the system, and our experimental
results demonstrate that the system is efficient
and promising.
