Abstract:
Traditional association rules mining algorithm
depends on two user-specified thresholds. One is the
minimum support, which is statistically required. Another
is the minimum confidence. Users or even experts are
difficult to specify this confidence value. The association
rules generated by using this classical confidence threshold
are not always interesting. We propose a transformation
from this traditional confidence to a so-called positive -
confidence. Users only need to give a database independent
threshold of confidence to mine the
association rules from databases, in despite of the details of
the databases. Experiments are conducted, showing that
our proposed method is useful for the automation of the
association rule mining systems, and the algorithm given
in this paper is effective.
