Mining Approximate Frequent Patterns from Noisy Databases

Abstract

As an important branch in the field of frequent pattern mining, approximate frequent pattern (AFP) mining attracts much attention recently. Various algorithms have been proposed to discover long true AFPs in presence of random noise. This paper considers the key issues of AFP mining in noisy databases, and categorizes the previous approaches according to the ways they cope with missing items in the transactions. And then a study of different data models on AFP is presented, in which the merits and defects are analyzed. Finally, we draw a conclusion and propose some solutions to deal with the problems in the field of AFP mining.