FP-MAXFLOW: An Algorithm for Mining Maximum Relevant Patterns

Arnaldo Díaz, Luciano García


Algorithms for itemsets recognition and classification have been widely studied. The state of the art shows that most of the algorithms obtain all possible itemsets, others only obtain maximal ones. Both approaches have limitations for getting relevant correlations. Obtaining all itemsets imply many irrelevant patterns. By obtaining only maximal patterns important information could be ignored. The goal of this work is to offer an algorithm for obtaining the patterns which more efficiently could describe the correlations. The new algorithm, called FP-MAXFLOW is able to extract these, information efficiently with one database scan. The comparative studies show that it is a competitive solution according to other algorithms which are among the most used.


Transactional databases, association mining, patterns recognition, frequent itemsets.

Full Text: PDF (Spanish)