Efficient theta-subsumption of sets of patterns
Jan Ramon, Jan Struyf · Lirias · 2004
Frequent pattern discovery is an important data mining task. Recently, there was an increasing amount of interest in relational frequent pattern mining. In previous work we discussed how one can speed up the matching of relational patterns against the examples. In this paper we present new optimizations that speed up the pattern generation phase. These new optimizations are based on a representation of the patterns in a tree structure and exploit this structure in the various tests that are necessary during pattern generation. We discuss some theoretical properties and present a number of experiments.