nonordfp: an FP-growth variation without rebuilding the FP-tree
Balázs Rácz · SZTAKI Publication Repository (Hungarian Academy of Sciences) · 2004
We describe a frequent itemset mining algorithm and implementation based on the well-known algorithm FPgrowth. The theoretical difference is the main data structure (tree), which is more compact and which we do not need to rebuild for each conditional step. We thoroughly deal with implementation issues, data structures, memory layout, I/O and library functions we use to achieve comparable performance as the best implementations of the 1st Frequent Itemset Mining Implementations (FIMI) Workshop.