Efficient Mining of Temporally Annotated Sequences.

摘要: Sequential patterns mining received much attention in recent years, thanks to its various potential application domains. A large part of them represent data as collections time-stamped itemsets, e.g., customers’ purchases, logged web accesses, etc. Most approaches sequence focus on sequentiality data, using time-stamps only order items and, some cases, constrain the temporal gap between items. In this paper, we propose an efficient algorithm for computing (temporally-)annotated sequential patterns, i.e., where each transition is annotated with a typical time derived from source data. The adopts prefix-projection approach mine candidate sequences, and it tightly integrated annotation process that associates sequences annotations. pruning capabilities two steps sum together, yielding significant improvements performances, demonstrated by set experiments performed synthetic datasets.