Parallel progressive sequential pattern (PPSP) mining

Abstract

Sequential pattern mining is one of the most important fields in data mining. The progressive sequential pattern mining problem explores the effect of deleting old data from the sequences in the database as well as adding new data to the sequences when sequential patterns are generated and hence generates the most accurate results in data mining applications. With the increasing amount of data, traditional algorithms running on uniprocessors have scalability troubles. We present a novel algorithm to optimize and scale the progressive mining of sequential patterns (PISA) problem on GPUs and achieve a speedup of more than 10× by efficiently utilizing coalesced memory accesses and SIMD execution on GPUs. © 2018 IEEE.