Dynamic Load Balancing Algorithms for Sequence Mining
Valerie Guralnik, George Karypis · University of Minnesota Digital Conservancy (University of Minnesota) · 2001
Discovery of sequential patterns is becoming increasingly useful and essential in manyscientific and commercial domains. Enormous sizes of available datasets and possibly large number of mined patterns demand efficient and scalable algorithms. In this paper we present a parallel formulation of a serial sequential pattern discovery algorithm based on tree projection that uses a novel dynamic load balancing algorithm whichiswell suited for distributed memory parallel computers. Our experimental evaluation on a 32 processor IBM SP show that this algorithms are capable of achieving good speedups, substantially reducing the amount of the required work to find sequential patterns in large databases.