Frequent Pattern Mining using Bipartite Graph
Duck Jin Chai, Long Jin, Bu-Hyun Hwang, Keun Ho Ryu · 2007
In this paper, we propose an efficient ALIB algorithm that can find frequent patterns by only onetime database scan. Frequent patterns are found without generation of candidate sets using LIB-graph. LIB-graph is generated simultaneously when the database is scanned for 1-frequent items generation. LIB-graph represents the relation between 1-frequent items and transactions including the 1-frequent items. That is, LIB-graph compresses database information into a much smaller data structure. We can quickly find frequent patterns because the proposed method conducts only onetime database scan and avoids the generation of candidate sets. Our performance study shows that the ALIB algorithm is efficient for mining frequent patterns, and is faster than the FP-growth.