Mining of high average-utility itemsets using novel list structure and pruning strategy

摘要: Abstract A novel algorithm for efficiently mining high average-utility itemsets is presented in this paper. The utilizes list structures, which compactly capture all information needed to calculate the actual average-utilities of so as mine without generation candidate itemsets. thus does not require any scanning a given transactional database after initial two scans constructing structures with 1-lengths. can generate through its depth-first search based process, conducted by recursively ( k + 1 ) -lengths from -lengths. In order avoid expansion unpromising that cannot be expanded itemsets, pruning technique using tight upper-bounds itemsets’ designed and applied algorithm. Therefore, runtime memory efficiencies are able enhanced significantly because space process considerably reduced. Various experiments were performed four real datasets groups synthetic datasets. Experimental results support proposed has runtime, memory, scalability performances superior those existing