A DISTRIBUTED ALGORITHM FOR MINING MULTILEVEL ASSOCIATION RULES

Association rule mining is an important task of data mining. In general an item is part of a hierarchy. More interesting rules can be found based on information taxonomy. The problem of multilevel association rule mining under distributed environment is studied, and a fast algorithm called MLFDM is proposed. Several techniques used in the algorithm are introduced, including encoded transaction table pruning, candidate iternsets generating and pruning, and the support count computing, etc. The principles and implementation and several variants of the algorithm are also discussed. The experiments prove the efficiency of these algorithms.