A Fast Scalable Algorithm to Build Closed Itemsets from Large Data

H. Fu and E.M. Nguifo (France)


Concept lattice, Closure operator, Closed itemset, Scalabil ity, Data mining, Partitionning


Large data still bothers us in the field of machine learn ing and data mining. The mining of very large database still needs more efficient algorithm. Concept lattice is an effective tool for data analysis and knowledge discovery. Many research works in classification or association rules increase the interest of Concept lattices structures for data mining. Several algorithms were proposed to generate con cepts of lattices, among which NextClosure algorithm is the best one for large context . However, it's also hard to face the complexity of large data with NextClosure algo rithm. So we propose a new efficient scalable algorithm: ScalingNextClosure which creates partitions within data and builds independently closed itemsets in each partition. There is no other communication among each partition, so that ScalingNextClosure algorithm can be easily par allelized. The experimental results show that the algorithm has efficient performance compared to NextClosure algo rithm.

Important Links:

Go Back