Abstract:

Compared with the classic frequent pattern mining (FPM) algorithms, the dynamic FPM algorithms on fast and massive data streams have become top research nowadays. A new batch mining algorithm in data streams called stream frequent closed labeled tree miner (SFCLTreeMiner) is proposed. SFCLTreeMiner uses a kind of addingremoving method between closed tree sets. Also it provides a time decay module for reasonable data updating. Experiment shows that SFCLTreeMiner is efficient in data streams mining by reducing consuming dramatically.

Key words: frequent subtree mining, data streams, time decay, labeled closed subtree, batch mining