Knowledge network based association discovery (Preliminary Version)

摘要

Association mining is an important field in data mining. In this paper, we present a knoweldge network approach to association mining which utilizes the active items and their links information stored in the network to facilitate fast association discovery. In this approach, frequent items and their links are stored in a one-layer knowledge network after scanning the database once. All potential frequent itemsets can be obtained from the activated network, and stored in a prefix tree which failitates the identification of free quent itemsets. The experimental results show that it is an efficient approach with low I/O cost and suitable for various mining goals.