AFOPT-Tax: An Efficient Method for Mining Generalized Frequent Itemsets

The wide existence of taxonomic structures among the attributes of database makes mining generalized association rules an important task. Determining how to utilize the characteristics of the taxonomic structures to improve performance of mining generalized association rules is challenging work. This paper proposes a new algorithm called AFOPT-tax for mining generalized association rules. It projects the transaction database to a compact structure - ascending frequency ordered prefix tree (AFOPT) with a series of optimization, which reduces the high cost of database scan and frequent itemsets generation greatly. The experiments with synthetic datasets show that our method significantly outperforms both the classic Apriori based algorithms and the current FP-Growth based algorithms.