FIMA: A New Fuzzy Item sets Miner Algorithm

Abstract

Frequent itemsets mining is the second step of Association rules mining which is the main task of Knowledge Discovery (KDD) and data mining (DM). There are three types of itemsets; Crisp (CI), Generalized (GI), and Fuzzy itemsets (FI). FI is the most recent type. This paper presents a new algorithm to mine fuzzy itemsets from large taxonomic databases depending on fuzzy taxonomies that reflect partial belongings among data items, also it depends on Item-Transaction layout, and shortest path finding between an item and its super classes. The proposed algorithm, Fuzzy Itemsets Miner Algorithm (FIMA) deals with the three types of fuzzy itemsets; taxonomic nodes, linguistic terms, and hedges. FIMA scans the database, under mining, only once. It excludes the need for complicated data structures, prunes the pruning steps of available algorithm, and avoids the weakness of manipulating low levels values of minimum support threshold. The algorithm performs much better than the available algorithm such that it reduces the complexity of mining FIs from exponential, O(an), to linear order of magnitude O(n).