PLT- Positional Lexicographic Tree: A New Structure for Mining Frequent Itemsets

Abstract

Association rules have proved their influence in different industrial fields, where their goal is to identify the relations existing among the events that are stored in large databases. However, in order to enumerate the association rules, there is a need to identify the frequent set of itemsets (i.e. those events that occur together in a sufficient number of transactions). In this paper, a new representation structure for the data stored in any transactional database is proposed. This structure, which we refer to as Positional Lexicographic Tree (PLT), provides an efficient mechanism for subset checking based on a summary of the data extracted from the database. This makes PLT a promising tool for most of the existing data mining approaches. Moreover, our proposed PLT structure regulates the data in the database so that they can be applicable to compression and indexing techniques, which makes PLT suitable for supporting large databases. First, we introduce the PLT construction process, then highlight the different mining approaches that can be modulated to take advantage of PLT. We then present our algorithm and finally prove its correctness.