Association Rule Mining using Path Systems in Directed Graphs
Subramanian Arumugam, S. Sabeen · International Journal of Computers Communications & Control · 2013
A transaction database (TDB) consists of a set $I$ of items and a multiset $\mathcal{D}$ of nonempty subsets of $I,$ whose elements are called transactions. There are several algorithms for solving the popular and computationally expensive task of association rule mining from a TDB. In this paper we propose a data structure which consists of a directed graph $D$ (loops and multiple arcs are permitted) and a system of directed paths in $D$ to represent a TDB. We give efficient algorithms for generating the data structure, for extracting frequent patterns and for association rule mining. We also propose several graph theoretic parameters which lead to a better understanding of the system.