Exploring interesting multi-level patterns using graph based approach

Abstract

Most of the previous studies on mining multi-level interesting patterns based on an Apriori approach required more number of databases scans and operations for counting pattern supports in the database. In this paper, we have focused on reducing database scans and avoiding candidate generation for extracting multilevel patterns. To achieve this objective a single level graph based approach has been used with a top-down progressive deepening method. At each concept level the whole database is compressed by converting into a directed graph which is stored in the form of an Adjacency Matrix. Further frequent pattern mining is done by performing operation on Adjacency Matrix of directed graph. The advantage of this method is that it requires only single scan of the database at each concept level for mining interesting patterns. Copyright � 2009 by IICAI.