AFARTICA: A Frequent Item-Set Mining Method Using Artificial Cell Division Algorithm

Introduction

Frequent item-set mining is considered to be one of the fundamental tasks of data mining apart from other frequent problems like the discovery of association rules, correlation mining, multidimensional pattern mining, etc. The essential problem of mining frequent item-sets is based on finding the frequent item-sets from a given set of items that occurs over a given threshold amount of times. The problem is challenging as it is to be performed over a large database in real time.

The target of data mining is to identify patterns in data by using various tools and algorithms (Rajagopalan & Krovi, 2002). Most of the data-mining algorithms designed to solve the problem is built upon the idea of Apriori algorithm (Agrawal et al., 1996). This algorithm uses a breadth first search strategy starting from the bottom of the search space to the top in order to find every possible frequent item-set. An enumeration of all possible subsets of length of the frequent pattern is found to be computationally hard in case of dense data. Recent research trends have focused on finding the maximal frequent item-sets to tackle the problems. The present work has been greatly influenced by the methodology of this algorithm. Zaki (Zaki, 2000; Zaki & Hsiao, 1999) carried experiments using both real and synthetic data with new efficient rule mining framework on closed set. Generation of rules only on demand and the method of reduction of redundant rules are found to be interesting.