A Graph-Based Ant Colony Optimization for Association Rule Mining

Abstract

This study is aimed at proposing a graph-based ant colony optimization (ACO) approach for association rule mining (ARM). The ACO-ARM is a two-phase approach comprising a Boolean transactional data representation scheme and the graph-based ACO. The first phase enhances the normal Apriori algorithm and engages in a data representation scheme. The data representation involves an adapted Boolean matrix representation of the transactional data. A standard Apriori algorithm is applied to the represented data, and n-frequent itemsets are generated. The second phase embellishes the ACO-ARM, which relies on the graph of 2-frequent items to generate the final frequent itemset. We have conducted two experiments. The outcomes of these tests reveal that the graph-based ACO-ARM enhances execution time compared to the standard Apriori algorithm. In addition, ACO-ARM improves the process of data representation in the Apriori algorithm.