Clustering Based Multi-Objective Rule Mining using Genetic Algorithm

Multi-Objective Genetic Algorithm (MOGA) is a new approach for association rule mining in the market-basket type databases. Finding the frequent itemsets is the most resource-consuming phase in association rule mining, and always does some extra comparisons against the whole database. This paper proposes a new algorithm, Cluster-Based MultiObjective Genetic Algorithm (CBMOGA) which optimizes the support counting phase by clustering the database. Clusters are based on the number of items in each transaction. Experiments on two different marketbasket type databases show that the CBMOGA outperforms the MOGA. However, the speedup highly depends on the distribution of transactions in the cluster tables. Hence, the optimization ratio is datasetdependent.