Abstract
Mining association rules is one of the most well studied problems in data mining. Current algorithms for finding association rules require several passes over the databases, and obviously the role of I/O overhead is significant for very large databases. In this paper, we present MARC (Mining Association Rules using Clustering), a new algorithm that makes only one full pass over the database. Firstly, we partition the collection of transactions so that similar transactions fall into the same cluster. Then we mine association rules on the summaries of clusters instead of the entire data set. Consequently, a proper method for summarizing a cluster of transactions is proposed. The results of experiments show that the proposed algorithm can learn association rules efficiently in single database pass, and also show that MARC algorithm does not affect too much the accuracy of the association rules learned.