With the growth of various types of data, mining useful association rules from large databases has been an important research topic nowadays. Previous works focus on the attributes of data to derive a variety of association rules. In this paper, we first organize the data as a multiple-attribute hierarchical tree by the attributes of transactions to efficiently derive the multiple-attribute association rules. Furthermore, we store the derived rules as a frequent hierarchical tree and allow users to specify various types of queries for finding interesting correlations named phenomena among the rules. We finally perform a series of experiments to evaluate the performance of our approach.