Abstract: The authors, from the view of relational data mining (RDM), propose a new method to excavate peculiar rules, which recognizes the peculiar data based on attribute-oriented method, and uses Chiz algorithm for reference to discretize the peculiar data and qualitatively describe the peculiar degree of the data. The new method, adopting the classical inductive logic programming design system FOIL, naturally mines the peculiar rules, breaking through the frame of traditional prepositional data mining. The experiment results show that the method can discover and mine the valuable knowledge ignored by the traditional mining algorithm of association rules.