Data mining techniques enable companies to explore meaningful patterns and rules from an enormous database. Association rule mining has been widely used. Many other relevant methods and researches have been published. However, most of the algorithms simply deal with the single level of association rules. In this paper, a new method "The FP-tree Algorithm for Mining the Multi-level Characteristic Rule" is proposed. The purpose of this study is to investigate the highly related association rules among customers’ characteristics and goods. It is expected that the FP-tree algorithm helps the users obtain meaningful association rules and put the data mining techniques into practice.