Data mining is a process or method of finding information, evidence, insight, knowledge and hypotheses in a huge database, such as marketing data. Recently, the association rule presented by R. Agrawal in 1983 has been used to rapidly expand a data mining method. This method is general and flexible and can be applied to both general data analysis and very wide surveys. In addition, the rules for this method are complicated. On the other hand, when the support value is minimal and the confidence value is high, the obtained value is already known and trivial. A breakthrough method is needed. The objective of this paper is to present a rough set model to overcome such issues. Employing the rough set model, we analyzed three different scales of databases and compared the results of simulation experiments using proposed and conventional models. The rough set model obtained an efficient number of association rules and usually took less computation time.