In order to obtain valuable knowledge from stored data on database systems, rule mining is considered as one of the usable mining method. However, almost current rule mining algorithms only use primary difference of a criterion to select attribute-value pairs to obtain a rule set to a given dataset. In this paper, we implemented a rule generation method based on secondary differences of two criteria. Then, we performed a case study using UCI common datasets. With regarding to the result, we compared the accuracies of rule sets learned by our algorithm with that of three representative algorithms.