Abstract: For the problems of poor fault-tolerant ability,small sample size and inconsistent decision results under equal conditions in information systems,an algorithm of extracting multi-hierarchical decision rules in attribute-oriented induction is proposed based on the variable precision rough set model.In the concept hierarchies of condition attributes,the relationships of the low approximations,positive regions,boundary regions and negative regions in the variable precision model between the high level decision table and its corresponding low level one are analyzed firstly.Based on the relationship diagram of different-level decision tables,the certain decision rules are obtained from the highest-level decision table to the bottom one using the traditional rough set model.Then more abstract decision rules are acquired from bottom to top along the relationship diagram based on the lowest-level decision table.The feasibility of the algorithm is validated by a case study.