Abstract: It is necessary to solve the two problems before applying association rule mining to affective design: firstly， how to identify proper parameters (the support and confidence thresholds) for association rule mining; secondly， how to find out useful information from a mass of association rules generated by association rule mining. In order to solve the two problems described above， a method which could refine association rules effectively is presented: at first， a set of raw rules are generated by specif ying low values for the support and confidence thresholds; then， these raw rules are evaluated to refine the most meaningful rules， in this step， a series of concrete sub-steps is presented. A case study was conducted to illustrate the proposed method.