Case based reasoning based on fuzzy rough set

This paper presents a new attribute significance measurement method based on the viewpoint of information theory of Fuzzy Rough Set. The method does not need to discretize continuous attributes included in cases, which can effectively reduce information loss, and it is suitable for the cases which contain both discrete and continuous attributes. Then it proposes a heuristic algorithm based on the new attribute significance for attribute reduction and weight distribution, and combines with the Nearest Neighbour (NN) method in case retrieve process. Finally, an example shows that this method improves accuracy and universality of the case retrieval.