Journal of Computational and Theoretical Nanoscience   2016 年 13 卷 7 期

2016.07.01

This paper proposes a novel case-based retrieval measure by using formal concept analysis. The proposed method employs the concept lattice to express the knowledge base of cases. The proposed retrieval method combines semantic, featural, and structural information into a decision. In this measure, we use the information content approach to obtain automatically the semantic part of the similarity scores of two concepts that constitute the traditional feature and structural similarityevaluating measures. Furthermore, the similarity for case retrieval can be calculated from the lower approximations on the basis of the rough set. Experiment results on the four UCI datasets show that our proposed similarity measure provides better accuracy than some existing similarity measures.