In this work two indexing approaches are presented for case-based reasoning. The first is a hybrid technique which uses a combination of a matrix structure and a tree structure to solve problems. The matrix and tree structures index cases by their discre-tised feature values. The second approach is based solely on the tree structure and never uses the matrix. The two techniques are evaluated in terms of their competency and efficiency with respect to nearest neighbor retrieval. Both approaches provide average efficiency gains of up to 20 fold in comparison to nearest neighbour with only a slight loss in competency, as averaged across all case-bases tested. It is argued that these approaches are appealing due to their simplicity, competency and efficiency.
展开▼
