Discovery of cases for case-based reasoning in engineering

At present, an important assumption about case-based reasoning (CBR) is that cases should be readily available. While it is true in areas such as Law and Medicine where past cases are usually well documented, it does not hold for some engineering applications in which previous cases have not been established properly. We propose to tackle this "case acquisition" problem by using data mining and knowledge discovery techniques to discover cases from historical engineering information stored in data files or databases. The task of mining cases is clearly defined, and an algorithm is proposed to discover cases from raw data. Also the Dempster-Shafer rule for combining evidence from independent sources is introduced to fuse compatible cases and manage uncertainties. The work reported in this paper is part of an ongoing project to build a case-based system to identify transmission faults in power distribution networks.