Representing incomplete knowledge in case-based reasoning

Many intelligent computer systems rely on complex computational entities that represent entities and events in the real world. In CBR such entities are expressed via representational primitives called case features. Reflecting varying degrees of uncertainty, the values of case features are usually specified by both inherently imprecise symbolic and exact numeric expressions. In addition to intra-mental phenomena (human subjectivity), the stochastic nature of the physical world may have a bearing on the conception and representation of cases via case features. This work applies, tests, and evaluates a flexible knowledge representation mechanism, namely a case feature concept frame, which integrates both types of incomplete knowledge, possibilistic as well as probabilistic knowledge, into the CBR framework.