Conversational Case-Based Reasoning in Medical Classification and Diagnosis
In case-based reasoning (CBR) approaches to classification and diagnosis, a description of the problem to be solved is often assumed to be available in advance. Conversational CBR (CCBR) is a more interactive approach in which the system is expected to play an active role in the selection of relevant tests to help minimize the number of problem features that the user needs to provide. We present a new algorithm for CCBR called iNN(k) and demonstrate its ability to achieve high levels of accuracy on a selection of datasets related to medicine and health care, while often requiring the user to provide only a small subset of the problem features required by a standard k-NN classifier. Another important benefit of iNN(k) is a goal-driven approach to feature selection that enables a CCBR system to explain the relevance of any question it asks the user in terms of its current goal.
Keywordscase-based reasoning classification diagnosis accuracy feature selection transparency explanation
Unable to display preview. Download preview PDF.