Abstract

We will consider set-covering models for diagnostic tasks handling uncertain
knowledge.Because these models have to be build manually by domain experts,we
will show how this effort can be reduced by an incremental development of the
set covering models.Thus simple models can be enhanced by similarities,weights
and uncertainty to increase the quality of the knowledge and the resulting
system.We will also present mechanisms for generating set covering models when
higher level knowledge about causation effects is available.  Finally, we will
motivate how our approach can be used for implementing diagnosis systems
including therapy effects.  
Keywords:  set-covering model; model-based diagnosis; abductive reasoning; applied uncertainty; applied causality