Inducing Diagnostic Inference Models from Case Data

Recent attention to using “case-based” reasoning for intelligent fault diagnosis has led to the development of very large, complex databases of diagnostic cases. The performance of case-based reasoners is dependent upon the size of the case base such that as case bases increase in size, it is usually reasonable to expect accuracy to improve but computational performance to degrade. Given one of these large case bases, it is advantageous to attempt to induce structure from the case base whereby the diagnostic process can be made more efficient. In addition, certainty properties of case based reasoning make fault diagnosis difficult, in which case inducing structure and applying a method for reasoning under uncertainty becomes advantageous. In this chapter, we discuss an approach to analyzing a diagnostic case base and inducing a compact knowledge base using the diagnostic inference model with which efficient diagnostics can be performed. We then apply an approach to reasoning using Dempster-Shafer theory to improve the diagnostics using the resultant model.