Unsupervised Rule Generation for Maintenance of a Diagnostic System

V. Ramani

A common limitation of diagnostic systems is their dependence on the training data. It is essential that the diagnostic system should be maintainable over time. In most practical applications the data used for training a diagnostic system does not cover the entire spectrum of faults that a system could encounter. Thus the system should be able to generate rules for new "unknown" faults. An effective methodology for a minimally supervised diagnostic system is developed and explained in this paper. Fuzzy logic is used for automatic generation and evolution of rules, based on the extracted feature set, for detection and identification of new faults.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.