Constructing Bayesian Belief Networks for Fault Management in Telecommunications Systems
Roy Sterritt, Weiru Liu · Research Portal (Queen's University Belfast) · 2001
This paper discusses the learning of Bayesian Belief Networks (BBN), probabilistic models of a system in which the independence relations between the variables of interest are represented explicitly, for high-speed telecommunications network alarm log data. Specifically the data is from the Synchronous Digital Hierarchy (SDH) Element Controller (EC-1) yet the issues of complexity and uncertainty under fault conditions and how BBNs can possibly be used to address them are generic. An example is provided using the PowerConstructor software developed by Cheng et al. (1997) and used for the induction of a BBN which represents a series of ten alarms in a SDH network.