Learning belief networks from data: an information theory based approach
CIKM '97 Proceedings of the sixth international conference on Information and knowledge management
Causality: models, reasoning, and inference
Causality: models, reasoning, and inference
Introduction to Bayesian Networks
Introduction to Bayesian Networks
Investigating Temporal Patterns of Fault Behaviour within Large Telephony Networks
IDA '01 Proceedings of the 4th International Conference on Advances in Intelligent Data Analysis
Hi-index | 0.00 |
With delivery of higher quality broadband services to residential customers via the copper access network, physical fitness of the plant is vital for meeting service obligations. Consequently, we need to quickly identify any poorly performing plant within what has become an aging geographically heterogeneous network. Intuitively simple approaches based on fault counts or electrical measurements may seem effective. Although they can lead to initial quick wins by picking out the worst, they soon become ineffective, failing to identify the ‘next worse’ cases. We clarify the problem, and then use a Bayesian network to construct a causal structure for fault propagation. The probabilistic structure is learnt without supervision to provide a unique view of the fault process for each item of plant. This can then be used to identify vulnerable plant and infers the causal mechanism for their respective susceptibility. Apart from helping to determine where the money should be spent, it also provides a straightforward way of estimating important cost aspects relating to programmes of improvement and quality reforms.