Intelligent Failure Domain Prediction in Complex Telecommunication Networks with Hybrid Rough Sets and Adaptive Neural Nets

Automated fault forecasting proactivity offers a promising closed-loop approach for conventional network failure management activities. Automated intelligent failure forcasting requires the capability to prefilter the observation data so as to remove irrelevant features or factors from multi-dimensional observation data. In this paper, we propose a new hybrid methodology of combining the rough-set logics with artificial neural networks for failure domain exploration in this complex system. Rough-set logics is the best solution to meet these requirements. It acts as a preprocessing machine to reduce the feature domain with a minimal factors sets being generated. In comparison with the conventional neural network approach, the Rough Neural Network Adaptive Information Prediction System (RNNAIPS) makes more efficient shortperiodical prediction for the failures domain of the internet. The simulation results show that the prediction accuracy and convergency are both greatly improved by applying this integration in comparison with the application using mere neural network.