Optimization of the knowledge base of a cognitive quality of transmission estimator for core optical networks

We have recently proposed a cognitive Quality of Transmission (QoT) Estimator for classifying lightpaths into high or low quality categories. In this paper, we enhance that work by incorporating learning and forgetting techniques with the aim of optimizing the underlying knowledge base (KB) on which the cognitive estimator relies. We demonstrate that by including these techniques, a higher percentage of successful classifications is obtained (which is higher than 99%), and moreover, that it also leads to a significant reduction on the computing time for on-line operation. In particular, the cognitive estimator, when relying on an optimized KB, is around one order of magnitude faster than when it does not use an optimized KB, and, moreover, it is around four orders of magnitude faster when compared with a different existing approach, the Q-Tool.