Web news articles are generated in continuous, time-varying, and rapid modes. This environment causes an explosion of information which needs to be stored, processed
and analyzed. Conventional machine learning algorithms that are applied in the web news mining work in an offline environment cannot efficiently handle data streams.
In this paper, we propose an evolving web news mining framework based on the recently published Evolving Type-2 Classifier (eT2Class). The eT2Class adopts an
open structure that can be used in non-stationary environments and works on a single pass learning mode that is applicable for online real-time applications. The
effectiveness of our evolving web news mining techniques is numerically validated and compared against state-of-the-art algorithms. The efficacy of our methodology
has been numerically validated with real local Australian news articles, namely the Age, spanning from 26/2/2016 to 13/3/2016 and has been compared with 6 state of
the art algorithms. Our algorithm outperforms other consolidated algorithms and achieves a tradeoff between complexity and accuracy with almost 10% improvement
in term of complexity.