Incremental Learning Framework for Mining Big Data Stream

At this current time, data stream classification plays a key role in big analytics due to its enormous growth. Most of the existing methods used ensemble learning, which is trustworthy but these are not effective face issues learning from imbalanced data, it also supposes that all pre-classified. Another weakness takes long evaluation time when target contains high number features. The main objective research develop new method for incremental based on proposed ant lion fuzzy-generative adversarial network model. model implemented spark architecture. For each stream, class output computed at slave nodes by training generative with back propagation error fuzzy bound computation. This overcomes limitations as can classify streams slightly or completely unlabeled and providing scalability efficiency. results show outperforms state-of-the-art performance terms accuracy (0.861) precision (0.9328) minimal MSE (0.0416).