An Incremental Learning Approach Using Long Short-Term Memory Neural Networks

Due to Big Data and the Internet of Things, machine learning algorithms targeted specifically to model evolving data streams have gained attention from both academia and industry. Although most of the proposed solutions in the literature have reported being successful in learning from non-stationary streaming settings, their complexity and the need for extra resources may be a constraint for their deployment in real applications. Aiming at less complexity without losing performance, this article proposes an incremental variant of the original LSTM neural networks with minor changes, that can tackle evolving data stream problems such as concept drift and the elasticity–plasticity dilemma without neither needing a dedicated drift detector nor a memory management system. Results achieved from benchmark datasets have shown that the proposed method is competitive in comparison with other incremental methods from the literature.