CURIE: a cellular automaton for concept drift detection

Data stream mining extracts information from large quantities of data flowing fast and continuously (data streams). They are usually affected by changes in the distribution, giving rise to a phenomenon referred as concept drift. Thus, learning models must detect adapt such changes, so exhibit good predictive performance after drift has occurred. In this regard, development effective detection algorithms becomes key factor mining. work we propose $$\textit{CURIE}$$ , detector relying on cellular automata. Specifically, distribution is represented grid automata, whose neighborhood rule can then be utilized possible over stream. Computer simulations presented discussed show that when hybridized with other base learners, renders competitive behavior terms metrics classification accuracy. compared well-established detectors synthetic datasets varying characteristics.