Adaptive Quick Reduct for Feature Drift Detection

Data streams are ubiquitous and related to the proliferation of low-cost mobile devices, sensors, wireless networks Internet Things. While it is well known that complex phenomena not stationary exhibit a concept drift when observed for sufficiently long time, relatively few studies have addressed problem feature drift. In this paper, variation QuickReduct algorithm suitable process data proposed tested: builds an evolving reduct dynamically selects relevant features in stream, removing redundant ones adding newly as soon they become such. Tests on five publicly available datasets with artificially injected confirmed effectiveness method.