Feature subset selection for data and feature streams: a review

Abstract Real-world problems are commonly characterized by a high feature dimensionality, which hinders the modelling and descriptive analysis of data. However, some these data may be irrelevant or redundant for learning process. Different approaches can used to reduce this information, improving not only speed building models but also their performance interpretability. In review, we focus on subset selection (FSS) techniques, select original set without making any transformation attributes. Traditional batch FSS algorithms adequate efficiently handle large volumes data, either because memory arise received in sequential manner. Thus, article aims survey state art incremental algorithms, perform more under circumstances. strategies described, such as incrementally updating weights, applying information theory using rough set-based FSS, well multiple supervised unsupervised tasks where application is interesting.