2017

encyclopaedic entry

english

EN Compared to batch learning from static data, constructing classifiers from data streams implies new requirements for algorithms, such as constraints on memory usage, restricted processing time, and one scan of incoming examples. Additionally, streams classifiers have to adapt to concept drifts. The entry discusses the following stream classification issues: data stream specific requirements, processing schemes, categorization of concept drifts, classifier evaluation criteria and procedures, forgetting mechanisms, change detection methods, main algorithms for supervised learning of single classifiers and ensembles, open problems, areas of application.

1191 - 1199

10.1007/978-1-4899-7687-1_908

https://link.springer.com/referenceworkentry/10.1007/978-1-4899-7687-1_908

Encyclopedia of Machine Learning and Data Mining