{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/mcdiarmid-drift-detection-methods-for","title":"McDiarmid Drift Detection Methods for Evolving Data Streams","arxiv_id":"1710.02030","date":"2017-10-05","proceeding":null,"authors":["Ali Pesaranghader","Herna Viktor","Eric Paquet"],"abstract":"Increasingly, Internet of Things (IoT) domains, such as sensor networks,\nsmart cities, and social networks, generate vast amounts of data. Such data are\nnot only unbounded and rapidly evolving. Rather, the content thereof\ndynamically evolves over time, often in unforeseen ways. These variations are\ndue to so-called concept drifts, caused by changes in the underlying data\ngeneration mechanisms. In a classification setting, concept drift causes the\npreviously learned models to become inaccurate, unsafe and even unusable.\nAccordingly, concept drifts need to be detected, and handled, as soon as\npossible. In medical applications and emergency response settings, for example,\nchange in behaviours should be detected in near real-time, to avoid potential\nloss of life. To this end, we introduce the McDiarmid Drift Detection Method\n(MDDM), which utilizes McDiarmid's inequality in order to detect concept drift.\nThe MDDM approach proceeds by sliding a window over prediction results, and\nassociate window entries with weights. Higher weights are assigned to the most\nrecent entries, in order to emphasize their importance. As instances are\nprocessed, the detection algorithm compares a weighted mean of elements inside\nthe sliding window with the maximum weighted mean observed so far. A\nsignificant difference between the two weighted means, upper-bounded by the\nMcDiarmid inequality, implies a concept drift. Our extensive experimentation\nagainst synthetic and real-world data streams show that our novel method\noutperforms the state-of-the-art. Specifically, MDDM yields shorter detection\ndelays as well as lower false negative rates, while maintaining high\nclassification accuracies.","url_abs":"http://arxiv.org/abs/1710.02030v2","url_pdf":"http://arxiv.org/pdf/1710.02030v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"mcdiarmid-drift-detection-methods-for","repo_url":"https://github.com/alipsgh/data_streams","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"mcdiarmid-drift-detection-methods-for","repo_url":"https://github.com/alipsgh/codes-for-moa","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"drift-detection","task_name":"Drift Detection"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1710.02030","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}