BEGIN:VCALENDAR VERSION:2.0 PRODID:-//TYPO3/NONSGML Calendarize//EN BEGIN:VEVENT UID:calendarize-concept-drift-detection-in-unlabeled-data-streams DTSTAMP:20260413T114427Z DTSTART:20260427T080000Z SUMMARY:Concept Drift Detection in Unlabelled Data Streams DESCRIPTION:On Monday\, 27 April 2026\, at 10:00 a.m.\,Daniel Lukats Unive rsity of Oldenburgwill give a lecture entitledConcept Drift Detection in U nlabelled Data StreamsThe lecture will take place online: https://studconf .uol.de/rooms/mpz-niv-bz5-40a/join The lecture will be held in English.\nA bstract:In long-running data streams\, an issue called concept drift may a rise. Concept drift denotes changes in the feature distribution P(X) or th e posterior distribution P(Y | X)\, such that the performance of classifie rs employed on the data stream may deteriorate severely. Besides detecting concept drifts for the adaptation of classifiers\, concept drifts themsel ves may be worth detecting\, e.g.\, in marine domains where changes such a s ocean fronts or algal blooms can be interpreted as concept drifts which merit further action.Although various concept drift detectors have been su ggested in the literature\, most detectors are supervised: They require an d assume the availability of the true label immediately after predicting i t. This assumption is hard to justify\, since labelling a potentially infi nite data stream in real time can become prohibitively expensive.While a n umber of unsupervised approaches to concept drift detection exist\, few ha ve been demonstrated on real-world applications. Instead\, these methods a re commonly evaluated on artificially created problems without ground trut h information about concept drift. Thus\, proxy metrics are commonly used in the literature. However\, these proxies are flawed\, as they often rewa rd frequent false positives more than correct detection.Based on this inte rpretation of the state of the art\, this work discusses the following res earch questions: How can unsupervised concept drift detectors detect conce pt drift on continuous features? How can the predictive quality and the ti meliness of concept drift detectors be evaluated in an unbiased fashion?To this end\, existing metrics such as the F1 Score are adapted for use in c oncept drift detection. Then a system encompassing an unsupervised concept drift detector and optional components to address common real-world prope rties such as seasonal data\, trend and noisy features is evaluated. These research questions are discussed on synthetic data streams and in the con text of oceanographic data streams with a case study on the detection of a lgal blooms in the Baltic Sea.Supervisor: Prof. Dr.-Ing. Axel Hahn X-ALT-DESC;FMTTYPE=text/html:

On Monday\, 27 April 2 026\, at 10:00 a.m.\,
Daniel Lukats University of Oldenburg
will give a lecture entitled
Concept Drift Detection in Unlabelled Data S treams
The lecture will take place online: https://studconf.uol.de/rooms/mpz-niv-b z5-40a/join
The lecture will be held in English.

\n

Abstract:
In long-running data streams\, an issue called co ncept drift may arise. Concept drift denotes changes in the feature distri bution P(X) or the posterior distribution P(Y | X)\, such that the perform ance of classifiers employed on the data stream may deteriorate severely. Besides detecting concept drifts for the adaptation of classifiers\, conce pt drifts themselves may be worth detecting\, e.g.\, in marine domains whe re changes such as ocean fronts or algal blooms can be interpreted as conc ept drifts which merit further action.
Although various concept drift detectors have been suggested in the literature\, most detectors are supe rvised: They require and assume the availability of the true label immedia tely after predicting it. This assumption is hard to justify\, since label ling a potentially infinite data stream in real time can become prohibitiv ely expensive.
While a number of unsupervised approaches to concept d rift detection exist\, few have been demonstrated on real-world applicatio ns. Instead\, these methods are commonly evaluated on artificially created problems without ground truth information about concept drift. Thus\, pro xy metrics are commonly used in the literature. However\, these proxies ar e flawed\, as they often reward frequent false positives more than correct detection.
Based on this interpretation of the state of the art\, th is work discusses the following research questions: How can unsupervised c oncept drift detectors detect concept drift on continuous features? How ca n the predictive quality and the timeliness of concept drift detectors be evaluated in an unbiased fashion?
To this end\, existing metrics such as the F1 Score are adapted for use in concept drift detection. Then a sy stem encompassing an unsupervised concept drift detector and optional comp onents to address common real-world properties such as seasonal data\, tre nd and noisy features is evaluated. These research questions are discussed on synthetic data streams and in the context of oceanographic data stream s with a case study on the detection of algal blooms in the Baltic Sea.
Supervisor: Prof. Dr.-Ing. Axel Hahn

LOCATION: END:VEVENT BEGIN:VEVENT UID:calendarize-departmentrat DTSTAMP:20260316T162603Z DTSTART:20260429T110000Z DTEND:20260429T120000Z SUMMARY:Department council DESCRIPTION: X-ALT-DESC;FMTTYPE=text/html: LOCATION:Meeting room in StudIP event 2.01.888 DPR Computing Science (http s://elearning.uni-oldenburg.de/dispatch.php/course/details?sem_id=e0180451 c9ee9edb1c3ba4fd34e5fbc6&sso=cas&cancel_login=1&ticket=ST-4524636-MY3KWgBy BQtaitZAKTcwvr3fwDUcas05) END:VEVENT END:VCALENDAR