Adaptive Stream Mining par Albert Bifet (Docteur de l'UPC-Barcelona et actuellement en post doc à l'Université de Waikato - Nouvelle Zélande)

Jeudi 17 décembre à 11h
LIRMM - salle 223

Résumé :
In the data stream model, data arrive at high speed, and the algorithms that must process them have very strict constraints of space and time. This talk presents new data mining algorithms for evolving data streams and for the extraction of closed frequent trees. It introduces a framework for developing algorithms that can adaptively learn from data streams that change over time. It presents an adaptive sliding window algorithm ADWIN for detecting change and keeping updated statistics from a data stream, and use it as a black-box in place or counters or accumulators in algorithms initially not designed for drifting data. It shows its application to decisions trees, ensemble classifiers, closed frequent tree mining and XML classification. Finally, it reviews MOA, the data stream mining software similar to WEKA developed at University of Waikato.