A Precise Statistical approach for concept change detection in unlabeled data streams

A Precise Statistical approach for concept change detection in unlabeled data streams

0.00 Avg rating0 Votes
Article ID: iaor20118213
Volume: 62
Issue: 4
Start Page Number: 1655
End Page Number: 1669
Publication Date: Aug 2011
Journal: Computers and Mathematics with Applications
Authors: , ,
Keywords: datamining, networks
Abstract:

• We present a statistical approach for change detection in unlabeled data stream. • In our approach, upon arrival of new data point, a hypothesis test takes place. • It is driven by a family of martingales which is based on Doob’s Maximal Inequality. • Our approach detects changes in all domains of categorical, numerical and mixed.

Reviews

Required fields are marked *. Your email address will not be published.