Maintaining the Maximum Normalized Mean and Applications in Data Stream Mining
ADMA '08 Proceedings of the 4th international conference on Advanced Data Mining and Applications
Supporting Customer Retention through Real-Time Monitoring of Individual Web Usage
ADMA '08 Proceedings of the 4th international conference on Advanced Data Mining and Applications
StreamKrimp: Detecting Change in Data Streams
ECML PKDD '08 Proceedings of the 2008 European Conference on Machine Learning and Knowledge Discovery in Databases - Part I
Toward autonomic grids: analyzing the job flow with affinity streaming
Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining
We're not in Kansas anymore: detecting domain changes in streams
EMNLP '10 Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing
Real time illumination invariant motion change detection
Proceedings of the first ACM international workshop on Analysis and retrieval of tracked events and motion in imagery streams
Change detection for temporal texture in the Fourier domain
ACCV'10 Proceedings of the 10th Asian conference on Computer vision - Volume Part I
Real time motion changes for new event detection and recognition
ACCV'10 Proceedings of the 2010 international conference on Computer vision - Volume Part I
A survey on concept drift adaptation
ACM Computing Surveys (CSUR)
Hi-index | 0.00 |
Model-based declarative queries are becoming an attractive paradigm for interacting with many data stream applications. This has led to the development of techniques to accurately answer the queries using distributional models rather than raw values. The quintessential problem with this is that of detecting when there is a change in the input stream, which makes models stale and inaccurate. We adopt the sound statistical method of sequential hypothesis testing to study this problem on streams, without independence assumption. It yields algorithms that are fast, space-efficient, and oblivious to data's underlying distributions. Our experiments demonstrate the effectiveness of our methods to not only determine the existence of a change, but also the point where the change is initiated, relative to the ground truth we obtain. Our methods work seamlessly without window limitations inherent in prior work, thus have clearly shorter delays compared to alternative window-based solutions.