- All Implemented Interfaces:
- Serializable, MOAObject
- Enclosing class:
- SeqDrift1ChangeDetector
public class SeqDrift1ChangeDetector.SeqDrift1
extends AbstractMOAObject
SeqDrift1 uses sliding window to build a sequential change detection model
that uses statistically sound guarantees defined using Bernstein Bound on
false positive and false negative rates. This is a block based approach and
checks for changes in the data values only at block boundaries as opposed to
the methods on per instance basis. SeqDrift1 maintains a sliding window and
repository. Repository gathers the new instances and sliding window stores
only the data values that are statistically not different, in other words
from the same distribution. If the data values in the repository are
consistent with the values in sliding window the data values of the
repository are copied to the sliding window applying reservoir algorithm. The
hypothesis is that the mean values of the sliding window and right repository
are not statistically different. In addition, SeqDrift1 declares a warning
state depending on warning significance level and increases sample size to
get a statistically more rigorous mean value
Sakthithasan, S., Pears, R., & Koh, Y. (2013). One Pass Concept Change
Detection for Data Streams. In J. Pei, V. Tseng, L. Cao, H. Motoda, & G. Xu
(Eds.), Advances in Knowledge Discovery and Data Mining (Vol. 7819, pp.
461-472): Springer Berlin Heidelberg.
- Author:
- Sakthithasan Sripirakas sripirakas363 at yahoo dot com
- See Also:
- Serialized Form