online read us now
Paper details
Number 2 - June 2013
Volume 23 - 2013
Random projections and Hotelling’s T2 statistics for change detection in high-dimensional data streams
Ewa Skubalska-Rafajłowicz
Abstract
The method of change (or anomaly) detection in high-dimensional discrete-time processes using a multivariate Hotelling
chart is presented. We use normal random projections as a method of dimensionality reduction. We indicate diagnostic
properties of the Hotelling control chart applied to data projected onto a random subspace of ℝn. We examine the random projection method using artificial noisy image sequences as examples.
Keywords
change detection, multidimensional control charts, dimensionality reduction, random projections, process monitoring