Drift is a key component in streaming data. As we start receiving the data, there may be a change in the nature of the data, say in terms of variance or other statistical properties. Let us say initially the stream throws data points with a particular mean and a standard deviation. Over time, the mean and the standard deviation of the data is changed. This constitutes a drift.
The streaming algorithm should be able to accommodate these drifts.
Refer to the following for more information about drifts: https://en.wikipedia.org/wiki/Concept_drift
DSD has some generators that can incorporate the drift:
> drift.stream <- DSD_MG(dim = 2)> cluster.1 <- MGC_Linear(dim = 2)> add_keyframe(cluster.1, time=1, dens=50, par=5, ...
