Fig. 5: Two examples of the occurring concept drift.
On the top panel, the daily time series of the PM2.5 concentrations of a reference instrument is shown, while on the lower panel, the daily time series of the NO2 concentrations is presented. The red verticals indicate a change in the concept of the series as identified by the adaptive windowing (ADWIN) algorithm224. The statistical properties of each segment are presented above the time series to demonstrate the effect of the concept drifts.
