Table 1. Algorithm 1.

Sampling algorithm
Input: The current privacy budget ϵi, the predicted value ${\hat{y}}_i^j$, the release data at the last sampling timestamp ${\tilde{x}}_l^j$
Output: Sampling or not
Calculation $\mathrm{dis}=\left|{\hat{y}}_i^j-{\tilde{x}}_l^j\right|$
Calculation ${\lambda }_i^j=1∕{ϵ}_i$
Ifdis > λi, then
Set i as the sampling point, update sampling interval
I =i −l
Else
i is not a sampling point, release predictive value ${\hat{y}}_i^j$
End if