Distinguishing time clustering of astrophysical bursts

Many astrophysical bursts can recur, and their time series structure or pattern could be closely tied to the emission system physics. While analysis of periodic events is well established, some sources, e.g. fast radio soft gamma-ray emitters, are suspected more subtle less explored windowed behavior: themselves not periodic, but activity only occurs during windows. We focus here on distinguishing behavior from merely clustered through clustering analysis, using techniques analogous spatial clustering, demonstrating methods for identifying characterizing behavior. An important aspect accounting ``curious incident dog in night time'' - lack carries information. As a worked example, we analyze six years data gamma repeater SGR1935+2154, deriving window period 231 days 55% duty cycle; this has now successfully predicted both active inactive periods.