Hi , it's a bit of a slow reply I realise, but I have just been looking at how many warnings we are logging and it seems to me that the baselining calculation doesn't allow for variations caused by night time and weekend lulls. The little graph it displays shows my supposedly deviating stats are following a fairly normal pattern, but the baseline is way too low for daytime activity levels. I can only assume that's because it's an average over all time and the variation between my day and night is huge, as I would guess it is for most people. It uses some standard deviation to calculate a tolerance, but that's far too conservative. Take an example of my logging rate, to the human eye you can see it's sticking to the normal pattern but because the rate drops to the low 100's overnight, the 2,000 rate in the daytime is way outside the baseline and tolerance. Weekends just add to that imbalance. Palo, can you change the algorithm to take into account time of day variations? I'm no mathmetician so don't know how, but at teh moment I am just having to ignore/filter out the deviating device logs as they trigger all the time.
... View more
