0

I'm tracking the value of a variable over time. I would like to detect if the value of the variable is bursting (having an unusual increase in it's value) in real-time. How can I do that?

Given this time series: 1, 3, 5, 6, 9, 8, 7, 4, 2, 1.

can someone please explain to me by providing the calculations how to detect the burst in the previous data?

Jack Twain
  • 6,273
  • 15
  • 67
  • 107

1 Answers1

0

You could start by calculating a first derivative (rate of change), and define a burst as a first derivative being above a certain value.

Alptigin Jalayr
  • 719
  • 4
  • 12
  • I edited the question to include a time series data example, can you please explain to me how to do that using the example I provided? – Jack Twain Jul 10 '13 at 11:30
  • Let's say the burst threshold == 3. |3-1| == 2 (< 3) so it's not a burst. However, |9-6| == 3 == burst threshold, so that's a burst. – Alptigin Jalayr Jul 10 '13 at 13:13