Remove noise from TimeSeries

Hi all,

I have a lot of time series containing Data about the position of the biologic cell (X, Y).
What I would like to do is to remove all noise from the time series or at least a big part of it. I don’t know which kind of noise (model) is present and I can’t predict the X, Y movement.

Which method can I use on KNIME? What do you suggest?


Hi @GiulioFerrari,

it’s an interesting question. I don’t know if there is an existing approach for doing this, however, I think after succeeding defining a distance there should be an idea.

So I would try the following…

  1. get the previous values of the specific potential noise column by using the lag-node (-1 and +1,…)
  2. define a distance between the current value and value(-1)…
  3. calculate a measure like distance(-1;0)/distance(-1;1) [just a simple one]
  4. use numeric outlier node to detect noise / outliers

just sketching a possible workflow.
It would be useful to invest some statistical know-how in step 3.

No solution, just starting the discussion :slight_smile:
Greetz, Tommy

1 Like