Sometimes, when the series is too chaotic, one can use the well known trick of cumulate the data. This magically whipes out all the fluctuation. You can then fit a model, and decumulate the result with the first forward difference.
Actual behavior
I could not find a way to cumulate the data, neither in the timeseries add-on, neither in the main Orange.
Steps to reproduce the behavior
Additional info (worksheets, data, screenshots, ...)
I guess it is a typical timeseries operation, as to cumulate on a general table you have to sort according to a column.
I also think that, while it is a window operation (with start at $-\infty$), the Moving transform widget which is already rather complex, adress an other kind of transformation.
Timeseries version
0.6.3
Orange version
3.36.2
Expected behavior
Sometimes, when the series is too chaotic, one can use the well known trick of cumulate the data. This magically whipes out all the fluctuation. You can then fit a model, and decumulate the result with the first forward difference.
Actual behavior
I could not find a way to cumulate the data, neither in the timeseries add-on, neither in the main Orange.
Steps to reproduce the behavior
Additional info (worksheets, data, screenshots, ...)
I guess it is a typical timeseries operation, as to cumulate on a general table you have to sort according to a column.
I also think that, while it is a window operation (with start at $-\infty$), the Moving transform widget which is already rather complex, adress an other kind of transformation.