Fit a dynamical system to a 2 variables time series

Hello,

I have data that consists of an N=1000 observations * 3 columns matrix.
the first two columns are 2 variables that condition the third one.

I used DynamicalSystems.jl to estimate the phase space portrait via delay embedding, and obtained a slow varying trajectory that I would like to fit a dynamical system model onto it.

image

Is there any functionality of a package to do that?

Hi, it sounds like you’re looking for this:

If you’re looking to identify the system just from the data, maybe take a look at Home · DataDrivenDiffEq.jl.

Thank you, but if I understood the packages correctly, they still need an equation to generate the data.

so… do you have data or do you need to generate data?