I’m working with a dataset containing IQ data in time domain. I’ve applied FFT to the IQ data to convert it into the frequency domain and further extract power, magnitude and phase angle from it. Each of the rows in these features contain an array of 1024 data points. I’m trying to train a Random Forest Classifier model on this data, but I encounter an error during fitting Image of the error (Apologies for the bluriness). All features have the same shape and size.

I tried concatenating the rows, flattening, converting to list but nothing seems to work. So I instead settled with the training the model with the statistical figures of these features. Is there a way I can pass the features directly obtained after FFT? Would it be too complex for the model?