how to use minirocket in production about minirocket

[angus924]

Hi @ramdhan1989, thanks for your question, sorry for the slow response.

This could make sense if the new dataset is drawn from more or less the same distribution as the old dataset.

In any case, what you would need to do is: (a) store the parameters generated by fit(...); and (b) store the ridge regression model.

The parameters are a tuple of three numpy arrays. You could use pickle on the tuple, or you could use numpy save / numpy savez to store the arrays via numpy.

In terms of the ridge regression model, I believe you can pickle the fitted model, or there are a couple of other options for storing the model: https://scikit-learn.org/stable/model_persistence.html. Another option which is a bit more convoluted would be to extract the learned parameters etc from the model, save these, and then assign them to a new classifier instance when loading.

I hope that helps a bit.

Let me know if you have any other questions.

Thanks.