Hi,
Nowadays, with the advent of tidymodels,
it's possible to write various models with similar code, right?
I find this very convenient because I can just specify the name of the model and the code is still available.
However, since tidymodels has appeared,
I don't know much about the difficulties I had before that.
What kind of code did you write when there were no tidymodels?
For example, I used to use library(dummies) and fastDummies when I used xgboost.
FactoMineR::PCA was used to create the features of the principal components.
When combining, (or model stacking), the predictions of the models, I used to add weights to the predictions and sum them myself.
When comparing multiple models, there were no tidymodels workflowsets, so I had to create them individually and then compare their performance.
I think I used the library(Metrics) package to compare the performance.
Would you mind sharing your experience with us?
I'd be very excited to see some before and after code using tidymodels, but I'd like to know what packages you were using and how you combined them.
I would like to know how you used the packages, how you combined them, how you stacked them, how you compared many models, whatever.
Thanks for reading.