I am working on a project in R where I’ve used the ranger function to fit a random forest model. I created training and test sets by sampling indices randomly. After fitting my model, I evaluated it using the Out-of-Bag (OOB) error and analyzed feature importance using Permutation Feature Importance (PFI).
When I upload my project on github other people who tried to reproduce my results
on different machines have encountered slightly different value.
Is there a way to ensure that the outputs (OOB error and PFI values) are consistent across different environments when my project is run? What can I do to ensure that my results are perfectly reproducible, is there even a way? (and yes I set a seed in before splitting the data)
Alex Unger is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.