happy new year everybody 🎇🥂 sharing here this LinkedIn post I keep coming back to over and over: https://www.linkedin.com/posts/sebastianraschka_random-forest-rf-is-my-favorite-baseline-activity-6967757978368180224-LvGU

the number of features at each node (here called "mtry") is the most influential hyperparameter for random forests. Increasing the number of features at each node improves predictive performance in the presence of noise.

However, using feature selection (here: recursive feature elimination, RFE) has an identical effect on the predictive performance