If you’ve ever tried to read a “from scratch” decision tree tutorial and immediately hit recursion, pointer-heavy node graphs, or a bunch of hand-wavy steps, this one’s a refreshing change.
This walkthrough builds a decision tree regressor in C# end to end, but does it in a very practical way: list-based node storage (no pointer chasing), a stack-based build process (no recursion), and row indices stored per node so you can actually inspect what the tree is doing. It’s also a nice reminder of why single trees tend to overfit, and why they usually show up as part of ensembles like random forests or boosting.
If you like seeing the whole thing laid out, including the “IF column X <= threshold THEN…” style interpretability, it’s worth a skim.
Full article here:
Decision Tree Regression from Scratch Without Pointers or Recursion Using C# - Visual Studio Magazine
Have you ever shipped a single decision tree to production, or is it ensembles only for you?