Random Forest

Random Forest Q&A

1What is random forest?
Answer: Ensemble of decision trees using bagging and random feature subsets.
2Why better than single tree?
Answer: Reduces variance and improves robustness.
3What is bagging?
Answer: Training models on bootstrap samples and averaging predictions.
4What is OOB error?
Answer: Internal validation using samples not included in bootstrap draw.
5Key hyperparameters?
Answer: n_estimators, max_depth, max_features, min_samples_leaf.
6Feature importance in RF?
Answer: Mean decrease impurity or permutation importance.
7Does RF need scaling?
Answer: Usually no.
8How RF handles nonlinearity?
Answer: Through many nonlinear tree splits.
9Main downside?
Answer: Less interpretable and larger memory/time than single trees.
10When RF struggles?
Answer: Extrapolation in regression beyond training range.
11RF for imbalanced data?
Answer: Use class weights/resampling and threshold tuning.
12One-line summary?
Answer: Random forest is a strong, low-tuning tabular baseline.