Random Forest

An ensemble machine learning model that builds many decision trees on random subsets of data and features, then averages their predictions — more accurate and robust than any single tree.

When you'd see it: A baseline for most tabular-data ML problems: churn prediction, fraud detection, customer scoring, demand forecasting. 'We trained a Random Forest' is often the first ML model a data scientist tries after a logistic regression baseline.

Why it matters: Random Forest outperforms single decision trees by reducing overfitting through averaging — individual trees memorize the training data, but the ensemble's errors partially cancel out. It requires relatively little tuning, handles non-linear relationships, and provides feature importance scores.

Common mistakes: Treating Random Forest as the last word in model selection. Gradient boosting (XGBoost, LightGBM) typically outperforms Random Forest on structured data when properly tuned. Random Forest is the right default for a quick strong baseline; it's rarely the best possible model for a well-resourced project.

Related decision tree, gradient boosting, overfitting, confusion matrix, AUC