Algorithm #4
Random Forest ๐ฒ
Ensemble Learning ยท Phase 1
๐ฒ๐ฒ๐ฒ
Random
Forest
Forest
๐ง
Tree 1
๐ง
Tree 2
๐ง
Tree 3
๐ง
Tree 4
๐ง
Tree 5
๐ง
Tree 6
How It Works โจ
๐ Data
โ
๐ฒ Bootstrap
โ
๐ฒ Trees
โ
๐ณ๏ธ Vote
โ
โ
Result
๐ฒ Bagging
Each tree trains on a random sample of the data (with replacement). This creates diversity among trees!
๐๏ธ Feature Randomness
Each split considers only a random subset of features. This prevents trees from all making the same mistakes.
๐ณ๏ธ Voting Mechanism
For classification: majority vote wins. For regression: average all predictions. Democracy of trees!
๐ฏ Wisdom of Crowds
Many weak learners become one strong learner. Individual trees may overfit, but the ensemble stays robust!
๐ง Key Parameters
โข
n_estimators โ Number of trees in the forest (more trees = more stable)
โข
max_features โ Number of features to consider at each split
โข
max_depth โ Maximum depth of each tree (controls overfitting)โ๏ธ Bias-Variance Tradeoff
๐ด
Single Tree
(High Variance)
(High Variance)
โ๏ธ
๐ฒ๐ฒ๐ฒ
Random Forest
(Balanced!)
(Balanced!)
Ensemble reduces variance without increasing bias