Feature Selection

Adding informative features improves the model. Each feature carries useful signal about the target.

Filter methods rank features by statistical measures like correlation — fast and scalable but they ignore feature interactions.

Forward selection greedily adds the feature that most improves the model at each step. Press Play or drag the slider to start.

Low penalty — all features retained. The model uses every available signal, including noise.

Tree-based models naturally rank features by how much they reduce impurity across all splits. Importance scores are becoming more stable as the ensemble grows.

The optimal feature count balances information gain against the curse of dimensionality. You're at the sweet spot! Maximum test accuracy with minimal overfitting.