PCA for Dimensionality Reduction

Why reduce dimensions

Too many features slow training, invite overfitting, and resist visualization. PCA finds new axes that capture the most variance, letting you keep a handful and drop the rest.

from sklearn.decomposition import PCA
pca = PCA(n_components=2)
reduced = pca.fit_transform(X_scaled)
print(pca.explained_variance_ratio_.sum())

Practical notes

• Scale first — PCA is sensitive to feature magnitude.
• Use the explained-variance ratio to pick how many components to keep.
• Great for visualization (reduce to 2D) and as a preprocessing step before modeling.