Parameter Management
All tunable values are stored in params.yaml. This allows experiments to be rerun without touching code.
Example:
data_ingestion:
test_size: 0.2
feature_engineering:
max_features: 50
model_building:
n_estimators: 25
learning_rate: 0.1
Changing any parameter automatically triggers only the affected pipeline stages when running dvc repro.