Machine Learning
Getting computers to find patterns from data instead of being explicitly programmed.
Rather than hand-coding rules, you feed a model examples and it adjusts internal parameters to minimize error. Once trained, it generalizes to new inputs. It powers recommendations, fraud detection, and image recognition, and its quality depends heavily on the data it learns from.
Performance depends on data quality, feature representation, and avoiding leakage; the bias-variance trade-off and proper train/validation/test splits are fundamental.
Spam filters, recommendation engines, fraud detection, scikit-learn, churn prediction.
Like teaching a child to recognize dogs by showing thousands of photos rather than describing every breed: eventually they spot a dog they've never seen before.