Machine learning

Machine learning is a set of techniques for building models that learn patterns directly from data, rather than being explicitly programmed with rules for every case. A model is trained on historical examples—inputs paired with known outcomes for supervised learning, or unlabeled data for unsupervised learning—and learns parameters that let it generalize to new, unseen inputs.

Common approaches include regression and tree-based models for structured, tabular data, clustering for unsupervised grouping, and neural networks for unstructured data like text and images. Models are evaluated on data withheld from training to check whether they generalize rather than simply memorize; performance on new data, not on the training set, is what matters.

In analytics, machine learning extends classical regression analysis into more flexible predictive analytics and powers anomaly detection, recommendation, and increasingly the natural-language interfaces behind augmented analytics tools. Its main pitfalls mirror those of any statistical model but at larger scale: overfitting to training data, propagating bias present in historical data into future decisions, and treating a model's output as ground truth rather than a probabilistic estimate that requires monitoring as real-world conditions drift away from what the model was trained on.