Unsupervised Learning
A machine learning approach where models identify patterns or groupings in unlabeled data without explicit outcome guidance.
Definition
Techniques (e.g., clustering, dimensionality reduction, anomaly detection) that discover intrinsic structure in data by optimizing objectives like cluster cohesion or reconstruction error. Governance considerations include validating discovered patterns against domain expertise, monitoring for spurious or biased groupings, and ensuring data quality since no labels exist to catch errors automatically. You must document algorithm choice, parameter settings, and interpretability tools used to inspect results.
Real-World Example
A retail chain uses k-means clustering on customer purchase histories to segment shoppers into behavior-based groups. Analysts validate segments against demographic surveys and adjust cluster counts to ensure meaningful marketing personas, avoiding misleading groupings driven by data artifacts.
.svg)
