COMPEL Glossary / supervised-learning
Supervised Learning
Supervised learning is the most widely deployed machine learning paradigm in enterprises.
What this means in practice
The model is trained on labeled examples -- inputs paired with known correct answers -- and learns to predict the correct output for new, unseen data. Supervised learning divides into classification (assigning categories, like spam detection) and regression (predicting numbers, like demand forecasting). The critical business implication is the labeling requirement: every supervised model needs data where the correct answer is known. For some tasks, labels exist naturally in enterprise systems (customer churn records, fraud outcomes). For others, labels must be created manually by human experts, which is often the most expensive part of an ML project. Data labeling costs should be a primary factor in COMPEL use case evaluation during the Model stage.
Why it matters
Supervised learning is the most widely deployed ML paradigm in enterprises, but its critical business implication is the labeling requirement: every model needs data where the correct answer is known. For some tasks labels exist naturally; for others, manual labeling by human experts is the most expensive project component. Organizations that understand this constraint make better investment decisions about which AI use cases to pursue and when.
How COMPEL uses it
Data labeling costs are a primary factor in COMPEL use case evaluation during the Model stage, where feasibility assessment includes label availability and creation effort. The Calibrate stage inventories labeled datasets across the organization. The Technology pillar assesses labeling infrastructure and processes. The Evaluate stage measures whether supervised models perform adequately on production data that may differ from training distributions.
Related Terms
Other glossary terms mentioned in this entry's definition and context.