GLOSSARY

Machine Learning (ML)

Machine learning trains algorithms to learn patterns from data — supervised, unsupervised, reinforcement — with CRISP-DM and MLOps governing the lifecycle.

Last updated:

Quick answer
Machine learning (ML) is a branch of AI in which algorithms learn patterns from data to make predictions or decisions on new data, rather than following explicitly coded rules. It covers supervised, unsupervised, semi-supervised, and reinforcement learning, and powers most of the enterprise value currently realized from AI — churn prediction, personalization, fraud detection, demand forecasting, pricing.

WHAT IT IS

The canonical ML lifecycle — framed by CRISP-DM and reinforced by modern MLOps — covers business understanding, data preparation, modeling, evaluation, deployment, and monitoring. Production ML adds versioning (data, features, models), feature stores, drift detection, and rollback — tools like MLflow, Weights & Biases, Feast, Kubeflow, Vertex AI, SageMaker, and Azure ML.

HOW IT WORKS

Common techniques include regression (linear, tree-based, boosted), classification (logistic regression, random forests, XGBoost, neural networks), clustering (k-means, hierarchical, DBSCAN), and deep learning for perception and generative tasks. Model choice is rarely the bottleneck — data, evaluation, and deployment usually are.

WHEN TO USE

Deploy ML when patterns are too varied for rules, when decisions repeat at high volume, and when the business has the labeled data or feedback loops to train and retrain.

RELATED

SOURCES

Related questions.

What is machine learning?
Machine learning (ML) is a branch of AI in which algorithms learn patterns from data to make predictions or decisions on new, unseen data, rather than following explicitly coded rules. It covers supervised learning (labeled data), unsupervised learning (unlabeled data), semi-supervised, and reinforcement-learning approaches.
How is machine learning different from generative AI?
Generative AI is a subset of ML focused on producing new content (text, images, audio, code) rather than classifying or predicting against fixed outputs. Traditional ML (churn prediction, fraud detection, demand forecasting) remains where the majority of enterprise value is realized today, despite the attention on generative AI.
What ML use cases deliver the most enterprise value?
Churn prediction, lifetime-value estimation, recommendation and personalization, fraud detection, demand forecasting, dynamic pricing, and propensity-to-buy scoring. These are the workhorses — less glamorous than generative AI, but consistently where measurable ROI shows up.
What does an ML engagement look like?
Data audit and feature engineering, model selection and training, offline evaluation against business metrics, online experimentation (shadow deployment, A/B), production deployment with monitoring, and retraining cadence. Most ML projects fail not at modeling but at deployment and drift management.
How does NUUN Digital deliver ML?
We scope to one named business outcome, build with vendor-neutral tooling (Python + sklearn, PyTorch, or cloud-native services), instrument monitoring for data and concept drift, and establish a retraining cadence before we declare the model shipped.

Need this term in action?

Book a call