Machine Learning Problem Types, Models, and Data Setup

Machine learning (ML)

AI subfield where computers learn patterns from data without being explicitly programmed for every rule.

Training data

Examples used to teach a model patterns for predictions or decisions.

Operational data

Data generated during normal system use for monitoring and improvement.

Quality data

Data accurate and relevant enough for the model to learn useful patterns.

Supervised learning

Learning from labeled examples, such as emails labeled spam/not spam.

Unsupervised learning

Finding patterns/groups in unlabeled data.

Clustering

Grouping similar examples together without a known label.

Reinforcement learning

Learning by taking actions and receiving rewards/penalties.

Semi-supervised learning

Using a small amount of labeled data with a larger amount of unlabeled data.

Weakly supervised learning

Learning when labels may be noisy, inaccurate, or imprecise.

Regression

Predicting a continuous numeric value, such as house price.

Classification

Predicting a category/class, such as spam vs. not spam.

Federated learning

Decentralized training where raw data stays on local devices/servers.

Problem formulation

Defining the ML problem, goals, outputs, constraints, and success metrics before modeling.

Success metrics

Measurements used to decide whether the model or project is successful.

Cloud infrastructure

Computing resources hosted in the cloud; important later but not the core ML suitability question.

Customer preferences

Information about what customers like; can be useful data but not the same as success metrics.

Data science best practices

Good practices for data/model work; important but broad.

Hybrid infrastructure

Combines on-premises and cloud resources; may support on-prem testing and cloud scalability.

On-premises testing capability

Ability to test systems on local/company infrastructure.

Cost of ownership

Total cost to build, run, maintain, and support a system.

Compliance issue

A concern about meeting legal/regulatory requirements.

Privacy concern

Concern that personal/sensitive data could be exposed or misused.