Modelling

Choosing an algorithm family and applying it. These pages are vendor-neutral.

Start with prerequisites for machine learning implementation, which covers what to establish before committing to a machine learning solution at all — data volume and quality, infrastructure, and whether the problem actually needs a model rather than a rule.

The algorithm families:

• Regression — predicting a continuous value
• Classification — predicting a discrete category, including the multiclass, multilabel and imbalanced variants
• Clustering — grouping unlabelled data
• Time series analysis — trend, seasonality, stationarity and forecasting

Feature preparation comes first: see data engineering and exploratory data analysis. Training and deploying the chosen model on AWS is covered under SageMaker AI.