NUMPA

LEARNING GUIDE

Machine Learning

Learn how datasets, validation strategies and model evaluation fit together in a reliable ML workflow.

Updated November 3, 2026

Learn how datasets, validation strategies and model evaluation fit together in a reliable ML workflow.

The workflow

A typical machine-learning project defines a task, inspects the available data, prepares features, selects a validation strategy, trains candidate models and evaluates results on data not used to fit the model. The workflow is iterative: poor results may reveal data-quality or problem-definition issues.

Why data quality matters

Missing values, duplicates, inconsistent categories, leakage and unrepresentative samples can distort evaluation. A readiness assessment can help identify risks before training, but no single score guarantees model performance.

Core topics to study

  • Classification and regression
  • Forecasting and time-aware validation
  • Clustering and unsupervised evaluation
  • Train, validation and test separation
  • Data leakage and reproducibility