MEPX
Chapter 6 of 10All chapters
  1. 1Framing the problem
  2. 2Data
  3. 3Training and testing
  4. 4Underfitting and overfitting
  5. 5Measuring performance
  6. 6Common algorithms
  7. 7How training works
  8. 8The pipeline
  9. 9After deployment
  10. 10Common pitfalls

Chapter 6 of 10

Common algorithms

A short tour.

The workhorses

Linear and logistic regression are interpretable baselines. Decision trees split on feature values. Random forests and gradient boosting combine many trees and win most tabular problems.

  • Gradient boosted trees remain the default for structured data.
  • Neural networks dominate images, audio and language, not spreadsheets.

Unsupervised

Clustering groups similar records without labels, and dimensionality reduction compresses many features into few. Both are exploratory rather than predictive.