MEPX
Chapter 2 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 2 of 10

Data

Where most of the work happens.

Quantity and quality

Models learn from examples, so gaps in the data become gaps in the model. Representative data beats abundant data, and both beat a cleverer algorithm.

  • Label quality sets a ceiling on performance no model can exceed.
  • Class imbalance needs deliberate handling, not just more data.

Features

Choosing and shaping inputs is where domain knowledge enters. A well constructed feature often does more than any change of model.