Skip to content
BENIBIX Learn

Underfitting

Also known as Underfit

The model is too simple. It does badly on old data and new data alike.

Think of it like

Think of it like drawing a cat as one plain circle. It is too simple to look right.

Example

Emma fits a straight line to house prices that jump sharply in the city centre. The line misses most houses, in training and in testing.

Why it matters

A model that cannot fit the data it learned from will never be useful. It is the first thing to rule out.

Definition

Underfitting is when a model is too simple to learn the pattern in its data. It does badly on the training data and badly on new data too. It might be a straight line used for a curved pattern. Or it might not have trained for long enough. A bigger model, better features or more training can help.
Underfit is the left panel. A straight line is too simple for dots that curve.

Common mistake

Adding more data to fix it. An underfit model needs more power, not more examples. Extra data is the fix for the other problem.

Key takeaways

  1. 1
    Too simple. Cannot follow the pattern.
  2. 2
    Bad everywhere. Old data and new data.
  3. 3
    Needs more power. A bigger model or better features.

Compare

Learn it properly