Underfitting
Definition
Underfitting means that a model has not yet sufficiently captured important patterns.
A model is underfitted if it does not adequately represent relevant structure even in training data.
Simply explained
The model has learned the task too coarsely. Therefore, it makes errors even on examples it has already trained on.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
A churn model systematically misses both historical training periods and new months.
Limits and pitfalls
- High training error can also result from faulty labels or optimization; more complexity is not automatically the solution.
Origin & history
The term follows from the bias-variance understanding of statistical learning.
Use in mid-sized companies
- Possible application: A paragraph model systematically misses both historical training periods and new months.
Frequently asked questions
What does underfitting mean?
A model is underfitted when it does not sufficiently represent relevant structure even in training data. What should you watch out for? High training error can also result from faulty labels or optimization; more complexity is not automatically the solution.
Sources
- Deep Learning, Capacity and Underfitting · Goodfellow et al., Deep Learning, Capacity and Underfitting
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: