DataTerminology.comNEW
A modeling error where the model is too simple to capture underlying patterns, yielding poor performance on training and test data.
DESCRIPTION
Underfitting occurs when a model lacks capacity or appropriate features to learn signal, resulting in high bias. Symptoms include low training accuracy and flat learning curves. Remedies include increasing model complexity, engineering features, or reducing regularization. Monitoring bias-variance tradeoffs is essential. Proper validation and baseline comparisons help detect underfitting early. It contrasts with overfitting, which has low training error but poor generalization.
ALTERNATIVE DEFINITIONS
Google Developers
Underfitting occurs when a model cannot capture the underlying trend of the data.
Concise description of the phenomenon.
IBM
Underfitting is a modeling error that occurs when a function is too simple to fit the data.
Emphasizes simplicity of the model as the cause.
RELATED TERMS
Bias-Variance Tradeoff Model Complexity Regularization Cross-Validation Superclass Learning Curve Feature Engineering Hyperparameter Tuning Baseline Model
COMPETING TERMS
RELATIONS
- Narrower: Model Underfitting
KNOWLEDGE AREAS
AI Data Science Testing & Validation MLOps
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer
EXTERNAL SOURCES
- Wikipedia - Overfitting#Underfitting
- Google Developers - Crash course on underfitting
- scikit-learn - Bias-variance tutorial
- IBM - Underfitting overview
TOP QUESTIONS
How do you detect Underfitting?
Look for low training performance and minimal improvement with more epochs; compare against simple baselines.
Category: Testing
How can Underfitting be fixed?
Increase model capacity, reduce regularization, add informative features, and tune hyperparameters.
Category: Requirements
What is the relationship between Underfitting and bias?
Underfitting reflects high bias where the model’s assumptions oversimplify the data.
Category: Opinion
What’s a common mistake leading to Underfitting?
Over-regularizing, using too few features, or setting overly conservative hyperparameters.
Category: Mistakes
What is the risk of Underfitting?
Poor predictions and missed patterns can harm decisions, even if the model appears stable.
Category: Risks
How does Underfitting differ from Overfitting?
Underfitting has high training error; overfitting has low training error but poor test performance.
Category: Alternatives
Where to start learning about Underfitting?
Study bias-variance tradeoff, regularization, and model capacity with tutorials and experiments.
Category: Example
Can data preprocessing cause Underfitting?
Yes; overly aggressive smoothing or loss of signal during cleaning can reduce learnable information.
Category: Risks
How does cross-validation help with Underfitting?
It reveals consistent poor performance across folds, guiding adjustments in model complexity.
Category: Testing