Underfitting occurs when a machine learning model is too simple to learn the underlying patterns in the training data. As a result, the model performs poorly on both the training data and new, unseen data.
In simple terms:
👉 Underfitting happens when a model has not learned enough from the data to make accurate predictions.
Why Does Underfitting Occur?
Underfitting typically occurs when the model cannot capture the complexity of the problem.
Common causes include:
- Using an overly simple model
- Insufficient training time
- Too few features in the dataset
- Excessive regularization
- Poor feature engineering
When these issues exist, the model fails to recognize important relationships within the data.
How Does Underfitting Affect Performance?
An underfitted model struggles to make accurate predictions because it has not learned the key patterns present in the dataset.
Common effects include:
- Low training accuracy
- Low testing accuracy
- High prediction errors
- Poor generalization
- Weak decision-making performance
Unlike overfitting, underfitting results in poor performance on both training and test datasets.
Signs of Underfitting
You may suspect underfitting when:
- Training accuracy remains low
- Validation accuracy is also low
- Training and validation errors are both high
- Learning curves show little improvement over time
These indicators suggest the model lacks sufficient complexity.
Example of Underfitting
Imagine trying to predict house prices using only the number of bedrooms.
A simple model may ignore other important factors such as:
- Location
- Property size
- Age of the house
- Nearby amenities
Because it uses limited information, the model may produce inaccurate predictions and fail to capture real-world trends.
How to Resolve Underfitting
Several techniques can help improve an underfitted model:
Increase Model Complexity
Use a more powerful algorithm that can learn complex relationships within the data.
Add More Relevant Features
Including meaningful variables can help the model better understand patterns.
Reduce Regularization
Excessive regularization may oversimplify the model and limit learning.
Train Longer
For some algorithms, additional training epochs can improve performance.
Improve Feature Engineering
Creating better features often helps models capture important information more effectively.
Underfitting vs Overfitting
- Underfitting: Model is too simple and performs poorly on both training and test data.
- Overfitting: Model is too complex and performs well on training data but poorly on unseen data.
The goal is to find the right balance where the model captures meaningful patterns while still generalizing well to new data.
Real-World Impact
Underfitting can negatively affect many machine learning applications, including:
- Sales forecasting
- Fraud detection
- Medical diagnosis
- Customer churn prediction
- Recommendation systems
In these cases, failing to learn important patterns can lead to inaccurate predictions and poor business outcomes.
Conclusion
Underfitting is a machine learning problem that occurs when a model is too simple to capture the underlying patterns in the data. It typically results in low accuracy on both training and testing datasets, indicating that the model has not learned enough from the available information. Common causes include simple model structures, insufficient features, excessive regularization, and inadequate training. By increasing model complexity, improving feature engineering, and optimizing training processes, underfitting can be reduced, leading to more accurate and reliable predictions.