What is Underfitting in Supervised Learning?
Underfitting is a common problem that occurs in supervised learning when a model is too simple to capture the underlying patterns in the data. It happens when the model has low complexity, resulting in poor predictive performance on both the training and test datasets. This situation is characterized by a high bias, where the model fails to learn from the training data adequately.
Causes of Underfitting
- Model Complexity: Using overly simplistic models, such as linear regression for a non-linear problem.
- Insufficient Features: Not including enough relevant features in the dataset that would help in prediction.
- Poor Hyperparameter Tuning: Default or inappropriate settings for model parameters can limit learning.
Consequences of Underfitting
When underfitting occurs, the model will perform poorly, leading to low accuracy on both training and validation datasets. This can prevent the model from being useful in real-world applications, necessitating a reevaluation of the model's architecture or the approach used in feature selection.
How to Overcome Underfitting
- Increase model complexity by choosing more sophisticated algorithms.
- Include additional relevant features in the dataset.
- Optimize hyperparameters using techniques such as grid search or random search.
In summary, addressing underfitting is crucial for developing accurate and effective machine learning models in supervised learning.