Description: Underfitting is a modeling error that occurs when a model is too simple to capture the underlying trend of the data. This phenomenon arises when the model lacks sufficient capacity to learn from the complexity of the data, resulting in poor performance on both the training and test sets. In machine learning terms, an underfitted model fails to generalize adequately, meaning it cannot make accurate predictions on new data. Key characteristics of underfitting include a high error rate on training data and an inability to capture significant patterns. This problem is common in linear models applied to nonlinear data, where the simplicity of the model prevents recognition of more complex relationships. In practice, underfitting can be identified by evaluating performance metrics, such as accuracy or mean squared error, which indicate that the model is not adequately leveraging the information available in the data. The relevance of underfitting lies in its impact on the effectiveness of predictive models, which can lead to erroneous decisions in critical applications such as fraud detection, medical diagnosis, or market trend prediction.
