Underfitting (Low Bias) is a modeling error that happens when the model does poor on the training data, doesn’t generalizes new data and does not perform well on the training set.

It can be caused by the fact that the model is simple for the training data or the data does not contains the things that you are trying to predict.


  • Use complex models
  • Select better features, increase data volume
  • Reduce the regularization
  • In ANNs increase epochs/training iterations.