Fitting Problem

Fitting problems in Machine Learning occur when a model performs very well on the training data but poorly on unseen or test data, indicating an unbalance Model.

Tldr

Fit refers to how well you approximate a target.


We deal with Fitting Problem in two stages of Machine Learning Development Lifecycle (MLDLC):

  1. Model Training: During the training phase, the model learns the patterns and relationships within the "Training Data".
  2. Model Evaluation on Test Data: The model is then evaluated on a separate set of test data to assess its performance using Evaluation Metrics.

States of a model's fitness and balance:

  • Overfitting
  • Underfitting
  • Balancing is when the model is not overfit or underfit. It's a trade-off between overfitting and underfitting and a key challenge in machine learning model development.
Example
  • Overfitting: Somebody who perfectly memorized driving test rule book, but fails the test because can’t answer questions as long as they aren’t exactly like the book.
  • Underfitting: A child that tries to memorize traffic signs. but fails to distinguish them because of lack of practice.

Fitting-Bias-Variance.png