K-fold cross validation to avoid overfitting¶

Python notebook: https://github.com/daviskregers/data-science-recap/blob/main/23-k-fold-cross-validation.ipynb

• One way to further protect against overfitting is K-fold cross validation
  • Split your data into K randomly assigned segments
  • Reserve one as your test data
  • Train on the combined remaining K-1 segments and measure their performance against the test set
  • Repeat for each segment
  • Take the average of the K r-squared scores
• Prevents you from overfitting to a single train/test split