• Missing values in regression trees
• Outliers:
  • Outliers will result in large values for MSE and SDR
  • This will result in further splits in order to meet the split stopping condition
  • Increase overfitting in depth of the tree
  • However, post-pruning will remove the tests leading to subset of outliers
• Noise:
  • If deviation of noise is small, then robust
  • The larger the noise deviation
    • the larger the MSE and SDR values
    • and the deeper the tree becomes
    • leading to more overfitting
  • Post-pruning will, however, result in robustness to noise

Built with DataHub Cloud