Supervised learning

3.15 Decision trees (also known as classification and regression trees [CART], conditional inference trees)

3.15.1 General explanations and interpretation

• Start with this paper and focus on the CART (Classification and Regression Tree) section mainly.

• A plainer-language explanation of how splits are chosen with a sample dataset on irises and some exerpts from ctree documentation.

3.15.2 Examples of CART in the wild

• These two papers use R’s ctree() from the partykit package

  • Available through OU’s subscriptions

  • Available via Interlibrary Loan

    • Access OU’s Interlibrary Loan Article Request Form

• A different algorithm to run CART: I suggest it because it has good figures and methods section

3.15.3 How and why to run the ctree version of CART/decision trees

• A short explanation of why ctree is a good general choice to minimize overfitting.

• The R vignette for the partykit package provides worked examples and code for both categorical responses (classification trees) and numeric responses (regression trees).

• ggparty allows for fancier and more informative plots

  • Another example using ggparty with a published dataset