Show me the tree!

A demonstrator application to illustrate the use of decision trees as predictive models

Created by Mikel Niño (February 2015)

What is ‘Show me the tree!’?

• Application to illustrate the use of decision trees as predictive models (‘rpart’ function)
• Uses examples of well-known statistics benchmarking datasets: mtcars, ToothGrowth
• Allows the user to tinker with some parameters to fine-tune the predictive model: training-testing set proportion, intervals/levels to categorize the continuous outcome variable

Example

This is the decision tree that it will be shown when using:

• the ToothGrowth dataset
• 75% of the observations for the training set
• 5 regular intervals to split the outcome variable (tooth length)

(This isn't a screenshot! It is generated with embedded R code, just like it is done in the application)

Benefits and potential uses

• Useful for pedagogical and teaching contexts (like this specialization :-) )
  • Easy-to-use demonstrator of decision trees
  • Illustrates how easy they are to interpret
• Easily adaptable - Source code provided! (see next slide)
  • New datasets and 'rpart' parameters can be easily added to tinker with

References

• Link to the application - give it a try! :-)

• GitHub repository with the source code

• Description of mtcars dataset

• Description of ToothGrowth dataset