|
Topics and Objectives:
Decision Trees are the third approach we will examine in the category of predictive models.
Given a dependent variable and a set of potential predictor variables, decision trees are computationally intensive algorithms that iteratively search through the predictor variables to identify those that are 'best' are predicting the dependent variable. In the process the algorithms produce a 'tree' (hence the name) and also a set of readily interpretable decision rules (for example, if professional female over age 30 with young children, then send offer for educational children's software club membership).
Readings:
Deriving Rules from Data: Machine Learning Algorithms, Chapter 10 from Seven Methods for Transforming Corporate Data into Business Intelligence
This a chapter from a book on data mining techniques - and was the best I could locate on Decision Trees.
It is a bit technical at times - do not worry too much about the details. Focus instead on getting a general overview of what decision trees do, when they might be useful and how they compare with other approaches such as linear or logistic regression.
Assignment 4: Predicting Response at BookBinders:
Decision Trees
This assignment returns to the BookBinders Book Club for one last comparison: how will decision trees compare with RFM and logistic
regression?
Back to top
|
