Hands-On Prescriptive Analytics

Chapter 6. Decision Tree Analysis: Overview

I focused on selecting from a menu of options in the previous chapters. The menu results from the Predictive Analytics stage of the Tripartite Analytics Paradigm I described in Chapter 1.

I showed in Chapter 5 how you can select one option from the menu using mathematical programming. Although effective, and certainly the approach heavily used in Prescriptive Analytics as noted by Lepeniotia et al. (2020), it has two problems that overly simplify, and perhaps obscure, how decisions are really made. These are:

1. No time dimension for decisions

2. Fixed (i.e., non-stochastic) coefficients for the mathematical programming functions

In this chapter, I will introduce temporal decision making via a decision tree paradigm. Trees are useful for clarifying conditions and showing the logical components of the thought process leading to a decision. Specifically, I will describe and illustrate a general framework for decision analysis with different states of the world (SOWs) and associated prior probabilities playing key roles. I will focus on:

• Terminology

• The role of subjective probabilities in decision making

• The construction and structure of decision trees for decision making

• The use and value of decision trees in decision making

The leading questions for this chapter are:

1. What is a decision tree?

2. Are there different types of trees?

3. How does a decision tree reflect a time dimension?

4. What is the structure of a decision tree? ...