6 Classification and Prediction Methods

6.1 INTRODUCTION

In the previous chapter, we discussed three types of tasks that analysts engage in: descriptive, predictive, and prescriptive. This chapter is devoted to predictive methods, including both classification and numerical prediction. We use the term classification when the task is to predict which class an individual record will occupy; we use the term prediction when the task is to predict a numerical outcome for an individual record. For example, classification applies when we wish to predict whether a particular customer will buy; prediction applies when we wish to forecast how much they will spend.

Previously, we offered the admonition to maintain a skeptical attitude toward data. With a skeptical attitude, we are motivated to spend the time needed to carefully explore our data. Here, we offer another admonition: do not expect magic from the methods presented in this chapter. The success of any of these methods depends on carefully selected, cleaned, and prepared data. Even then there are pitfalls that only the skeptical analyst can avoid.

The methods we present here originated in a variety of fields. Some come from classical statistics, others from the more recently developed fields of machine learning, artificial intelligence, or decision analysis. As a broad generalization, we could say that in the world of classical statistics, computing was difficult and data were scarce. Moreover, the focus was more on building explanatory ...

Get Management Science: The Art of Modeling with Spreadsheets, 4th Edition now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.