The Analytics Lifecycle Toolkit

CHAPTER 8Analytics Model Development

We may say most aptly that the Analytical Engine weaves algebraic patterns just as the Jacquard loom weaves flowers and leaves.

Ada Lovelace, English Mathematician (1815–1852)

PROCESS OVERVIEW

The purpose for describing analytics model development is not to recast the whole of quantitative science including statistical methods, machine learning algorithms, artificial intelligence, text analytics, operations research, or other mathematical endeavors. There is little doubt that analytics methods will continue to change at an increasing pace, making a comprehensive review of the analytical methods nearly impossible at any point in time. Furthermore, there are dozens if not hundreds of books on each of the analytics methods described here. Instead, the focus will be on the overarching best practices used in analysis, and on the people and process aspects of analytics model development.

The goal is to provide a framework for how to consider these techniques so that they are accessible and understandable by a wider audience of data champions, and to demystify them so that others may participate more fully in the development, testing, and use of analytics models.

Typically, treatments of analytics methods tend to focus on the tactics of the method and the hype of machine learning, which can cause confusion. For example, I was approached recently by a client about helping with a “predictive modeling” problem. After some discussion, it was clear ...