This book, Berry and Linoff's first, was the first management-oriented description of data mining for marketing published. At the time, there were only vendor white papers, seminar materials, and a handful of academic papers that described the data mining process from a business person's point-of-view. In many ways they got it right months before anyone else. This book has been widely used both as a business seminar reference and as an academic text.
Like most introductory chapters, this one is good to skim. Unless you are completely new to data mining, don't spend a lot of time here.
This book was written before the Esprit CRISP-DM or SAS SEMMA white papers were published. One can certainly argue that Berry and Linoff's "Virtuous Cycle," based on projects they ran at Thinking Machines and MRJ, significantly influenced both. Of particular importance in these chapters is the author's unabashed willingness to expose the "dirty," time-consuming, unglamorous side of data mining. They also stress the importance of of the iterative nature of data mining in the workplace. In class we will discuss all of these data mining methodologies and integrate them into a complete framework. Chapter 18 is a nice complement to these chapters.
This chapter includes a very simple case study example of a data mining project. The most important material in this chapter, however, is a brief discussion of data mining tasks: Classification, Estimation, Prediction, Affinity Grouping, Clustering, and Description. We will explore these tasks in more detail, tying material from other chapters into our discussion. In many ways its a theme the authors pick up again in Chapter 17, Choosing the Right Tool for the Job. This is essential material for the practicing data miner.
Chapter 5 introduces a number of data analysis best practices that are critical to successful data mining. We will cover all of this material thoroughly.
This is where the bottom line is revealed if you've done all the prior project work completely and unambiguously. Do newly developed models perform better than the old ones? Will the new models provide positive financial results? Which models are best? Would a combination of models be an improvement? Etc. This is the information senior management wants to see! The various measurement techniques covered in this chapter are used not only to assess the outcome of a data mining exercise, but they are also used to estimate and direct the overall project.
This chapter introduces the major data mining techniques and is a complement to Chapter 17. It also relates closely to Chapter 4. We will attempt to provide an integrated view of all of the various techniques, their underlying data analysis task, and their CRM application.
These chapters contain descriptions of 7 of the most common data mining techniques used today. Berry and Linoff's treatment takes a more technical, hands-on approach to their description than Berson, Smith, and Thearling. Combining the material in both books gives a good picture of how the technique works and for what marketing function they are used. We will cover Chapters 8, 10, 12, and 13 in this class.
Chapter 15 discusses the relationship between data warehouses and data mining. Data warehouses are built to provide a single, comprehensive source of data within an organization. As such, they share a symbiotic relationship to data mining: data warehouses are the natural source for most of the data to be analyzed, and data mining is a value-generating activity that justifies the resources spent to build these expensive databases. We will not be covering this chapter in class.
OLAP stands for "Online Analytical Processing." There is a class of database management systems that have specialized algorithms built into them to support this data analysis activity. This chapter compares and contrasts OLAP with data mining and would probably be of interest to students working in organizations where these databases are being deployed. We will not be covering this chapter in class.
Chapter 17 touches on an important aspect of data mining: when and where to use each technique and what the implications of that choice will be. It is no different than knowing which statistical procedure to use given specific types of data. We will combine the material from this chapter with chapter 7 and material from our other text to provide as complete a picture as possible for this important topic.
This chapter complements the material in Chapters 2 and 3 by exploring the organizational ramifications of data mining. In many ways, this chapter is really a description of the Organizational Change Management requirements for developing an effective data mining function within an enterprise. We will only lightly cover this chapter in class.