Descriptions and Prediction: Profiling and Predictive Modeling
Chapter 3 introduces data mining in terms of goals, tasks, and techniques. This chapter focuses on the most common type of data mining, directed data mining, and introduces a methodology for descriptive and predictive models. The next six chapters enhance the themes in this chapter by describing specific techniques.
Regardless of the data mining techniques being used, all descriptive models (also called profiling models) and prediction models — the two main types of directed data mining models — have much in common. When using such modeling techniques, the data miner has a goal in mind, and known examples of this goal are available in the model set. The following are all examples of directed data mining:
- Which customers are likely to stop?
- When will customers make their next purchase?
- What is the best next offer for each customer?
- What distinguishes high-end customers from the average customer?
All of these are examples where the historical data contains, respectively, customers who stop, the intervals between purchases, the products different customers purchased, and customers in segments. The goal of the data mining task is usually to find more customers like the good ones in the data.
The chapter starts by explaining what directed data mining models look like. An important part of this discussion is the subject of model stability — models should work not only in the lab but also in the real world. The ...