Overview

It has been said that "knowledge is power." Business knowledge produces power by enabling its possessor to take the right action at the right time. It is knowledge made actionable that wields power.

Knowledge that is not applied might as well not be known. Once knowledge has been obtained from a human expert or extracted through discovery techniques, it must be put to work to provide value.

Recall that data mining has two components: discovery, during which meaningful patterns are detected and characterized in data; and exploitation, during which meaningful patterns are used to create useful applications (models).

Data mining models typically classify or predict, so are called predictive models. The term "predictive" does not suggest that there is any prognostication involved. Predictive models are basically "pattern classifiers." Applications that estimate credit worthiness based upon a credit report, or that sift through thousands of transactions and identify instances of fraud, or predict tomorrow's closing S&P500 index, are all examples of predictive models.

The process of building predictive models using data mining methods and tools is referred to as predictive modeling.

One of the most attractive facets of data mining is the development of insight into the business process that makes possible the development of predictive models. In competitive markets, niche selection, product release timing, and marketing focus all stand to benefit from the use of predictive models.

Predictive models can be based upon conventional methods (control charts, Gantt charts, spreadsheets, etc.), but these are...