Data Mining with Microsoft® SQL Server® 2008

Chapter 12. Microsoft Neural Network and Logistic Regression

Imagine the way the human mind works when presented with a problem. At first, the problem's facts are analyzed and weighed at some sensorial level. Next, these facts are passed through neural paths, which act as filters and are based on previously known patterns. This leads to conclusions, which may be possible solutions to the problem or may serve as additional facts for a new iteration over the neural paths.

Artificial neural networks are mathematic models for the process just described. The Microsoft Neural Network algorithm is such an artificial neural network. It works by creating and training artificial neural paths (relationships between inputs and outputs) that are used as patterns for further predictions. The algorithm does a better job than other algorithms in detecting very complex relationships between inputs and outputs. Detecting such relationships is a more intricate process than in most of the other algorithms, and training a neural network is generally more time-consuming than using any other model.

In this chapter you will learn about the following:

• How to use Microsoft Neural Network and Microsoft Logistic Regression

• How to interpret models using these algorithms

• The principles of the Microsoft Neural Network

Examples, data sets, and projects for this chapter may be found in its downloadable companion, Chapter12.zip, which is available on the book's companion website at www.wiley.com/go/data_mining_SQL_2008/ ...