Togaware DATA MINING
Desktop Survival Guide
by Graham Williams
Google

Logistic Regression

Linear regression is a successful framework for building models. However, not all data fits the assumptions underlying linear regression.

Logistic Regression is appropriate when the target variable is binary. It is used to build a linear model involving the input variables to predict a transformation of the target variable, in particular, the logit function, which is the natural logarithm of what is called the ``odds'' ( $log(\frac{p}{1-p}$).

The inverse of the logit function is the logistic function, $\frac{1}{1+e^{-z}}$, to return the results of the linear model back to the 0-1 range.

We can see the effect of the logistic function in the following plot. Essentially, it maps numbers from a range from minus infinity to plus infinity, to the range 0 to 1.


http://rattle.togaware.com/code/rplot-logistic.R

Note: the way it goes is if it is numeric target and logistic regression it expects 0/1. If it is a categoric target, with two values, then it will convert to 0/1 internally.

Because trees are not such a numerically oriented model builder they worry less abuot value ranges.



Subsections
Copyright © 2004-2010 Togaware Pty Ltd
Support further development through the purchase of the PDF version of the book.
The PDF version is a formatted comprehensive draft book (with over 800 pages).
Brought to you by Togaware. This page generated: Sunday, 22 August 2010