Logistic regression

A classification algorithm is a process whose input is a training set, as previously described, and whose output is a function that classifies data points. The ID3 algorithm produces a decision tree for the classification function. The naive Bayes algorithm produces a function that classifies by computing ratios from the training set. The SVM algorithm produces an equation of a hyperplane (or hypersurface) that classifies a point by computing on which side of the hyperplane the point lies.

In all three of these algorithms, we assumed that all the attributes of the training set were nominal. If the attributes are instead numeric, we can apply linear regression, as we did in Chapter 6, Regression Analysis. The idea of logistic regression is to transform a problem whose target attribute is Boolean (that is, its value is either 0 or 1) into a numeric variable, run linear regression on that transformed problem, and then transform the solution back into the terms of the given...