Machine Learning in Java - Second Edition

Logistic regression

Linear regression works when the dependent variable is continuous. If, however, the dependent variable is binary in nature, that is, 0 or 1, success or failure, yes or no, true or false, survived or died, and so on, then logistic regression is used instead. One such example is a clinical trial of drugs where the subject under study either responds to the drugs or does not respond. It is also used in fraud detection where the transaction is either a fraud or not fraud. Normally, a logistic function is used to measure the relationship between dependent and independent variables. It is seen as a Bernoulli distribution and, when plotted, looks similar to a curve in the shape of characters.