2
$\begingroup$

This question already has an answer here:

I understand that logistic regression has some nice properties that works well for classification problems, such as the S-curve shape, the output value being between [0,1], and continuous across X. But there are some other functions which also share such properties, for example the Complementary log-log function or even the CDF of Gaussian distribution.

Why is the logistic function most commonly used in classification problems? Is it because the logistic function is relatively easy to implement?

|cite|improve this question
$\endgroup$

marked as duplicate by kjetil b halvorsen, Michael Chernick, Haitao Du, Stephan Kolassa, Peter Flom♦ regression Mar 6 '18 at 12:03

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

2
$\begingroup$

Why is the logistic function most commonly used in classification problems? Is it because the logistic function is relatively easy to implement?

The logit function is the canonical link function and thus an obvious default (see Wiki on generalized linear models). See this post for some of the properties of canonical link functions which makes them nice.

Do see the post Tim links to.

|cite|improve this answer
$\endgroup$
  • 1
    $\begingroup$ The question is confusing. Logistic regression has next to nothing to do with classification. It is a direct probability estimation model. See fharrell.com/post/classification $\endgroup$ – Frank Harrell Mar 5 '18 at 20:46

Not the answer you're looking for? Browse other questions tagged or ask your own question.