-2
$\begingroup$

I want to know that what property or attributes of logistic regression make it to useful for modeling user behavior.

$\endgroup$
  • 2
    $\begingroup$ You use logistic regression when your dependent variable is dichotomous (or ordinal). Your question is too vague to be answerable as it currently stands. $\endgroup$ – Patrick Coulombe Jul 4 '15 at 2:46
  • 1
    $\begingroup$ Is this for some class? $\endgroup$ – Glen_b -Reinstate Monica Jul 4 '15 at 3:35
  • $\begingroup$ i mean how to model user behavior using logistic regression? $\endgroup$ – Rahul Jul 4 '15 at 3:54
  • $\begingroup$ thank you, Patrick Coulombe for your comment. i know that logistic regression is used for modeling user behavior, so how it model user behavior. there are some other methods e.g. linear regression, but linear regression can not model user behavior so what property present in logistic regression which will model user behavior. please reply me... $\endgroup$ – Rahul Jul 4 '15 at 4:14
1
$\begingroup$

You'll have to be more specific about what you mean by "user behavior", but I think you're actually wanting to know what is logistics regression in the first place. Lightly speaking, logistics regression is a method of regression where we have a binary response variable; that is, a 1/0 response variable. More specifically, it predicts the probability that an observation is either a 1 or a 0. This is useful if you want to classify an observation into one of two categories. For example, in terms of user behavior, we might want to predict whether a link will be clicked on, whether a customer will purchase an item, whether a voter will vote for one candidate or another, etc.

Basic linear regression cannot be used to predict probabilities due to the constraints of ordinary least squares (you should know these). Now, basic linear regression can still be useful - your response variable won't always be a 1/0 variable. For example, you can model consumer spending on luxury items as a function of income.

I hope this answers your question. Linear regression can be used to model user behavior, so long as you can represent the response variable in the appropriate way. But you may want to measure the probably of whether or not a user will do something, and logistics is useful for that.

$\endgroup$
  • 2
    $\begingroup$ The correct term is logistic regression and it is not used to classify observations; it is used to predict probabilities. $\endgroup$ – Frank Harrell Jul 4 '15 at 12:17

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