0

I have studied Logistic regression and I am still confused about certain things:

does logistic regression use linear regression to predict a continuous output, but since linear regression can output values above 1 or below 0 (here we are dealing with binary classification so the output must be between 0 and 1) so to squeeze the output we use the sigmoid function. the output of the sigmoid is the probability of it belonging to class 1. to set a threshold for the output of the sigmoid (for example if the value above 0.5 it belongs to class 1 and vice versa, that is where the logistic regression uses decision boundary to separate the data.

did I understand it correctly so far?

this next question is the main part I am confused about:

so when we say z = wx + b, is this the formula for linear regression or is it just weights for the logistic regression or is it the formula for the decision boundary line ?

Mohammed Younis
  • 11
  • logistic regression models the log odds, which although has one-to-one relationship with probability is not the same as probability. – bdeonovic Jun 29 '22 at 15:29
  • if $p$ is probability then log odds (often referred to as logit) is $logit(p) = log(p/(1-p))$ – bdeonovic Jun 29 '22 at 15:29

0 Answers0