0

Is logistic regression a linear model which separates the two classes with linear decision boundary?

Can we use logistic regression for non linear problems as well?

I looked at python sklearn and I don't see how to make a non linear logistic regression, so wondering if logistic regression cab be applicable for non linear problems as well.

Considering the hypothesis function h(theta) = 1/1+exp(z)

where z is theraT * X. I hope we can make z as a non linear function by increasing the degree. Does it affect the linear decision boundary to be a non linear?

yome
  • 219
  • 2
  • 3
  • 9
  • Logistic regression is a linear model. While they may be helpful, Scikitlearn docs is the wrong place to look for explanations on statistical concepts. Have a look at: Generalised Linear Models which subsumes both simple linear regression and logistic regression. – Zhubarb May 12 '16 at 12:48
  • And take a look at regression splines. – Frank Harrell May 12 '16 at 14:14
  • @Zhubarb thanks. If we change the degree in the sigmoid func still it be a linear model? – yome May 12 '16 at 14:53
  • In this setting the logistic function is the link function, have a read here: https://en.wikipedia.org/wiki/Generalized_linear_model#Link_function. The linearity comes from the predictor variables and how they are used, it is not to do with the response variable (and therefore the link function) – Zhubarb May 12 '16 at 15:59

0 Answers0