I have one dependent binary categorical variable, and one independent continuous variable. There is a lot of randomness deciding the result of the dependent variable.

The relationship between the independent variable and the dependent variable is linear. I have 2,000 data points to train data on. Some possibilities are:

  • Logistic regression - simplest option
  • SVM (support vector machines)
  • Naive bayes
  • Random forests - I see this does well on kaggle, but I have a simple one variable linear relationship, so it seems random trees isn't necessary here.
  • 2
    $\begingroup$ It's next to impossible that the binary response is actually linear in the independent variable, unless the IV is very restricted in range --- and if it were actually linear, why would you use logistic regression? Isn't that nonlinear? Lastly, you state that it's linear with apparent certainty. Where does that certainty arise? $\endgroup$
    – Glen_b
    Jun 11, 2013 at 1:13

1 Answer 1


You should try all of the models you listed, and cross-validate them. It's the name of the site!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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