1
$\begingroup$

I am working on a prediction question: what's the percentage of Y = 1 using a number of features?

The output Y values I have for training are in binary. In this case, should the prediction be treated as classification or regression?
Would logistic regression that returns the probability be suited or the probability of SVM, e.g., this?

$\endgroup$
0
$\begingroup$

Applying regression methods would render useless, as we usually feed binary (or multiple) categories as target variable in classification problems. However, if you have somehow got probability values for each sample, you can perform regression as well - I can't think of such a rare case, though.

If you are looking for prediction probabilities, Logistic Regression is the best call. Having some experience with sklearn, I can tell that SVM, by default, does not compute probabilities. If you wish to obtain those prediction probabilities, SVM uses another set of algorithms, which cause serious slowdown in process and extra computational cost.

| improve this answer | |
$\endgroup$

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.