0
$\begingroup$

My model should take students' info and achievements from the preparatory group and with machine learning offer the faculties (there are 7 of them) in which the percent of his success are the highest.

I'm new to ML but I understand that this is a multi-classification. Can Random Forest or SVM handle such a task? I don't need only one "the best" faculty but rather to rank them from best to worst for each student.

My data consists of 3500 students, 25 variables (haven't feature-engineered them yet) and 7 classes (the faculties, which are the target value)

$\endgroup$
1
$\begingroup$

Yes, random forests (or any classifier which can output the probability (score) for each class) can be used to rank the faculties. You can rank them in order of probabilities.

Think of this problem in the following way: you want to find a score of every faculty for each student. This score gives you the ranking of faculties for each student.

$\endgroup$
  • $\begingroup$ I guess I do it with predict() when type="prob"? In that case, won't all classes together give me a total of 100%? My concern is that a strong student can finish (for example) 3 faculties with 90% equally. $\endgroup$ – Riddle-Master Aug 12 '17 at 14:16
  • 1
    $\begingroup$ There is a method called predict_prob() for all predictors. In your case you could build individual models for each faculty (like a binary classifier, think logistic regression) and hence, strong students will get 90% probability in more faculties. $\endgroup$ – Shishir Pandey Aug 14 '17 at 14:17

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.