1
$\begingroup$

I am a biggener student in Machine learning, and I want to ask if is it possible to convert a binary classifier label (y) by applying some condition on column1 to get a third situation.

I.e. Instead of having just 2 cases "Easy" and "Difficult" as output, I need to apply a condition on an additional feature in order to get as output "Easy", "Normal", "Difficult".

and I need also some keyword that can I use on Google as a request to look for solution.

$\endgroup$
3
$\begingroup$

Yes it is.

For multiclass classification problems, you can use 2 strategies: transformation to binary and extension from binary.

In approaches based on transformation to binary, you have:

  • OVA (one versus all), which is based on training k binary classifiers (k = #classes), where the i-th classifier is specialized on distinguishing the i-th class from all the other k-1 classes.
  • OVO (ove versus one), which is based on training k * (k-1) / 2 classifiers, where each classifier learns to distinguish 2 classes only. When a prediction is required, each clasisfier votes on the class it thinks it's correct, and the class with more votes is selected as the output.

On the other hand, you have extension from binary approaches: some classification algorithms are already capable of dealing with these multiclass problems. Some examples: kNN, decision trees, naive bayes...

You can find a bunch of resources on this. For more practical purposes, please check out the following resource: https://scikit-learn.org/stable/modules/multiclass.html

$\endgroup$
1
  • $\begingroup$ thanks Fabio for your answer, I do appreciate your kind consideration $\endgroup$ – Nirmine Mar 3 '19 at 21:22

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.