I have been trying to find a good summary for the usage of popular classifiers, kind of like rules of thumb for when to use which classifier. For example, if there are lots of features, if there are millions of samples, if there are streaming samples coming in, etc., which classifier would be better suited in which scenarios?

Any help is appreciated.

  • 5
    $\begingroup$ This is too broad to be answerable. Different classifiers will perform best depending on the true data generating process, how the classifier will be used, etc. Your best bet is to read some books that discuss this. The Elements of Statistical Learning is probably a good place to start. $\endgroup$ – gung - Reinstate Monica Mar 6 '15 at 18:39
  • 1
    $\begingroup$ @gung I agree that the question is too broad, but Marc Claesen pointed the cheat sheet that summarizes the topic in a very convenient way so it convinced me to retract my close vote. $\endgroup$ – Tim Mar 6 '15 at 18:50
  • 1
    $\begingroup$ @Tim I agree the answer by Marc Claesen provides a convenient summary, however I think the question itself is too broad. I don't a single answer can address it. $\endgroup$ – Ellis Valentiner Mar 6 '15 at 18:58

Rules of thumb can only get you so far, but scikit-learn's cheat sheet is quite helpful for basic guidance. Here's a blog post by the creator of said diagram.

enter image description here

  • $\begingroup$ @AD.Net yes, that is a nice cheat sheet but don't forget that the reality is more complicated than that. So better use it as an aid but start with a statistics handbook (e.g. the one pointed out by gung in his comment). $\endgroup$ – Tim Mar 6 '15 at 18:55
  • 1
    $\begingroup$ As an ANN fanboy I'd like to point out that this sheet contains nothing about ANNs/Deep Learning so it really is incomplete (as the blog post itself admits) $\endgroup$ – runDOSrun Mar 6 '15 at 19:18
  • 2
    $\begingroup$ @runDOSrun yep, it is definitely incomplete. Other common approaches like Naive Bayes, decision trees and even logistic regression are omitted. $\endgroup$ – Marc Claesen Mar 6 '15 at 19:25
  • $\begingroup$ @AD.Net, if this diagram sufficiently resolves your question, please consider accepting it. If what you need is more than this, your question is probably too broad to be answerable. $\endgroup$ – gung - Reinstate Monica Mar 6 '15 at 20:09

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.