I was wondering (about a more semantic question), is there a difference between data-driven methods and machine learning? Or is it more correct to state that machine learning is a category of data-driven methods (and what then are other categories)?

  • 1
    $\begingroup$ Can you give an example of a data-driven method that is not a machine-learning method? $\endgroup$
    – ginge
    Commented Jan 27, 2017 at 18:10
  • $\begingroup$ You could see a simple t-test as a data-driven method, but I don't perceive this as a machine learning method (although of course they are related). $\endgroup$
    – Archie
    Commented Jan 27, 2017 at 18:20
  • 2
    $\begingroup$ stats.stackexchange.com/questions/6/… $\endgroup$
    – Emre
    Commented Jan 27, 2017 at 18:39
  • $\begingroup$ Thanks for the interesting discusion, though perhaps not conclusive for this question? $\endgroup$
    – Archie
    Commented Jan 27, 2017 at 19:04
  • $\begingroup$ "data-driven method" doesn't seem to be a well-defined term. so this question seems rather opinion based... maybe you can give a reference to where you came across this term repeatedly $\endgroup$
    – oW_
    Commented Jan 27, 2017 at 19:55

2 Answers 2


Based on the quotation you have added in your comments, data-driven approaches are approaches where you use data that describes past states ("historical data") to get a (not defined) system to give a desired output.

To understand whether this definition includes machine learning or not we will have to define "machine learning", and while there could be plenty of ways to define it I expect that it will be quite difficult to come up with a definition that does not include within it "Using a system that, based on given states will give a desired output".

Note that in this last definition I use "given states" and not "past states" as to include approaches such as online learning.

Bottom line is that unless you really want to hold to a narrow definition of "past states" it seems that machine learning approaches are a subset of data-driven approaches.


In my view, Empirical Likelihood method is a very data-driven method but it has nothing to do with machine learning. Here is a link talking about the Empirical Likelihood method:

Empirical Likelihood Method


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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