To give you some backstory, there is a particular publically-available hospital dataset that has been published about by other groups. I have recently come up with a novel machine learning method for predictive modeling within the dataset, and it is here where I am not sure where to go.

I believe that what I have is novel and significant enough to publish a paper about. However, I have more of a clinical/biology background and no formal background in machine learning. What would be the best way to go about the publication process here? I'm fairly certain in my ability to write the whole paper, but I am somewhat worried that there may be a technical mistake that may not be apparent at first glance. I'm also not even sure who is an "expert" that I could contact about this.

Any guidance is appreciated!

  • 3
    find a collaborator who has a background in this new field?
    – ACarter
    Commented Nov 20, 2023 at 15:59
  • 2
    who are you? a student? a phd? a postdoc? a practitioner? you have peers, now you need to find peers from a different field. If you have universities close-by, put an ad on the billboard of Computer Science, or of Informatics, or of Statistics (best ML people are statisticians ... they will tell you why you do not need it). Think about time: how much time do you need to learn pitfalls of ML? how much do you value your time? and your free time? offer compensation of at least 1/2 of your free time. FYI: I usually value my free time as 5x my working rate ;)
    – EarlGrey
    Commented Nov 20, 2023 at 16:51
  • 6
    "but I am somewhat worried that there may be a technical mistake that may not be apparent at first glance" - another thing that might well happen is that your method already exists in the literature (maybe under a name that you wouldn't expect, and maybe in statistics rather than ML). Commented Nov 20, 2023 at 18:26
  • You may need to talk to a machine learning expert.
    – Nobody
    Commented Nov 21, 2023 at 7:43
  • ML academic here. Is it that your approach is very interesting for that particular data set? Or is it that your approach is applicable and potentially yields great results in other data sets as well? Then are you worried that you make a mistake on the method (math) or on the evaluation process? For feedback: One of the things that you could do is that you could send your paper to a conference in order to get as much feedback as possible (reviewers and then audience). Commented Nov 24, 2023 at 7:54

1 Answer 1


I agree with the consensus in the comments: you really should try to collaborate with an expert in Machine Learning (or, as I would really recommend, in statistics, but I'm biased).

On the one hand, as Christian Hennig writes, there is quite a chance that you rediscovered something that is well known in the field, probably under some name you simply don't know - either as something that works, or as something that doesn't. That in itself does not mean you don't have something publishable, you might certainly be able to contribute something like a new application, or offer a new variation on an existing tool. However, you really need to speak the ML language. Reviewers will not like it if you propose something they already know; it will look like you didn't do your homework. ("Doing your homework" is pretty much what we are recommending here.)

On the other hand, an ML/stats expert will be able to suggest venues for publication, write the paper in a way that is commonly accepted in this field, request useful reviewers, and so forth.

So your question is really how to go about finding such an expert who is willing to collaborate with you. And that is a different question. We might already have something on that here at Academia.SE, try searching. How you go about this will depend heavily on your situation, on whether you are a student (how far along?), or an established researchers, whether you are at a university or at some other institution, or a private researcher, whether you are a member of some professional association that might have a ML/stats interest group, and so forth.

  • 1
    Thank you to you and all the commenters! This is the advice I was looking for. Commented Nov 21, 2023 at 18:28

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .