I'm working on a topic in machine learning, and along with formal published journal and conference papers, there are some different approaches which have yielded decent results in Kaggle - but these aren't published in any formal way, the top results are just listed on the kaggle leaderboard and the approaches are described in the discussion section of the competition.

On one hand I feel my review would be incomplete without brining up these approaches, yet I have no idea how to reference them?

On the other hand, I wonder if the lack of rigor that's typical when solving a Kaggle problem (lots of ad-hoc stacking and feature engineering,...) means the results can't be used as a reference anyway?


I have cited these resources before in academic work. Usually I cite them in a general and holistic way, not a specific way.

Several of the results presented in this paper were inspired by the discussion and resources on the Iowa Housing Project contest on Kaggle.com [citation with URL]. These resources were invaluable in determining a new approach to fitting a random forest model as found in this work.

I would avoid citing specific solutions, as these tend to be transient and informal. Such generalities as the one I've given above provide you with a blanket cover for any code which might resemble code on Kaggle. Citations are usually done for two main purposes:

  1. Provide attribution and recognition for a researcher.
  2. Provide insight into the research processes of the author of the paper.

The anonymity and changing nature of resources on a site like Kaggle make it very hard to track citations attempting to accomplish the first purpose. Thus a general citation will at least give insight into how you wrote the paper and what though processes you went through.

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.