I'm looking for a dataset that has some properties:

  • it has at least one feature that is categorical and that takes more than 1000 different values (when I say 1000, it's an order of magnitude, it can be 900)
  • these features' values have a meaning in english (like city names, ingredients, auto parts...)
  • it has to contain a target value so that it would be possible to make predictions that have some meaning

I'm looking for a dataset like this because I want to test different methods to handle categorical data (one-hot-encoding, embeddings, knowledge graphs...) and do some benchmarks.

I already found one dataset that correspond to what I am looking for, but I would love to have a choice between several. Here it is: https://www.kaggle.com/c/whats-cooking-kernels-only/

I looked on Kaggle, OpenML and several "Awesome lists" of datasets on GitHub, but I did not find anything except the one dataset I talked about.

  • Would the census.gov data do what you want? They measure many values, though I'm not sure they really have a "target" value. – user3856 Aug 9 '18 at 15:48

Open Payments Data are maintained by the US gov and contain a public record of payments from drug manufacturers and distributors to medical providers. They contain different features, including names and locations. You can, for example, try to predict who is paid above some amount.


| improve this answer | |

Ok, I found some datasets on Kaggle.

Here are some examples of datasets that I have found that correspond to my criteria:

| improve this answer | |

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.