Using ROCK with Categorical Data

[PascalEconomouCoupa]

Hi,

I have a dataset with purely categorical features. After some research I found that the ROCK algorithm can be used to cluster categorical data, and has some advantages over other algorithms like K-modes.

I spent some time trying to get your implementation to work, before reading this closed issue #373 where you say that your implementation has been adapted and works only with numerical data. I was quite surprised by this, given that the original algorithm is specifically for categorical data.

So I have two suggestions: 1) If possible, it would be nice to have an implementation of the original algorithm that works with purely categorical data. 2) If not, it seems sensible to make it a bit more obvious in the documentation that the algorithm has been adapted and does not work with categorical data.

Thanks,
Pascal

[annoviko]

Hello @PascalEconomouCoupa ,

Yes, that's right, ROCK is adapted for numerical data only in pyclustering library. It is possible to support categorical data, but I do not have enough time to work on the library right now as I was doing it before due to personal circumstances, so I cannot promise anything right now.

Thank you for your feedback about the documentation, I will highlight this moment in the documentation.

[skullyhoofd]

This is still not clear from the documentation.