I have a clustering task at hand. The data that I have contains only categorical variables. So, k-modes seemed like the best option. But I am not sure what are the data pre processing steps required for the same ?

What I am doing right now is the following:

  • label encoding features which have ordinal values.

  • one hot encoding the others.

and that's all I am doing as part of data preprocessing steps. My feature space gets to 50 from original 4 after doing above mentioned steps. I am getting 17 clusters as best number of clusters for silhouette score 0.60.

Also, I think doing Principal Component Analysis (PCA) to reduce dimensions and feature scaling doesn't make sense here as if I do this, I might as well use K-means. Would it be a good decision to run PCA and then use k-means for categorical variables ?


No - PCA and k-means can not be used on categorical variables. Both PCA and k-means require numerical variables.


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.