feature_engine
feature_engine copied to clipboard
entity embedding for categorical features with large cardinality
I think this would be useful. I would know how to do this using TF + Keras. Not sure if this could be used?
Hi @csetzkorn
thanks for creating the issue. Could you add some links or more details regarding what this functionality is about?
The technique tends to be used in the nlp world and is know there as word embedding. Here are some links:
- https://gitlab.com/praj88/deepembeddings/-/blob/master/Scripts/deepEmbeddings_Keras.ipynb
- https://medium.com/@roeibahumi/keras-regression-with-categorical-variable-embeddings-dfc28616e7fe
- https://mmuratarat.github.io/2019-06-12/embeddings-with-numeric-variables-Keras
- https://www.youtube.com/watch?v=EATAM3BOD_E&list=RDLVOuNH5kT-aD0&index=4
Thank you @csetzkorn !