I recently participated in the hackathon. The dataset includes drug name, sentiments about the drug, unique id. The target variable is the sentiment. It was a sentiment classification or analysis problem.
The drug feature has 413 unique names. But majorly data was about 15 drugs. Rest has only one or two rows. So, I made bins out of it and then did the encoding.
Now, my question is: should I do encoding or vectorization of these variables?
If encoding then, which one : label encoding or one hot encoding?
once encoding is done how to combine the text vector features and this encoded feature into a model