I am working on a project in which I use data of movies and I represent each movie as a vector of length of 15. So there are 15 features ranging from genre to director. Most of the features are categorical and I want to use target encoding, but I am not sure if I must weight target encoded data or they are kinda already weighted. What I am trying to make is a movie recommendation system which is fully unsupervised. One hot encoding is out of question for me as one movie vector becomes easily 1 million D. Also I do not understand the reason for weighting at all. I am a rookie. Can you help me understand? By the way, categorical and numerical data(such as duration of the movie) is placed together in a vector. I hope it does not cause problems as I am planning to normalize/constrain numerical values to 0-1 range.

  • $\begingroup$ You can one-hot encode individual features like if the director feature has 50 unique values, you'll end up with a 50D vector for that feature. Also, how can the movie vector become 1M D when you have only 15 features? $\endgroup$ – Shubham Panchal May 5 at 6:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.