0

I have been working with a basic CB rec sys and I already made this steps:

--Create TF-IDF vector for each item. --Get user profile according to the ratings each user has given.

And after that it seems that I can just get the similarity between the user profile and the items features vectors which fine to recommend or get a ranking. However, I need to compute a prediction for the rating because I will have to compare the results in terms of RMSE among different algorithms. In this case, I need a value (1 to 5) so I can calculate the error. What would be the best formula to use?

Daniela Varela
  • 46
  • 3
  • Hi, it would be helpful to write answer when seeing the piece of your data – puhuk Jun 25 '20 at 00:51
  • 1
    I am working with many datasets, all of them have ratings (user,item,ratings) and some item content. For example, the description of a book. – Daniela Varela Jun 25 '20 at 12:47

0 Answers0