Open AnhPC03 opened 4 years ago
i don't think you should split the data set. i think that just eliminate the first recommandation should be enough, the more data you have the better the model should performe. in this case the classical "overfitting" problem is a good thing , you want the most specific racommandation for that particular movie.
for testing the recommandation system you should have a set of movie with a predetermine racommandation and maybe watch how similar the "human racommandation" and the knn racommandatio are, but i think that this is very impractical
How can i evaluate the book recommandation system? Must i split the dataset into train and val or test set?