#na# in dls.classes["title"] in Collaborative filtering

Part 1 (2020)
1

I noticed that the first element of the dls.classes[“title”] is an “#na#”. Same for dls.classes[“user”].

Can someone explain where the #na# comes from? I do not see any missing values in the ratings Dataframe for either of the 2 columns (“user” and “title”).

2

In the notebook, only one batch is printed for reference but the #na# may be present in some row in the full dataset. Did you try searching the whole dataframe?

3

Yes, I searched the whole ratings dataframe for missing values and found none.

Also, the length of the dls.classes[“title”] matches that of the embedding (movie_factors) which makes sense and gives me some assurance that each line of the dls.classes[“title”] corresponds to each line of the embedding matrix, however I do not know what it means to have an embedding for #na# (first line of the embedding matrix.

1 Like
4

This is something I am also confused about. Why are there #na# values at the list beginnings of dls.classes for ‘user’ and ‘title’? This question comes from the collab chapter, i.e., chapter 8.

Thanks!

5

Yes! Especially since I do not see any nas in the data itself. If anyone can help, we’re all ears!

6

How do I make a post on my own topic?

7

Hi everyone,

I noticed that the docs show #na# in the example, but it’s not clear why it appears. In my case, I’m also getting this value, even though there are no NAs in my original DataFrame.

Any insights on why this might happen?

Thanks in advance!