nDCG per user other than per fold about rrecsys

[ludovikcoba]

That wouldn't be that much trouble for me. But consider the fact that I am doing k-fold cross validation, so you will get k values of the nDCG for each user.
If you are interested only in the nDCG(on each user), I can implement a new function only for that.
Let me know.

[jonnylee719]

Yea, I had a look at the code, there are k values of the nDCG for each user. But you normally only get the nDCG per fold as the result rather than the individual k values of the nDCG for each user or the nDCG (on each user). Please do tell me if I am mistaken.

[ludovikcoba]

You are right, the nDCG function computes it on each user and later it is computed over the entire fold(the reported value) on the evalPred function, this operation is repeated k-times.

[jonnylee719]

So do you think it would be useful to also provide individual nDCF per user so that we can compare the performance between algorithms by statistical tests like t-test or Wilcoxon rank sum test?

[ludovikcoba]

Depends on what you are trying to achieve. In this implementation I followed literature and as well other implementation in popular recommender environments. Let me point out that this is not an Issue but it stands more as a request, which I would be glad to help you out with.

[jonnylee719]

Yea it is a request, and I am very grateful and glad to hear that you would be able to help. Thank you! And I apologize if I came across as being ill-mannered at any point.

[ludovikcoba]

You weren't ill-mannered at all, but let me close the issue and we can continue the discussion via e-mail.