wujcan / gcm Goto Github PK

Hi, I encounter a problem when I try to analyze the data.

There are many "0" in the city's values in NC datasets and I am wondering if 0 is the padding of this context feature.

In addition, I want to know how you deal with the missing context id in each sample in the data preprocessing.

Hi, when is the source code expected to be released, or is it possible to obtain a copy in advance?

Dear all,
Thanks for the repository.

I found the paper interesting and I am trying to reproduce your GCM work on other datasets.
I have been able to create train.dat and test.dat with needed structure: "u_" for user features, "i****" for item features and "**c**" for context features. Hopefully that is correct.

Could you please elaborate more on how item_dict and user_dict are created?
Why test.dat file contains only one test item per user? Does it work if we have more than 1 test item per user?

Looking forward to hearing back from you.
Thank you in advance.
Best regards.
Manel.