I have started going over the code and the simulated data and trying to replicate it all in python-pandas in a local repo to ensure I understand all the steps and can derive things on my own.
I have come across this issue where in the textbook, as well as line 58 of implicit_rating_calculator.py, you designate "moredetails" as w2 (50), but in line 93 you have it switched with "details".
There's one other issue I will create a PR for once I find it, but in listing 4.1 of the textbook, you do a count(CASE... using moredetails as the event, when in fact in should be moreDetails. This is causing the simulated data to contain all zeroes under moreDetails when counting events for users. I have checked this locally and can confirm. This will skew all implicit ratings.
Hello Kim,
I have started going over the code and the simulated data and trying to replicate it all in python-pandas in a local repo to ensure I understand all the steps and can derive things on my own.
I have come across this issue where in the textbook, as well as line 58 of
implicit_rating_calculator.py, you designate"moredetails"asw2(50), but in line 93 you have it switched with"details".There's one other issue I will create a PR for once I find it, but in listing 4.1 of the textbook, you do a
count(CASE...usingmoredetailsas the event, when in fact in should bemoreDetails. This is causing the simulated data to contain all zeroes undermoreDetailswhen counting events for users. I have checked this locally and can confirm. This will skew all implicit ratings.Cheers!