Galina E
31 posts

Galina E
@gendcode
Data scientist and bookworm on the lookout for next travel destination

Holy crap. They're not accounting for weekend periodicity and therefore are overfitting the drop in reporting from last weekend and predicting as if that's going to continue. This is "worst student in freshman stats"-level malpractice, and people are going to die from it. Unreal.


Talk done at #Big Data and #AI World this morning. Will now be preparing to moderate an afternoon panel discussion on "driving revenue whilst creating a positive social impact" #bdaiw #womenintech





Suppose workers at company X strongly prefer to work with men. So, the algorithm learns men do well at company X, and women don’t. The algorithm will be predictive of performance, but not for the expected reasons. This is why naive algorithmic optimization is problematic.






