Show this thread
-
You should probably learn stats that way but you probably shouldn't actually use those things in prod because you're a lot more likely to have bugs in your code then heavily used libraries.
Something like Prophet is more than just a heavily used library though. It deliberately hides complexity/subtlety from the user. Complexity that might turn out to be important..?
-
-
This is the whole bear case for using mac. *ducks
End of conversation
New conversation -
-
Like, it’s great if you’re a sysadmin who needs to predict load/usage every hour for the next month. Less good if you’re planning to spend hundreds of millions of dollars in an adversarial environment based on the forecasts.
-
I think this is more a function of how they chose to enter the specific market. The library itself isn't really the issue and they likely would have ran into the same problems even if they ran their own ML implementation
- Show replies
New conversation -
-
-
tbh I don't know anything about prophet, obviously sometimes you need to make your own stuff but standard stats/math/models you're probably better off using a library. Less likely to have weird bugs and if someone leaves/is hit by bus/etc you aren't left in a bad place
-
prophet is just a regression on sin() and cos() for weeks and month, a logistic function time trend, and a radial basis for holiday dummies. that's it!
End of conversation
New conversation -
-
-
The devil is in the details.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.