Serious question: how much data do you need before data science is useful? I doubt there's a hard and fast rule, but I can't help remember that statistical significance testing, a cornerstone of modern statistics, was invented to make better decisions about industrial processeshttps://twitter.com/JohnDCook/status/1348658451665322005 …
-
My rule: 1) literature/google/net search according to business problem. 2) has someone done something? Y: good, try w/ their rules. N: make my own rules according to business owners requirements and internal data. Problem framing takes time but it is definitely worthwhile..
-
-
Regarding meaningful data size: hardly ever. More specifically I found twice about satellite imagery and other few w/ talkings to business analysts (net or colleagues). Also, it was weird to find out that in some specific problems the model architecture played a bigger role.
-
Eg: lending prediction - time series problem. LSTM was better with more data (despite lending system change and data quality - totally weird); Random forest was better with less data (only considering current lending system).
End of conversation
New conversation -
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.