Anyone else having a small difference between LB & CV score? if so what's your highest CV score?
Same thing happened to me I got a higher accuracy score but when submitting to my surprise I got a bigger error rate (hence a higher score) !
same to me, i wonder why
it's nearly impossible for the cv to match the lb, and if it does, something is probably wrong. small differences are fine, big differences are worrisome. The difference in this particular challenge was probably due to small differences between train target distribution and test target distribution.
alright thanks much excuse me i'm a newbie to this nice platform and i kindly want to know which particular EDA i can use to try improve my model