search

DS420-Cobras / DS420-Cobras

Python code for our KDD Cup 2018 (Cup of Fresh Air) compeititon
The Unlicense
6 stars 0 forks source link

Feature engineer darksky forecast data (fill NaNs) #57

Open ghand1 opened 6 years ago

ghand1 commented 6 years ago

The darksky.net data has a number of NaNs. We need to fill the NaNs with maybe mean by day/station_id (for example)

LeoSalemann commented 6 years ago

First opportunity is to look at where we're dropping our existing data, due to nan's.

Next opportunity is around line 167 where it says,

ToDo: Undo. Removed because they have a lot of empty cells. Replace with more meaningful values

The other section is around line 213 where it says,

ToDo: Come up with a strategy for handling na values