Feature scaling should be avoided

akalikadien / shell-ai-hackaton-2023

Repo for the agricultural waste challenge of Shell.ai 2023 https://www.shell.com/energy-and-innovation/digitalisation/digital-and-ai-competitions/shell-ai-hackathon-for-sustainable-and-affordable-energy.html

0 stars 0 forks source link

Feature scaling should be avoided #5

Closed vsinha027 closed 1 year ago

vsinha027 commented 1 year ago

https://github.com/akalikadien/shell-ai-hackaton-2023/blame/96f4d4721210dd5bfbfc90faf08dc17a9c5ca04f/place_storage_and_refineries_optimize_flow.py#L85 I would avoid feature scaling. The kmedoids is stable without scaling also and then there could be issues in translating back the original values which are needed to calculate the exact biomass flow.

akalikadien commented 1 year ago

I checked this and with/without scaling the cost is in the same order of magnitude. Without scaling it was lower though, so I will remove it in the next commit.