DATASCIENCEARCHIVE Telegram 94
说到特征降维/选择的问题,大部分EDA的套路都是从model训练的loss来判断feature importance。其实有一个简单易行而且很有效的办法是在CV里面用做feature permutation,对原始特征shuffle得到shadow(也可以加一些噪音),在通过zscore比较两者差异来判断importance,不断遍历筛选。在ESLII中593页有提到这个办法。R里面有一个包Boruta可以做这件事,py也有:https://github.com/scikit-learn-contrib/boruta_py



tgoop.com/DataScienceArchive/94
Create:
Last Update:

说到特征降维/选择的问题,大部分EDA的套路都是从model训练的loss来判断feature importance。其实有一个简单易行而且很有效的办法是在CV里面用做feature permutation,对原始特征shuffle得到shadow(也可以加一些噪音),在通过zscore比较两者差异来判断importance,不断遍历筛选。在ESLII中593页有提到这个办法。R里面有一个包Boruta可以做这件事,py也有:https://github.com/scikit-learn-contrib/boruta_py

BY Data Science Archive




Share with your friend now:
tgoop.com/DataScienceArchive/94

View MORE
Open in Telegram


Telegram News

Date: |

End-to-end encryption is an important feature in messaging, as it's the first step in protecting users from surveillance. There have been several contributions to the group with members posting voice notes of screaming, yelling, groaning, and wailing in different rhythms and pitches. Calling out the “degenerate” community or the crypto obsessives that engage in high-risk trading, Co-founder of NFT renting protocol Rentable World emiliano.eth shared this group on his Twitter. He wrote: “hey degen, are you stressed? Just let it out all out. Voice only tg channel for screaming”. 4How to customize a Telegram channel? The main design elements of your Telegram channel include a name, bio (brief description), and avatar. Your bio should be: Developing social channels based on exchanging a single message isn’t exactly new, of course. Back in 2014, the “Yo” app was launched with the sole purpose of enabling users to send each other the greeting “Yo.”
from us


Telegram Data Science Archive
FROM American