Video není dostupné.
Omlouváme se.
What should you do with outliers?
Vložit
- čas přidán 14. 08. 2024
- An outlier is an unusual data point, a weirdo that's not like the others. What should you do if you find one in your dataset?
Here's what not to do - just prune it away because it looks inconvenient. Stop! This is data science, not bonsai. Instead, let's look at 4 options.
Learn more about regression analysis, mentioned in the video:
bit.ly/mfml_reg...
Learn more about statistics:
Blog - bit.ly/quaesita...
Video - bit.ly/statthin...
Minicourse - bit.ly/quaesita...
Learn more about machine learning:
Blog - bit.ly/quaesita...
Course - bit.ly/mfml_000
Don't forget to share, like, and subscribe if you'd like to encourage more videos like this one.
I love that you said upfront that there won't be a one size fits all solution, great content all around the board!!
Its better if she talked more about how domain experitise can be used to figure out what to do. The key is to understand the overall biz metric you are trying to move. If the outlier(s) represent a significant impact to your overall metric, then you absolutely dont want to drop them and in fact may want to understand them deeply and focus on the outliers. This is basically the 80-20 rule. If the outliers only represent a small portion of your overall metric and keeping them vs. dropping them has no real impact to the overall KPI metric, then drop them.
Beautifully explained!!
thanks a lot, you answered my question about "normal outliers" thank you very much
Very clear and informative
"What if it's not a mistake" more like "what if she actually weighs like an eliphant"
Awesome! The first thing people do is to get rid of them... shame!
I suddenly realise who really should have been on the Iron Throne. Brilliant.
That Marie Kondo meme lmao
Everytime, I struck somewhere. I get my answer here ❤️❤️❤️
Eureka moment! Thank you
very cool cassie
Excellent content
loved the content
Oohh. ..... she's looking like Anastasia from fifty shades of grey.
Welcome to theoretical finance: bubbles don’t exist!