Don't Replace Missing Values In Your Dataset.

Показать описание

Everyone knows they must replace missing values in their dataset before training a machine learning model.

Most people, however, miss one critical step.

This video will show you what you are missing and how to do it better.

📚 My 3 favorite Machine Learning books:

Disclaimer: Some of the links included in this description are affiliate links where I'll earn a small commission if you purchase something. There's no cost to you.

Рекомендации по теме

Комментарии

Your posts (Twitter + Youtube) are more helpful than any other content for gaining intuition about data. Brief and excellent! Thank you, Santiago!

kemalariboga

this channel will be a gem in times to come

ashwinshetgaonkar

You're so brave man! For real, well done! Keep it up, we will follow !

dimasveliz

Great content, clear, intuitive and to the point. Refreshing to see this kind of content not 30mins long...

adkh

Followed your twitter, signed up for bnomial once it was launched, and now I am in love with your channel :) Thank you for the value you are creating.

hanweiz

That's a great point I learned today..
Thank you man....

sandeeptuluri

Awesome content. I have a suggestion. In your bnomial series sometimes the readers haven't covered a certain topic so it'd be helpful if after giving them the feedback you could link them to a good resource that explains that concept or may link them to one of these videos. It'd be a great help.

arskas

Super stuff 🔥🔥. Keep this thing rolling

sahanakaweraniyagoda

Great video. Always telling my students this and really hoping they stay aware of this in the future!

Fransphoenix

Incredibly insightful. Can counting the number of unanswered questions (e. 3, 0, 0, 1, 0, 2...) work too?

samarthsaxena

Thank you. Rarest kind of advice in ML field that I ever got (not like I have been in the field for too long anyway, still an undergrad student). I have questions though. That means that given N columns-table, the maximum number of columns possible is 2N right? Also, what if we just replace the missing values of categorical columns with a new category? Do you think the idea/intuition still works? Because I think that adding columns might increase the cost especially in a very large table with massive amounts of both row and column.

rinogrego

I had a survey I was working with that had a bunch of check boxes, and the data was 1 or missing. This example pretty much blows up all standard methods.

edmundfreeman