What happens if RF n_estimators(individual decision trees) have conflicting outcome as in 50% of them voted/predicted class A while the other 50% voted/predicted class B. In this situation, what would be the final outcome??
Many of you have asked me to share my presentation notes, and now… I have them for you! Download all the PDFs of my Notion pages at www.emmading.com/get-all-my-free-resources. Enjoy!
Can you clarify how the random feature subset selection happens "without replacement"? Is it that e.g. we have 20 features and tree 1 takes 10 features, tree 2 takes the remaining10 features and now tree 3 can take 10 from the original 20?
Hi Emma. Thanks for the video. Have a question. I am not sure about how this statement is true. "random forest constructs a large number of trees with random bootstrap samples from the training data". If sample size = replacement, we have all observations in every bootstrap sample. Then, it's not random bootstrap samples. Can you please elaborate what that line is saying?
Great Video! Thanks for making it. One minor comment is that at 6:56, sigma^2/k is actually not from CLT, essentially it's just from the basic property of variance.
Keep up the awesome work!, Emma I watched your video one year ago and I got a data science job. Now I start to forget some ML models that I don't use often, it is a very good way to refresh my memory on them!!!
Very clear and informative
Very clear and well structured
Thanks, Allie! Glad you found it helpful. 😊
Would you mind sharing the notion page with us? Would really appreciate it. :)
Of course! I'm working on getting all notes organized and sharable in one location, will let you know as soon as they are ready! :)
What happens if RF n_estimators(individual decision trees) have conflicting outcome as in 50% of them voted/predicted class A while the other 50% voted/predicted class B.
In this situation, what would be the final outcome??
Up to your logic at that point. But if that is a common occurrence in your model, perhaps try increasing the number of estimators.
Many of you have asked me to share my presentation notes, and now… I have them for you! Download all the PDFs of my Notion pages at www.emmading.com/get-all-my-free-resources. Enjoy!
Can you clarify how the random feature subset selection happens "without replacement"? Is it that e.g. we have 20 features and tree 1 takes 10 features, tree 2 takes the remaining10 features and now tree 3 can take 10 from the original 20?
No. It means any tree in the forest has no duplicate features.
Nice lecture, can we get the resource you used. It will be very helpful.
Very clear explaination! Thank you so much!
Hi Emma. Thanks for the video. Have a question. I am not sure about how this statement is true. "random forest constructs a large number of trees with random bootstrap samples from the training data". If sample size = replacement, we have all observations in every bootstrap sample. Then, it's not random bootstrap samples. Can you please elaborate what that line is saying?
Great Video! Thanks for making it. One minor comment is that at 6:56, sigma^2/k is actually not from CLT, essentially it's just from the basic property of variance.
Looking forward to the notes!! Thanks for sharing, Emma!!!
can we say if interview ask which algorithm can be used here , and we don't know the Ans we can surely apply random forest here.🤔😜
x)
A very helpful video on RF. Hi Emma, would you mind actually making a video on how to go about mastering new ML concepts from zero to hero?
Thanks for the suggestion, Ayuumi! I'll add it to my list of video ideas. 😊
Awesome video. Super helpful.
thanks!!1
Love it!
great video!
Thanks for the kind comment, Yunge! 😊
Keep up the awesome work!, Emma I watched your video one year ago and I got a data science job. Now I start to forget some ML models that I don't use often, it is a very good way to refresh my memory on them!!!