Feature Selection Wrapper and Embedded techniques | Feature Selection Playlist

[@UnfoldDataScience]

Access English, Hindi Course here - www.unfolddatascience.com/store Don't forget to register on the website, it's free🙂

[@polinalee9128]

One of the best videos explaining the difference between wrapper and embedded models I've seen. You cleared up all confusion due to your balance of conciseness and level of detail. Bravo sir.

[@UnfoldDataScience]

Thank you

[@saharyarmohammadtoosky]

Very good explanation Aman, you are a good teacher, I follow your videos, very simple and understanding explanation, good luck!

[@UnfoldDataScience]

Your comments motivate me. Thank you so much.

[@mansibisht557]

Thank you Aman!! Such crisp explanation!

[@UnfoldDataScience]

My pleasure 😊

[@subhajitroy4869]

Awesome Sir!!!! Thanks a lot. You are a perfect Guru for any DS learner. Another request Sir, kindly make a detailed video on SVM. It would be really helpful for many of us.

[@UnfoldDataScience]

Thanks a lot Subhajit, sure

[@umasharma6119]

Thanku Sir for this great explanation.

[@pavansingara9408]

very good explanation of the concepts

[@UnfoldDataScience]

Thank you Pavan

[@israrobinson5175]

Thank you for this

[@sudhanshusoni1524]

thanks for the awesome work!

[@UnfoldDataScience]

Glad you found it helpful.

[@sriamani]

Very Informative video,i have some doubts regarding forward feature selection 1. PCA with forward feature selection 2. feature names we have to select, k_features we have to give exactly 3 or 4, then how algorithm will select,and which features will select

[@beautyisinmind2163]

Sir, the combination of feature you got in your result is applicable for KNN only or same combination works for other model as well?????

[@leamon9024]

Hello Aman, thanks so much for the detailed explanation. Could you also talk about clustering based feature selection technique?

[@Sagar_Tachtode_777]

Great video Aman! Thanks for sharing! Can u please tell which algorithm to use for product recommendation using demographic data like age, Salary, Gender, Occupation etc….???

[@alishaparveen4735]

Can we do wrapper method for feature selection in unsupervised learning data?

[@akjasmin90]

Hy, I really loved your video and appreciate your efforts in making such informative videos. I have 3 questions though. 1. In the video you have used the methods on numerical data can we use it on categorical? 2. We should use it before or after feature engineering? Like after making dummy variables and binning are data it requires? 3. In RFE -CV all the variables were showing as 1 i.e. important. Can you explain it a little bit? Or if you can direct me to some video.

[@UnfoldDataScience]

Thanks Ayushi. 1. Some test can be used on numerical only. 2. Before only 3. Try with other data this will change, here the difference is not that much.

[@Shivay_1357]

Thanks

[@fahadnasir1605]

Aman, you said, in RFE, it is internally decided how the variables will be eliminated and in backward selection, we are passing knn model to remove the variables. BUT, in RFE you are passing a Linear Rgeression model, please explain

[@FranklinKondum]

This is awesome! Please, I have a question: In the backward wrapper method of feature selection, how can I use my own "user defined" model. I have an already existing model, but i want to reduce the features. It is a linear equation: Y = 0.22D + 0.19E + 0.16F + 0.15G + 0.16H + 0.12K I want to do feature elimination without changing the coefficients.

[@ajaykushwaha-je6mw]

Hi Aman, once we get the number of importance feature then we have to remove unwanted featured from X_train and X_test right ?

[@UnfoldDataScience]

Yes, Both places. No need of these features ahead, just keep a track of what all we removed so that next time new data comes we know what to keep/remove.

[@beautyisinmind2163]

Sir do we need to apply all techniques(filter, wrapper, embedded) and see that which feature is important?

[@UnfoldDataScience]

Yes if you have the infrastructure to support especially if your model is not doing good.

[@skvali3810]

can we do all this techniques inside a pipeline

[@UnfoldDataScience]

You can do.

[@UnfoldDataScience]

And take a call in the end based on results

[@skvali3810]

@@UnfoldDataScience thank you man

[@zeeshankhanyousafzai5229]

Hello sir What if features are categorical and discrete?

[@UnfoldDataScience]

Test like chi square and some model based technique will be used.

[@zeeshankhanyousafzai5229]

@@UnfoldDataScience ok sir can you make a video on that as well?

[@umasharma6119]

If we have the domain knowledge I think we don't need to perform feature selection techniques ?

[@UnfoldDataScience]

Then also we need to see , domain knowledge is what we know, "Data must tell its own story"

[@umasharma6119]

@@UnfoldDataScience Okay Sir.

[@Fatima-gw7sm]

Cost of your data science course?

[@UnfoldDataScience]

Please fill the form attached in the description of the video.

[@Griddymerchant7]

hello Aman, pls can I have your personal mail

[@mohamedesmailelsalahaty6050]

Thanks