Why and When Should we Perform Feature Normalization?

[@simon20002]

clear, concise and exhaustive. Very useful, i appreciate your work. salutations from Italy!

[@akshaykrishnan7985]

Superb sir.. This was the video I was looking for. You cleared all my doubts related to feature scaling in the video

[@VVV-wx3ui]

Thanks Krish. Such insights are the key and makes the difference. Thanks once again for Sharing with All.

[@anubhavsood1510]

Hi krish, Could you please make a video on end to end steps to be followed for a ML project. For example... 1. Import dara 2. Impute data 3. Remove outliers 4. Perform univariate EDA 5. Train test split 6. Scale features 7. Call algorithm classes like KNN, random forest, linearregression... 8. Perform k Cross validation to validate which algorithm is suitable 9. Perdorm feature reduction 10.perform feature selection technique

[@anujtiwari5084]

hey brother how are you doing ? I have started learning machine learning few months ago, seeing your 4 year old comments makes me curious , how are you doing right now? are you working as a data scientiest , a AI developer or you have changed your field somewhere else?

[@mrrasel871]

Just amazing ❤️

[@kalangiprasad4073]

9 min great explanation sir.

[@chimadivine7715]

Quick and concise. Thank you.

[@gvsmchaithanya2847]

Feature Scaling will be applicable to SVM also I think because in that mostly i taken the data set like the binary classification of the dataset like 0 or 1. In this case of that problem and i have also applied min - max scala by using sklearn library. I think my usage of feature scaling what i applied this may correct.

[@anandacharya9919]

Thank you sir for this one more excellent video. I request you to make a video, on how to choose a correct machine learning algorithms . Thank you very much!

[@maxwellpatten9227]

I LOVE these videos!!!! Excellent!!!

[@syedtasleem2827]

Sir, please try to solve analytics vidhya amexpert hackthon challenge problem..!!

[@the_imposter_analyst]

This was so helpful!! Thank you so much sir.

[@louerleseigneur4532]

Thanks Krish

[@kaushikumang]

Thanks for the information sir.

[@naehalmulazim]

Sir, your lectures are always very helpful. Have a subscriber. But maybe you might consider writing some of the points on the whiteboard. It became a little hard for me to keep up.

[@siddarthbali12]

You mentioned that where ever we use euclidean distance we should apply feature scaling, but what about the computation that involves the cosine similarity, does we should perform normalization in that case and the also the score return by the cosine similarity should be normalized or not???

[@AmitYadav-ig8yt]

How about classification problem problems or In Algorithms like Logistic Regression ? Can we use Feature Scaling?

[@ashraf_isb]

Thank you

[@junaidlatif2881]

Sir. I used my data for RF without scaling. Model performance is worst and in - ve value. Because my Y (output) variable range from 10E14-10E20. I tried many ways. But i found better accuracy when i used standard scaler of Y variable. Other ways my model performance is worst! Still my question is: if i used scaler in RF and ensamble method is it right? Can I use it?

[@husamjon5971]

so we do scaling either for ensuring correct results (like K mean algorithms or for efficiency like anyone which uses gradient decent - correct?

[@ayubaalim2201]

helpful..............my DREAMS from Somalia

[@pravinacharya4411]

Do we need to do scaling for time series data and which scaling method would be best

[@ishantyagi2701]

can i apply standariztion for logstic regression

[@muhammadusmanakram406]

can you make a video on how can we make our own dataset and how much rows will be enough for that particular dataset??

[@ashishsangwan5925]

Awesome 👍 Do we need to do feature scaling in logistic regression?

[@VVV-wx3ui]

yes

[@atiladursun7075]

Hi Karish, do you do Normaluzation before pca or after? Thanks

[@sunnysavita9071]

sir please upload the various feature scaling technique practicals

[@divyanshchaudhary7063]

what to use first Scaling or Transformation ?

[@rusteze4710]

What happens when we apply feature scaling when there are outliers?

[@krishnaik06]

First we need to exclude the outliers through some way.

[@rusteze4710]

In few cases especially health care domain outliers are very important what happens when we scale?

[@vishalkap62]

Use robust scaler..

[@ASNPersonal]

What about Logistic Regression, SVM, Naive Bayes? We should apply standardization or not required?

[@hemantsah8567]

Is course started?... where to pay ?

[@mahery_ranaivoson]

Do we need to scale values we have when we do prediction?

[@VVV-wx3ui]

Its required based on the Algorithm pack one chooses. Except Tree Based and Ensemble Algorithms, for the rest it would be required. In Case of CNN, we do it differently as the value can be between 0 and 255 (pixels).

[@mahery_ranaivoson]

@@VVV-wx3ui Yes CNN obviously it required. But how we scale the unseen sample given only the built-in model(No knowledge of mean-max-min anymore). Assume the model is already deployed, how can we handle unseen un-scaled, unnormalized, ... ?

[@VVV-wx3ui]

@@mahery_ranaivoson the unseen future input for prediction has to go thru the same transformation that the train data has being put thru. We provide the interface for prediction and so, should be transforming it for the prediction to work as trained. Hope this clears.

[@hridayanandadas3528]

I have trained my model using feature scaling. now i want to predict the real data. do i need to scale whose data also?

[@akshaypratap7390]

Sir plz edwisor k baare m btaye

[@RamCharan-bn5cb]

Bhai I am in ML 3 group how to pay amouny

[@vineshcool5826]

Can you please explain on local outlier factor concept. It will be helpful..

[@MrSawan27]

Quite useful information. Can you also explain when to use which type of scaling mechanism like standard scaler and minmax scaler?

[@anubhavsood1510]

Between these two, it doesn't matter much coz we are scaling the features to same scale. However, if you want outliers to have less effect on your data, go for RobustScaler()

[@GAURAVRAUL95]

So can I say, scaling is just needed to improve running time of a model but it doesn't affect the accuracy or other metrics?

[@subhashgupta9847]

Please suggest some website or WhatsApp group where I can ask machine learning Doubts

[@krishnaik06]

Join here t.me/joinchat/N77M7xURBNyGxafNhCmnTw

[@vijaynale7893]

I am only waiting for next kaggle competition - House price prediction video , to see..

[@akhandpratap__]

2:10

[@arrow94]

Thobada dikhane se acha pdf thought dikhana