Standardization Vs Normalization- Feature Scaling 

Thank you Krish for this video! It was fantastic in helping me understand the difference between these 2 things and some additional advice regarding how it helps with some other things (e.g. helping some kinds of models optimize faster!)

One of the best and detailed explanation on Scaling Techniques. Thank You so much Krish ji.

I'm really in love with the way you explain. So nice :)

I have completed Statistics Playlist. You explained in a very good way. Thanks for this. :)

Thank you. I wish this world will be fulled of people like you!

I love you dude, thanks for the explaining you saved me, greetings from México wey you have a new sub

Incredible explanation, thank you very much!

the best video on Standardization & Normalization

You explain it very simply. I love it. I even even recommend your videos to other guyz in ML.

Hi sir. I have seen lots of videos on machine learning but I personally feel like u r d only one who’s making the videos in very fantastic way. u explains all the things in such a way that even the person who is from non technical background can understand it. Just a small req for you. Can u pls make video on all the techniques that can be apply on single data set. Like when to scale the data & apply PCA, clusters, algorithms, when to do label encoding instead of one hot. Can u pls apply all these things on any dataset so that i can have clear insight on model building. Can u pls make video on this for end to end model building

Brilliant explanation. Thank you sir!

One of the best teachings on this subject. Thanks Krish

Great way of teaching...really helpful!! :)

Great explanation. Very well conveyed with proper examples

Finally completed your statistics playlist and can definitely say learned much more than other online courses

Thanks for the intuitive explaining.

@Krish Naik Sir Github Link is not there in the description link, for the Jupyter Notebook shown in this video, Can you Share the Same ?? Thanks & Regards, CHINMAY N BHAT

Great!...very good explanation...plz keep posting...thanks

Hi Krish, one quick question. I was going through some tutorials for batch normalization and got confused with which technique is used there. It seem like they first do min-max followed by standardization. Can you please help me here?

I have gone through other speakers videos but they are hard to follow. I really liked the way of explanation in a very simple way with great examples. Thank you brother.

Thank you! Perfectly explained.

HI, can you please add the github link in the description? the github address is missing.

Clear message, clear structure, easy to understand, thank you

You're a great teacher. Thank you.

Thanks for your suggestion 🙏

Sir, you should also mention the dataset link in the description. This will help us to follow you.

Very informative and helpful, thanks a lot Krish

Great content. Thank you for explaining in the best way possible. However a small suggestion, please include the links of dataset your are using in the description box. It will be helpful to practice along while watching the video. Thanks again, cheers!!

Awesome sir, you are explaining very easy way .

So clear explanation, thanks Krish

This short video has helped me understand a great deal of feature engineering. God bless you. I wish to learn more from you. I recommend you do a video on a full data science project and focus more on the thought process. While you also do a soft touch on various alternatives to whatever method you have used. This is Great!

I hope one day I will become data scientist like you , you are really helpful for aspiring data scientist like me

Hello Sir, Generally the range for normaization is 0 to 1. But I have read papers where different ranges are used Say 1 to 7. I want to know the logic/criteria behind selecting the target range?

Hi Krish.. I recently use standardization in classifier model and it improve my accury. I am aware that for classification method standardization is not recommended but in my case its improving accury. What would u suggest?

thanks for your excellent explanation, but it confuses me when I try to filter lower-variance features, standardization isn't suitable because it scales all features' variances to 1, so in this case I should try normalization, but then is it right to perform PCR or PLSR next?

wow! great explanation .. Thank you 🙏

Thanks to you Chanel... it's so helpful for my UNI Lesson

Great Explanation, Thank you!

Hi Krishna.You're saviour.Apologies in advance if it is already asked question.(please advise if you have already answered and will find out the video). 1.Do you have any usecase where you do standardisation (with mean & std ) followed by min-max normalization so that you can compare same scalled features and then fit them into 0 to 1 or let's say 0 to 100 or -50 to +50 etc ? 2. any pros and cons of standardisation followed by min-max normalization ? 3. am i missing any logic by asking ? is there any solution for a scanario where you have more than 5 + features and user want it to scale in a single number so that instead of viewing the movement or change of 5 features,you will only focus on final score by means of min-max norm....hope it's clear out my question looking forward to see your answer.Regards & TIA + Thanks for this video.

Thanks for the awesome explanation

Hi, I really enjoyed the video. I was wondering is this the same as normalisation on keras.

thank you so much you are the man !!

Hi Krish, What will happen to notmalization if outliers are present in the data? Outlier treatment is necessary before applying notmalization? There a method in sklearn normalize, will it same as minmaxnormalizer.

Hi, Krish. For instance, I train my model on normalized dataset and I need to use this model. I have to classify ONLY ONE test object. How should I normalize this object?

Great explanation !! Thanks

best explanation, keep it up

sir, please make video on difference between GD,SGD,SGD (mini batch),SGD with momentum.

Thank you. Great explanation.

after fitting model and predicting values using normalised/ standardised data, how to get back the original values for predicted results. no where reversing of normalisation is shown??? have had any video on this?

Here you go. Hope it can help you guys df = pd.read_csv('raw.githubusercontent.com/rasbt/pattern_classification/master/data/wine_data.csv', header= None, usecols=[0,1,2])

What is the most adequate way of features scaling for ANFIS algorithm, normalization or standardization?

Krish whenever i get confused for any Data Science topic, i search it on YT, if your video pops up for it, i definitely select your explanation for that topic.

I had been watching all your previous statistics videos and understood each concept well. Since I am not from mathematics background, In this video I couldn't understand what you explained in the part while telling what process to use when. Will this be a matter to bother in my data science learning journey?

Thank you sir..nice explanation :)

Good one 🤘🤘🤘 Actually z score is much widely used for most of the algorithms as i have seen. And I do practice the same all the time. The reason is the affect of the outliers. Outliers can be easily detected by z score. Normalistion between 0 to 1 just shrinks curves.

sir if the features are not normally distributed and we apply standard scalar then does it become distributed according to a standard normal???

Thank you this was very helpful.

I love you mate, thx Cheers!

I am trying to build a digit recognizer using SVM. So I should use Min max normalization right sir ?

@krish - i didn't quite understand when to use Normalisation and when to use Standard Scaler. Can u share with an example why standard scaler was used and another one where normalizer or min max scaler was used, and why.

Great Job very well explained

Hi...Please clear my doubt that data normalisation or standardisation should be done along the column i.e along each dimension of data. Right?. I'm a little bit confused, whether to do row-wise or column-wise data normalisation in excel. Please reply in this thread.

Today only started this playlist and today only completed, it is possible because the way sir❤️ explain is just amazing..❤️ Now I move to next part.

super guru! U made it a cake walk

sir my db is in mysql using MYISAM.... No foreign keys implemented with some columns repeating in tables as well... i have approximatelly 40 tables and all containns approx 500 rows.... but two or 3 tables contains above one million rows....db is not normalized.....but it is still working fine...on LAN and on cloud as well ... what will be fuiture of my db... can u help me...

Thanks Sir for sharing all wonderful videos, kindly provide the github link to download the dataset ,not getting from description box

Thank you so much !!!

thanks for the great explanation sir .the link of channel you are talking about is not working pls help with that

In most of the cases I reproduce this kind of videos at 1.25x velocity, this one 0.75x haha nice videos Krish!

always the best explanation!

Absolutely excellent explanation

Krish Naik, what if the data has outliers. When we do normalization or standardisation, then the extreme value will probably get a value close to 1, due to this, most of the data gets assigned close to 0, how can we handle this ?

Thanks a lot Krish :)

Thank you for your video Krish, it was really helpful!

Very informative thank u

How to normalize gridded datasets (X1 and X2) for neural network in python?

Paji tusi great ho :)

Commenting exactly on the same date, such a coincidence though, Thank you Krish for this video!

Excellent explanation!

Nice and useful information 👍

Is scaling required when we are doing regression with Decision trees (Including RF, GBM. XGBM etc.,)?. I got this doubt because of variance is key to split the node in Regression case. Please correct me In case I am wrong. Thanks.

Thank you for the video. It was useful. Can you please provide the github link?

Have to say that your presentations really stand out , basically because of the distilled informations and to the point suggestions you make. One question though. At some point you talk about CNNs and that we have to use MinMax scaler. I am using CNN but on non image data, basically I see my data as an image of point values. Should I go with MinMax scaler or I could also use Standard scaler? And in order to be more specific lets say that I have an image of 7x7 where I want to keep the relative value differences between a pixel and its neighbours. Which scaling should I use in your opinion? Can we use standarization on the dataset in order to train a CNN or the values should be in [0,1] so we have to use minmax scaler. I am really interested to hear your opinion based on you experience.

hello krish .awesome video .But lease provide the GitHub link for practice.I have searched over in GitHub profile of yours but could not find it

Great explanation!

what are the algorithms in which gradient descent is not used?

Thanks sir for explanation

I like your why of explaination

Sir my test data and training data has different distribution ...what to do sir

Can you send the Github link for this code

Good evening sir.. I have a doubt.. When you say collecting data, how exactly is it done? Is it done through market research methods or any other methods are there for collecting it? Please do elaborate..

Great explanation Sir

does transformstion required affter standardising the data to make it gaussian?

Thank you sir ♥️

In linear regression, one common assumption is that all the features have 0 mean same variance. Which is similar to standardization. Hence it works.

Clean explanation Thanks

Pls tell difference in machine learning , AI , Deep learning n NLP

Great explanation

Should the features are related to each other if we want to normalize or standardize them?

can somebody tell me that it is possible that standard normal distribution cannot be normal that is mean is not equals to median for standard normal distribution