Feature Scaling - Normalization | MinMaxScaling | MaxAbsScaling | RobustScaling

[@ayeshasyedKhan]

You make learning so enjoyable and accessible! Your clear explanations, engaging examples, and passion for teaching truly set you apart. Thank you for putting so much effort into creating content that not only educates but also inspires. You're the best teacher on KZbin-keep up the amazing work!

[@fit_tubes_365]

Course Started : ML Lecture-01: 14/08/2024 Lecture-02: 14/08/2024 Lecture-03: 14/08/2024 Lecture-04: 14/08/2024 Lecture-05: 14/08/2024 Lecture-06: 15/08/2024 Lecture-07: 15/08/2024 Lecture-08: 15/08/2024 Lecture-09: 15/08/2024 Lecture-10: 15/08/2024 Lecture-11: 16/08/2024 Lecture-12: 16/08/2024 Lecture-13: 17/08/2024 Lecture-14: 17/08/2024 Lecture-15: 18/08/2024 Lecture-16: 19/08/2024 Lecture-17: 20/08/2024 Lecture-18: 20/08/2024 Lecture-19: 21/08/2024 Lecture-20: 21/08/2024 Lecture-21: 22/08/2024 Lecture-22: 22/08/2024 Lecture-23: 23/08/2024 Lecture-24: 23/08/2024 Lecture-25: 24/08/2024

[@shadowmind9346]

where now?

[@shadowmind9346]

completed the full playlist? working on a real project?

[@unityleveldesign4878]

That is most valuable things I ever come across, thanks for this great content.

[@learnenglish699]

hello i have some doubts

[@sabalniroula26]

Min-abs scaling is often used in situations where the signs of the original values are important and should be preserved, such as when working with financial data or when the scale of the original variables is not important and all that matters is the relative ranking of the values.

[@elyaabbas7216]

they way you teach every thing is just amazing i love it really. i used to learn from many platforms but you are the best of all in conveying the exact meaning in a beautiful way thanks a lot sir and stay bless.

[@vaibhavchaudhary5571]

I have seen others explaining Data science topics ..but you are way far from everyone.. ❤️

[@hamdansiddiqui3294]

Very informative, best Ml explanation, step by step on KZbin .

[@geekyprogrammer4831]

This is very underrated channel!

[@narendraparmar1631]

Added some useful knowledge today Thanks for this good work😀

[@goyanii]

free me bahot accha padhate ho sir app

[@Abraham33286]

I realy love this channel what a great explanation

[@G.VPrasannaAnjaneyulu]

hi sir... i am very much impressed... following thoroughly... but 1 thing is , to get much awareness how we can get your codes

[@poojadesai2826]

Very nice explanation in Feature Scaling. I have one doubt though, as it is mentioned, Feature Scaling is applied very last once everything is done like handling missing data, categorical data, detecting and removal of ourliers etc. In that case, when we always handle outliers first and then apply scaling, why do we need of RobustScaling for scenarios like outliers? We would not need to think of outliers while applying scaling.

[@manojrangera]

If outliers is our dataset are more then that outliers play an important role in ML algorithm.. May be some important information that y we didn't remove outliers and use robust scaler... Am I correct?.. Just clarify it..

[@osho_magic]

Outliers can’t always be omitted entirely

[@WAMIQMUSHTAQ-p9f]

Any detailed video on getting started with sk learn? Pls

[@Adarshhb767]

why didn't use or transform the y_train and y_test is that not necessary can someone explain?

[@Nudaykumar]

Sir, I can understand Hindi little bit, but still can grasp maximum based on your skills. I am having one question. I have introduced 3 outliers records one each for 'Class label' and applied MinMaxScalar. As you teached values are scaled between 0 and 1. But when i compare using kdeplot before and after scaling still i see those outliers between 0 and 1 spread. I am thinking those 3 outliers will be mingled with other values and that is the way we are going to eliminate outliers. plz correct me if i am wrong. Thanks in advance for this stuff.

[@navinebhatt4014]

We will never know the true min and max because we are performing the test split and applying fit. But what of the true max or min comes in test data. Then the range of the feature will be beyond the 0 to 1 range. What should we do then?

[@kadambalrajkachru8933]

Great teaching sir.. Thanks for such great content...

[@hmikraminfo7019]

sir thanks for this amazing play list. students I face some issue while plotting the sns plot at 8:50 then i try this line of code. it helps and resolve. i put here for some help. sns.scatterplot(data =df, x='alcohol',y= 'malic_acid', hue=df['class_label'])

[@zainfaisal3153]

Hello Sir! I want few minutes of yours. I am following this series and it's amazing. I just want to ask something that can you suggest me any book or any project source so that I can practice all concepts practically as well Thank you so much Please reply

[@saurabhdas2234]

This video was incredibly helpful

[@anupprasad695]

Sir, kya ho agar minimum ya maximum ya phir dono test data me ho...

[@freshersadda8176]

I'm Addicted to your channel ❤️

[@monikrayu2546]

ok

[@rockykumarverma980]

Thank you so much sir 🙏🙏🙏

[@hassamkafeel]

Hello! if we are splitting before applying MinMax Scaling, it is possible that maximum value of one feature say 250 end up in Test split. How would then MinMax scaling work considering we are only fitting it on Training dataset.

[@ayushbhanu-c2j]

when to use standardization and normalization , but are sqeezing data but when to use which one.

[@sonal008]

Why in last graph the scale is not from 0-1 .. it shows value of -0.2 to 1.2 ?

[@anshagarwal9826]

@campusX Just A Question should we scale the target variable also or it's only for the features that are inputs to the models

[@shahinanjum5287]

Only for features (independent columns)

[@sankettidke6060]

is scaling done only for continous features ?

[@saumyashah6622]

Hello sir, this is a suggestion, can you please make a video explaining the pipeline concept of the sklearn library. I have tried to learn from other videos from YT and official documentation, but I can't understand and implement pipelining in my code.

[@campusx-official]

Will do it in a few days

[@shreejanshrestha1931]

Yes sir its would be great. 😄 cause you explain the best

[@RaushanKumar-y1i9k]

The test data should be normalized using the same min and max values that were used to normalize the training data. This ensures consistency between training and test data. If you normalize the test data independently, it might be scaled differently than the training data, leading to inaccurate predictions. ... is this correct statement?? anyone plz confirm

[@surajghogare8931]

Teaching at its best... superb sir 🙌

[@KeigoEdits]

Suppose there is a data feature containing height, ranging from 10 to 50 now lets suppose we did split the data and according to the random seed we took the training set got the range of height from 11 to 48 but those data points having 10 and 50 heights went into test set, now the data is fit on 11 as min and 48 as max, now if we transform the test data these points will results in less than 0 and more than 1 values data points after transforming

[@Adventurebhat]

Thats why the concept of seeding comes , while train test split , so that the train and test splitting can be random

[@WowFactor2023]

Hii, where can I find the OneNote for this playlist

[@ezaanamin8479]

Hi I been watching your videos and also been making notes on the side and improving my math as well the problem is I didn't focus in my university since I was mostly busy in making web development projects can I laid a job as a data scientist without a master degree cause my GPA is low very low

[@esakkimuthu7650]

Sir, where we got these notes, which are u teaching

[@adarshvlog2409]

completed it!!

[@muhammadarhamadeel]

Sir can we do min max/standard scaling on y or target columns? If the target data is in continuous or regression form.

[@keshavkarki7775]

X_train, X_test, y_train, y_test = train_test_split(df.drop('Class label', axis=1), df['Class label'], test_size=0.3, random_state=0) I DIDN'T GET first two steps of lines of train test split because sir ne pehle ke videos me (x,y,test_size) ye format bataya tha split ke liye ye class label drop kaha se a gaye?

[@tanzeelmohammed9157]

df.drop('Classlabel',axis=1) is basically your X because you're dropping your feature variable, while df["ClassLabel] is your y

[@poojadesai2826]

I have one more question: why do we need to train_test_split first before applying scaling. What if Data is very huge and learning of mean, SD from training data would give wrong idea because test data set has some different observations which could hamper already learned mean and SD. I know this is very rare scenario but this could happen.

[@manojrangera]

That is my question also.. Some time we do train test split after that use scaling sone time we don't do and use in while dataset ... Y I that?... Can you explain me that

[@Fighte321]

Bro what U say about sparse data I don't understand it

[@el-drago9688]

Sparse data is that data, which contains a significant number of Zeroes in it

[@ZuhaibUlhassan-t7z]

Nice bro

[@sandipansarkar9211]

finished watching

[@zkhan2023]

Thanks sir

[@satishmutke8199]

Great 👍

[@Code-Pedia]

Love you sir from Pakistan

[@rubayetalam8759]

can you please update the dataset?

[@SaqibKhan-f1r9i]

anyone watching in the 2024 ❤

[@tanb13]

Could you please confirm if we normalisation/standardisation of target variable should also be done along with input variables? Kindly explain with an explanation or link to resources which explain this question.

[@osho_magic]

Sparse like in digit data

[@tusarmundhra5560]

awesome

[@Ganeshjadhav2808]

thank you sir

[@AzharKhan-wc1et]

Great Videos Thank you 👍

[@JustPython]

💗💗💗

[@Star-xk5jp]

day2-date:10/1/24

[@arshad1781]

Thanks

[@_iamankitt_]

thanks bro

[@abdulmanan17529]

🎉

[@harshsaxena1115]

sir why do we do fit our train data only to scaler object and know we need to transform train and test data but why only train data is to be fit?

[@salonikedia1891]

Could you please share the onenote link?

[@smitpatel1358]

Thank you sir!!

[@jahaansingh8627]

sorted

[@MRAgundli]

done

[@MuhammadJunaid-yr8jd]

I have seen others explaining Data science topics ..but you are way far from everyone..

[@ds.zubair]

Bessssssssssst

[@Mehedihasan-b4r9y]

I'm Addicted to your channel ❤

[@monikrayu2546]

ok

[@YashGaneriwal-je6rh]

done