Feature Scaling | Standardization Vs Normalization | Data Preprocessing | Python | Data Magic

[@Iranvlog_]

hi thank you. Totally you mean we can use standaridzation for any type of data distribution for example non normal distribution like uniform. and it is recommended to use normalization just for normally distributed data sets? Does the normalization or standasdization change the distribution of a data?

[@DataMagicAI]

-Yes, you can use standardization for any type of data distribution
-Normalization or Standardization never changes the distribution of your data.
Hope I answered your questions.

[@Iranvlog_]

@@DataMagicAI Thank you.

[@alndr4u]

Thank you so much.... How to convert standardized values between 0 to 100 ? Plz help me 🙏🏻

[@DataMagicAI]

I am not sure what's your purpose behind it. But you can use simple maths, multiply each scaled value bet ween 0 to 1 with 100.

[@alndr4u]

@@DataMagicAI z score values are between -3 to +3 range, then that multiply with 100, then values are remain -ve and +ve, but i want only 0 to 100 range

[@alndr4u]

@@DataMagicAI thank you for rply

[@MohdFadzilHassanIsmail]

Hi. Can I perform this feature scaling before spliting training and testing dataset? Any effects regarding data leakage?

[@DataMagicAI]

Perform feature scaling after dataset splitting to avoid data leakage and ensure realistic model evaluation.

[@fernandoborges3044]

Thank you, very helpful to me, I liked it a lot!

[@DataMagicAI]

Glad, you enjoyed it.👍

[@rmb1106]

Hi I want to use normalisation for 4 numbers, to make it between 0&1 , but Xmin should be 0 and Xmax should be the max number from the 4 numbers we are taking.
Can you tell me the code for that

[@DataMagicAI]

You can solve with simple math equation. ((Xmax - Xmin)/100)*(Xvalue-Xmin) = scaled Xvalue

[@rmb1106]

@@DataMagicAI I meant is Xmin =0( for all set of values) and our formular of normalisation (x - Xmin)/(Xmax-Xmin) becomes x/Xmax , so is it right to use this x/Xmax and call it normalisation

[@sudippandit1]

Awesome videos

[@DataMagicAI]

Glad you like them!

[@davideje8841]

Why do we still need to fit our test data, will it be improper to just transform both the train and test data after fitting only the train data?

[@DataMagicAI]

If you are training your model with scaled dataset and try to validate with test dataset which not scaled then you will get the wrong results. You always scale your test dataset along with train dataset so you will get the correct results for prediction/validation.

[@davideje8841]

@@DataMagicAI Alright, thanks

[@ali.mazeed]

thank you but where is method for Normalization to make the value between 0 to 1
where is the code and how make it

[@DataMagicAI]

Thanks you ! Will upload soon

[@zainularifin-ll4th]

Umm sir where's the next step where after scaling we use some regression model?

[@DataMagicAI]

You may checkout the ML Algorithms playlist on Data Magic UA-cam channel for the same.

[@AshokKumar-qh5vc]

sir you didn't told about normalisation

[@DataMagicAI]

I will add new short episode on it