What can we do if even after the transformation, there are outliers..am kinda puzzled over this notion of natural outliers. Like we are supposed to treat them separately.. can you give some pointers..
Suppose in one of my outcome measures pre is normal but post is not normal, so should I log transform only the post recording or should I transform both the pre and post values for further analysis?
Sir what is the correct sequence of variable transformation. First we need to do feature scaling then Gaussian transformation or First Gaussian transformation then feature scaling ?
Sir small dout I have two variables(independent and Dependent) represented in percentage. If I apply log for only one variable. Will result differs. Is it the correct way of transformation/analysis
Hi, Thanks for the great video. Is it necessary to convert all features into normally distributed, before modeling? Is it a compulsory step to follow in feature engineering?
I have a doubt like what is the optimal method to do remove the outliers [Z-score , IQR method] or use transformation methods like log normal or inverse Can someone tell ?
Hello and thank you for this nice video. Could you please clarify that what are the axis X and Y before and after log transformation. Thank you in advance
after we transformed the column values using log10. if we build a app using flask what values we should pass for that column to predict the output?? the original value or first we need to transform that value using log 10 and then insert??
What are the other techniques
I can use to treat outliers or convert negative or positive skewed data into normal distribution form?
roots, exponents, inverse methods..
In Linear Regression suppose both the variables or features are positively skewed, then we should apply log10 to both of them
How about log 1 plus?
What can we do if even after the transformation, there are outliers..am kinda puzzled over this notion of natural outliers. Like we are supposed to treat them separately.. can you give some pointers..
Suppose in one of my outcome measures pre is normal but post is not normal, so should I log transform only the post recording or should I transform both the pre and post values for further analysis?
Sir what is the correct sequence of variable transformation.
First we need to do feature scaling then Gaussian transformation or First Gaussian transformation then feature scaling ?
Sir small dout I have two variables(independent and Dependent) represented in percentage. If I apply log for only one variable. Will result differs. Is it the correct way of transformation/analysis
Pallavi Jagtap
1 second ago
Sir, Once you transform the variables, do we have to use same transformed columns in further process of modelling?
Thank you.. Could you please let me know how to convert natural log back to the original value
Can you fix a custom bin And filter data til upper quartile.
What are the functions to be applied for negative skews and also if the data has zero
Sir, Once you transform the variables, do we have to use same transformed columns in further process of melling?
Hi, Thanks for the great video. Is it necessary to convert all features into normally distributed, before modeling? Is it a compulsory step to follow in feature engineering?
It confuses me too. tell me if you know now
I have a doubt like what is the optimal method to do remove the outliers [Z-score , IQR method] or use transformation methods like log normal or inverse
Can someone tell ?
Hello and thank you for this nice video.
Could you please clarify that what are the axis X and Y before and after log transformation. Thank you in advance
Frequency distribution graph
Excellent
Log transformation applied to train set, and when out of sample data comes in do we apply same transformation...
using log10 transformation, it didnt give normal distribution.
How to deal with this?
after we transformed the column values using log10. if we build a app using flask what values we should pass for that column to predict the output?? the original value or first we need to transform that value using log 10 and then insert??
no,the values are inserted and then transformed in the code
@Prathamesh Mistry can u please explain more clearly because iam also having the same doubt
negatively skewed data to normal distribution?
Good info
other methods square root, cube root , binning
Why should not taken log with base e and y base 10
The information you communicated to us was fine but your delivery could use some work. Trying to repeat yourself less might help.
Can u share your github link about this codes....
Here you go... github.com/nitinkaushik01/Machine_Learning_Data_Preprocessing_Python/find/master?q=
I listened very carefully, cause I can't understand anything at 1.5x Speed
Why do Indians have to use the word OK so much?