Log Transformation for Outliers | Convert Skewed data to Normal Distribution

Поделиться
HTML-код
  • Опубликовано: 10 окт 2024

Комментарии • 33

  • @TheAIUniversity
    @TheAIUniversity  5 лет назад +1

    What are the other techniques
    I can use to treat outliers or convert negative or positive skewed data into normal distribution form?

    • @nabeelnaseer5592
      @nabeelnaseer5592 5 лет назад

      roots, exponents, inverse methods..

    • @nizamlootera3163
      @nizamlootera3163 4 года назад

      In Linear Regression suppose both the variables or features are positively skewed, then we should apply log10 to both of them

    • @XuanTran-ri1hn
      @XuanTran-ri1hn 2 года назад

      How about log 1 plus?

  • @nabeelnaseer5592
    @nabeelnaseer5592 5 лет назад +2

    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..

  • @edphi
    @edphi 2 года назад +1

    Excellent

  • @ajaykushwaha-je6mw
    @ajaykushwaha-je6mw 3 года назад

    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 ?

  • @aakashv4594
    @aakashv4594 4 года назад +1

    What are the functions to be applied for negative skews and also if the data has zero

  • @creativesurgeinfidel
    @creativesurgeinfidel 3 года назад

    Thank you.. Could you please let me know how to convert natural log back to the original value

  • @elyasmohammadi8409
    @elyasmohammadi8409 4 года назад +1

    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

    • @edphi
      @edphi 2 года назад

      Frequency distribution graph

  • @vinayvvalaboju
    @vinayvvalaboju 2 года назад

    Can you fix a custom bin And filter data til upper quartile.

  • @mansirawat1052
    @mansirawat1052 Год назад

    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?

  • @Karthik_info_vlogs
    @Karthik_info_vlogs 4 года назад +1

    Good info

  • @pallavijagtap8140
    @pallavijagtap8140 3 года назад

    Sir, Once you transform the variables, do we have to use same transformed columns in further process of melling?

  • @pallavijagtap8140
    @pallavijagtap8140 3 года назад

    Pallavi Jagtap
    1 second ago
    Sir, Once you transform the variables, do we have to use same transformed columns in further process of modelling?

  • @aniketsultan9497
    @aniketsultan9497 5 лет назад +3

    other methods square root, cube root , binning

  • @durgadeviarulrajan4560
    @durgadeviarulrajan4560 2 года назад

    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?

    • @usmanriaz6241
      @usmanriaz6241 Год назад

      It confuses me too. tell me if you know now

  • @balamurali75
    @balamurali75 2 года назад

    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

  • @shrutimadan4451
    @shrutimadan4451 3 года назад

    using log10 transformation, it didnt give normal distribution.
    How to deal with this?

  • @MrNabiwishes
    @MrNabiwishes 4 года назад

    Log transformation applied to train set, and when out of sample data comes in do we apply same transformation...

  • @amitbudhiraja7498
    @amitbudhiraja7498 2 года назад

    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 ?

  • @nicholaslipanovich827
    @nicholaslipanovich827 3 года назад

    The information you communicated to us was fine but your delivery could use some work. Trying to repeat yourself less might help.

  • @independent7212
    @independent7212 4 года назад

    negatively skewed data to normal distribution?

  • @ankurkamthan5854
    @ankurkamthan5854 4 года назад

    Why should not taken log with base e and y base 10

  • @nikhilgaikwad9954
    @nikhilgaikwad9954 4 года назад

    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??

    • @prathameshmistry3868
      @prathameshmistry3868 4 года назад

      no,the values are inserted and then transformed in the code

    • @vineethp8925
      @vineethp8925 3 года назад

      @Prathamesh Mistry can u please explain more clearly because iam also having the same doubt

  • @vuminhquanle1426
    @vuminhquanle1426 4 года назад +1

    I listened very carefully, cause I can't understand anything at 1.5x Speed

  • @rohitjaiswal6102
    @rohitjaiswal6102 4 года назад +1

    Can u share your github link about this codes....

    • @TheAIUniversity
      @TheAIUniversity  4 года назад +1

      Here you go... github.com/nitinkaushik01/Machine_Learning_Data_Preprocessing_Python/find/master?q=

  • @fakhrik
    @fakhrik 4 года назад

    Why do Indians have to use the word OK so much?