Live Discussion On Outlier And Its Impacts On Machine Learning UseCases

Поділитися
Вставка
  • Опубліковано 19 гру 2024

КОМЕНТАРІ •

  • @mahesh1234m
    @mahesh1234m 4 роки тому +11

    In Naive Bayes , Outliers will affect the shape of Gaussian distribution. That is the Outliers would have impact on the mean and standard deviation of the bell curve., So Naive Bayes is sensitive to Outliers.

  • @beingrishi6450
    @beingrishi6450 2 роки тому +1

    Love u Krrish Sir always been ur fan.
    Data science expert

  • @ArpitYadav-ws5xe
    @ArpitYadav-ws5xe 4 роки тому +4

    Excellent krish. God bless you. Keep working like this

  • @chinmaybhat9636
    @chinmaybhat9636 4 роки тому +2

    Yes @KrishNaik Sir KNN is sensitive to outliers and Imbalanced dataset which was said by you in the video of KNN In Depth Intuition Video
    Thanks & Regards,
    CHINMAY N BHAT

    • @KnowledgeAmplifier1
      @KnowledgeAmplifier1 4 роки тому

      KNN may be affected by outliers , but weighted knn is not for sure as weight proportional to 1/distance & for outlier , it has abnormal high distance from rest of the dataset , so weight is going to be very less & as a result , outliers will be less impactive :-)

    • @sajidchoudhary1165
      @sajidchoudhary1165 4 роки тому

      @@KnowledgeAmplifier1 weights don't used for handle outliers , weights are use for giving importance to nearest points

    • @sajidchoudhary1165
      @sajidchoudhary1165 4 роки тому

      yes , he said

    • @KnowledgeAmplifier1
      @KnowledgeAmplifier1 4 роки тому

      @@sajidchoudhary1165 Yes , I know well why we use weighted knn , I was telling , if we go with weighted knn , then the outlier can not affect that much on prediction as you said also weights are use for giving importance to nearest points , so outliers will get less importance

    • @sajidchoudhary1165
      @sajidchoudhary1165 4 роки тому

      @@KnowledgeAmplifier1 what if outliers are much nearer toh future data point

  • @akarkabkarim
    @akarkabkarim 3 роки тому +1

    thank you mister krish

  • @ayush.pratap.singh.
    @ayush.pratap.singh. 4 роки тому +1

    Thanks Krish! Really nice video!

  • @mounicak5923
    @mounicak5923 4 роки тому +1

    Excellent Krish sir

  • @KnowledgeAmplifier1
    @KnowledgeAmplifier1 4 роки тому +1

    KNN may be affected by outliers , but weighted knn is not for sure as weight proportional to 1/distance & for outlier , it has abnormal high distance from rest of the dataset , so weight is going to be very less & as a result , outliers will be less impactive :-)

  • @equbalmustafa
    @equbalmustafa 3 роки тому +1

    KNN
    if k=low, the model is susceptible to outliers
    if k=high, robust to outliers

  • @eduhomebyshubh5445
    @eduhomebyshubh5445 4 роки тому +1

    Excellent sir

  • @KnowledgeAmplifier1
    @KnowledgeAmplifier1 4 роки тому +3

    Sir , I have one doubt regarding the Robustness of Regression Tree towards Outliers.
    Regression tree split criterium depends on the averages of the two groups that are splitted, and, as the average is severely affected by outliers, then the regression tree will suffer from outliers.
    Whereas Classification Tree uses Mode which in not affected by outliers ...
    So when we say "Decision trees are robust to outliers", we mean Decision Tree Classifier , not Regressor.
    So I think Regression Tree are affected by outliers :-/

  • @victorhenostroza1871
    @victorhenostroza1871 3 роки тому +1

    Hello, Why no tuse a multivariate outliers detector instead of univariate analysis?

  • @kousikdutta5889
    @kousikdutta5889 2 роки тому +1

    Sir, DBSCAN is not sensitive to outlier, it is actually robust to outlier.

  • @venkateshmunagala8089
    @venkateshmunagala8089 2 роки тому

    SVMs are sensitive to outliers the support vectors are calculated based on distance only

  • @joeljoseph26
    @joeljoseph26 3 роки тому +1

    why can't we apply transformation techniques to remove skewness and verifying using Q-Q plot?

  • @bmodi700
    @bmodi700 4 роки тому +3

    Is it advisable to use Log scaling on target variable to over come outlier?

  • @rakeshkumarsaini9474
    @rakeshkumarsaini9474 3 роки тому +1

    can u please help I need to know if LPNT is a reliable cryptocurrency as the Probit exchange shows its status as Outlier? please give it a glance as I have invested a lot of money in it.

  • @sajidchoudhary1165
    @sajidchoudhary1165 4 роки тому +1

    i also think so KNN Classifier is sensitive to outliers and you also told in one video KNN Classifier is Sensitive and DBSCAN is'nt sensitive to outliers because it handle outliers by itself it don't consider outliers in Clusters

  • @sandipansarkar9211
    @sandipansarkar9211 3 роки тому

    finished watching

  • @shreyasb.s3819
    @shreyasb.s3819 4 роки тому +4

    krish unfortunately u did one mistake here. Actually KNN is very sensitive to outliers. Guys u can google it.

  • @midhunmurali3620
    @midhunmurali3620 3 роки тому +1

    sir if we have 80 columns and 5 rows and how to find outliers for each every columns and how to remove it sir

  • @salehabdullahi9356
    @salehabdullahi9356 Місяць тому

    What of applying log1p to the skewed distribution?

  • @parthagarwal4592
    @parthagarwal4592 3 роки тому +2

    What's wrong sir at 1:31:54 ??

    • @sameer9045
      @sameer9045 Рік тому

      😂🤣😁 he forgot he is teaching.

  • @karanbhuva7733
    @karanbhuva7733 3 роки тому +1

    could you please explain why you replaced outliers with upper bound value? can we replace it with mean?

    • @KnowledgeAmplifier1
      @KnowledgeAmplifier1 2 роки тому +1

      mean is highly affected by outliers , like for example , mean of 1,2,3,4,6,10000 is 1669.33 something , that means one single outlier is affecting the mean such that , the mean is not representing the majority part of the dataset properly , so when outlier is present , you should never use mean , rather median is not that much affected by outliers , you can try that or some other advanced algorithm , hope this will be helpful ! Happy Learning :-)

  • @abhinaygupta8243
    @abhinaygupta8243 4 роки тому +1

    if distribution is left skewed then we have to replace all left hand values less than -3S.D. with -3S.D. to make it normally distributed?

  • @mrigankshekhar384
    @mrigankshekhar384 4 роки тому +1

    What if in multinomial naive Bayes , you have given a test data containing a word which is not in our document matrix as a feature

  • @basavarajag1901
    @basavarajag1901 2 роки тому

    In I am doing EDA , as per discussion EDA will be done on raw data, in raw data many outliers and we may or may not made decision on outliers, before doing eda , we need treat outliers ..please clarify ...Thanks

  • @vijayayaduvanshi1036
    @vijayayaduvanshi1036 2 роки тому

    Please discuss MATLAB Coding as you have discussed python coding

  • @bismaishfaq1164
    @bismaishfaq1164 4 роки тому +2

    i am working on covid 19 prediction.. should it is needed to remove outliers or not.

    • @sridhar6358
      @sridhar6358 4 роки тому

      if the outliers are genuine you should not remove them it depends on the business use case too

  • @asawanted
    @asawanted 3 роки тому +1

    Sir I have one doubt. Is there a difference between min value and lower bridge? For fare my min value shows zero but lower bridge shows negative 61.

  • @abhisekdatta8887
    @abhisekdatta8887 4 роки тому +2

    with due respect sir
    one of the videos of KNN you told that KNN is sensitive to outliers but you have written that KNN is not sensitive to outliers, it makes me more confusing.
    sir kindly please clarify
    PS huge fan of yours

    • @c.b.t6738
      @c.b.t6738 4 роки тому +3

      knn actually is really sensitive to outliers you can easily see that as it is fully based on norms

    • @KnowledgeAmplifier1
      @KnowledgeAmplifier1 4 роки тому

      KNN may be affected by outliers , but weighted knn is not for sure as weight proportional to 1/distance & for outlier , it has abnormal high distance from rest of the dataset , so weight is going to be very less & as a result , outliers will be less impactive :-)

  • @sandeepmutkule4644
    @sandeepmutkule4644 4 роки тому +1

    sir can we replace outliers with mean value of that variable? or it will cause any error or something , if we have large data and lots of outliers.

    • @KnowledgeAmplifier1
      @KnowledgeAmplifier1 2 роки тому

      mean is highly affected by outliers , like for example , mean of 1,2,3,4,6,10000 is 1669.33 something , that means one single outlier is affecting the mean such that , the mean is not representing the majority part of the dataset properly , so when outlier is present , you should never use mean , rather median is not that much affected by outliers , you can try that or some other advanced algorithm , hope this will be helpful ! Happy Learning :-)

  • @karthiksundaram544
    @karthiksundaram544 2 роки тому

    Sir can u explain travelling salesman

  • @chhavigaur4608
    @chhavigaur4608 3 роки тому +1

    Hey @Krish Naik do we have to make end-to-end projects just like you or make custom functions for steps. Which looks more professional Please reply.

  • @vigneshnagaraj7137
    @vigneshnagaraj7137 3 роки тому +1

    How about outlier analysis in text data ?. Is there any specific method Krish

  • @purvijardos3249
    @purvijardos3249 4 роки тому +1

    hello sir, i have one query. may be not relevent to this.. but if my data is not normally distributed and if wants to apply gussian naive bayes what method can be apply to convert data into normally distributed

  • @ty_b_63_prajwalwaykos86
    @ty_b_63_prajwalwaykos86 2 роки тому

    Hello , can some one please answer...weather i can consider a categorical variable in which one class has very low frequency as compared to other classes as an outlier????

  • @anjaligautam7528
    @anjaligautam7528 4 роки тому +1

    Is it possible to apply RNNs for the Outlier Detection?

  • @i_amanrajput
    @i_amanrajput 4 роки тому +1

    print('Thanks a lot, Krish !)

  • @sandipansarkar9211
    @sandipansarkar9211 3 роки тому

    finished practicing code

  • @basavarajag1901
    @basavarajag1901 2 роки тому

    If the dataset contains more 30% of outliers , how to impute them ...( if we cannot drop also and not able to replace with mean/mode/ median)

  • @sushovanmallick6833
    @sushovanmallick6833 4 роки тому +1

    Sir i want to transit my current sofware domain to data science domain. Enrolling in aplied ai course for total data science is it ok or this course will only teach the ai and ml?

    • @srinivasaraob4168
      @srinivasaraob4168 4 роки тому

      Hi sushovan, you have asked the wrong window. do to applied ai channel and ask. you may get answer

  • @snehasingh9399
    @snehasingh9399 Рік тому

    Where do we get materials

  • @padmapriya805
    @padmapriya805 2 роки тому

    Can 3 quartiles(20%, 50% and 75%) be the same?

  • @yashvichauhan4137
    @yashvichauhan4137 4 роки тому +1

    Hey, make an unboxing video of titan rtx please!!

  • @chinmaythorat5207
    @chinmaythorat5207 4 роки тому +1

    Hi sir

  • @sayeerahad9215
    @sayeerahad9215 2 роки тому

    KNN is sensitive sir

  • @joeljoseph26
    @joeljoseph26 3 роки тому

    SVM is sensitive to outliers.

  • @sumitdas7489
    @sumitdas7489 3 роки тому

    Not Sure about this video everywhere i serached it says naive bayes is sensitive to outliers even SVM and linear regression already so much confusion in Data science topic and this is more confusing and misguiding

  • @revathyv5038
    @revathyv5038 4 роки тому +1

    Hii