Evaluating Classifiers: Confusion Matrix for Multiple Classes

Поделиться
HTML-код
  • Опубликовано: 17 ноя 2024

Комментарии • 101

  • @YogaByMarcelGroux
    @YogaByMarcelGroux 8 лет назад +11

    Hey I just wanted to tell you that our prof referenced to this video, in order for us to do a specific assignment. Cool video :)

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  8 лет назад +1

      +c4tTi many thanks for letting me know .. which university are you studying at?

  • @user-cc8kb
    @user-cc8kb 6 лет назад +3

    Thank you very much. Your explanation was much needed for my thesis :)

  • @user-or6mz4gy6i
    @user-or6mz4gy6i 4 года назад +2

    I beg to differ on the TN calculation, but maybe am I wrong?
    For me, True negatives for a class are the numbers of the matrix on the diagonal which are not TP (example: for class A, it will be BB, CC, DD, EE, but not AA which are TP). At 9:21, your matrix shows the white cells are not all filled with blue values, which are well classified values. Hence it should not be all cells but only diagonals except the true positive for this class, no?

  • @ETcodenet
    @ETcodenet 2 года назад

    It is So cool tutorial.. thanks so much. very clear and very interesting lecture..

  • @FatimaH-iu8mn
    @FatimaH-iu8mn 8 лет назад +6

    Thank you for this very clear tutorial!
    I wonder if there is any specific method to calculate the total F-measure for multi-class classification?(do we just calculate the average of the F-measure values for all the classes, or there is another better way?)

  • @liveartrainbow
    @liveartrainbow 10 лет назад +5

    Thank you for spreading the knowledge.

  • @masoudrezaei331
    @masoudrezaei331 3 года назад

    This video helped me a lot. Wish you all the best sir.

  • @karwandbg1
    @karwandbg1 9 лет назад +2

    Thanks a lot it is really valuable tutorial. Wish you all the best

  • @HeaderGFX
    @HeaderGFX 5 лет назад +3

    Thanks for the video the explanation how you can get the fp tn and fn from the confusion matrix helped me a lot for my university exercise

  • @MaheshBabu-ys3vh
    @MaheshBabu-ys3vh 8 лет назад +7

    @10.3, Precision A = 25/(25+3+1)

  • @alexgiank1960
    @alexgiank1960 Год назад

    great video , easy to understand . Keep it up

  • @emiliogenaro90
    @emiliogenaro90 8 лет назад +2

    Nice lecture! It's very clear the explanation :)
    I wonder if you have some video or explanation about precison and recall clustering, that is to say, something like an average or similar among all the class precisions and recalls.

  • @rinkssinghal488
    @rinkssinghal488 3 года назад

    A true value add to my knowledge

  • @hildamayrose
    @hildamayrose 4 года назад

    This video was very helpful for me. Thank you.

  • @bulusbali1798
    @bulusbali1798 2 года назад

    Thanks a lot. This is confusion matrix made easy.

  • @sateeshkumar1145
    @sateeshkumar1145 5 лет назад +5

    accuracy = (TP +TN)/(TP + TN + FN + FP) ..... but you said sum all TP and divide it by sum of everything else ... answers are different.... i think we will have to calculate TN for every coloumn and sum it same for FP and FN ... what do you think ??

    • @hianjana
      @hianjana 5 лет назад +2

      I too agree. In the video, numerator is only TP. Ideally, it should be TP + TN. Denominator = entire table which is correct I think.

    • @desikpoplover
      @desikpoplover 5 лет назад

      Were you able to get a reasonable explanation to this?

    • @sateeshkumar1145
      @sateeshkumar1145 5 лет назад

      @@desikpoplover not yet

  • @monikabansal5808
    @monikabansal5808 4 года назад

    Nice Explanation, thanks for sharing a great tutorial

  • @shaelanderchauhan1963
    @shaelanderchauhan1963 3 года назад

    Perfectly explained!

  • @taruninbox
    @taruninbox 7 лет назад

    Very nice explanation of multi-class confusion matrix...

  • @hrkhaff
    @hrkhaff 9 лет назад +1

    Hi, I calculated Accuracy, Sensitivity,Specificity and F-1 score using your explanation for multiclass, but the specificity values are always way higher than the rest of them (10+% higher sometimes 40% higher!). Is that normal?

  • @FOR4MUSIC
    @FOR4MUSIC 7 лет назад +2

    where was this video when i was doing my project...extremely helpfull!

  • @jhonmoncaleanotorres1975
    @jhonmoncaleanotorres1975 4 года назад

    This is a very clear tutorial, thank you!!

  • @VenusOnlineSoftware
    @VenusOnlineSoftware 3 года назад

    you saved my time thanks alot

  • @lobbielobbie1766
    @lobbielobbie1766 7 лет назад +4

    Hi Dr Sadawi, thanks for your great explanation. Are there ways to calculate the overall precision, recall and specificity? thanks.

    • @amernew3ful
      @amernew3ful 3 года назад

      Just get the average

    • @SHADABALAM2002
      @SHADABALAM2002 3 года назад

      @@amernew3ful .. r u confirm about it. Because i am also stuck in taking overall scores of whole model. Plz confirm regards.

  • @nag9502
    @nag9502 3 года назад +1

    very good and useful

  • @khuzaimamughal2226
    @khuzaimamughal2226 3 года назад

    Thankyou sir for this explanation

  • @juannieves8078
    @juannieves8078 3 года назад

    Such a good explanation!!!

  • @felipefscalco
    @felipefscalco 5 лет назад

    Thank you very much for this explanation!

  • @kevinsetiawan7638
    @kevinsetiawan7638 4 года назад

    This video helped me alot, thank you

  • @madhumathikannan2930
    @madhumathikannan2930 9 лет назад +7

    please improve your Audio Quality,there are too many air blows.Great lecture.

  • @wrightvcx2249
    @wrightvcx2249 4 года назад +1

    Would have been better if you had explained this with an example in R.

  • @muaidahmed4019
    @muaidahmed4019 9 лет назад +2

    Great and well explained lecture. can I construct ROC from multiclass confusion matrix? or does the ROC is for one class? tq

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  9 лет назад +2

      +Muaid Ahmed see here: ruclips.net/video/fKo-HSBx4G0/видео.html

    • @goldnutter412
      @goldnutter412 3 года назад

      NO idea what you are talking about.. thanks Google
      Ahh false positive testbed, how I miss those long scans between iterations on a very limited but.. promising custom detection signature method for PE (.exe/dll) files. Compilers always have markers, opcodes like SHR SHL only get used for specific reasons, code section entropy etc etc

  • @aouatefmahani232
    @aouatefmahani232 5 лет назад

    Thank you very much for this explanation. Please, how to calculate the confusion matrix? Do you have some examples? Thank you for your help.

  • @dewanta1197
    @dewanta1197 3 года назад

    i want to ask, how to calculate the accuracy for each class on multiclass classification. It is possible ?

  • @mustafasalah9491
    @mustafasalah9491 6 лет назад

    thank you so much... the presentation is so great.
    please, could you share the presentation slides?

  • @mhetzenberger
    @mhetzenberger 8 лет назад

    Is there any official paper about this topic, i.e. how to get the 2x2 confusion matrix of a single class?

  • @a.v.a.bharatkumar5417
    @a.v.a.bharatkumar5417 6 лет назад

    how do we calculate F-measure, chi-square, p-value from the example shown here in this video and explanation about their significance in the analysis would be helpful......

  • @brandonkieft8310
    @brandonkieft8310 9 лет назад +1

    great tutorial!

  • @SHADABALAM2002
    @SHADABALAM2002 3 года назад

    Hi. How can we calculate accuracy, percision, recall, specificity and F1 of whole model(not class wise) in multi class classification. Thanks.

    • @nedim8403
      @nedim8403 3 года назад

      Mr.Shabad did you get any aswer to your question.It is also my question?

    • @SHADABALAM2002
      @SHADABALAM2002 3 года назад

      @@nedim8403 . I took thek average of all individual classes.. is it right way??

    • @nedim8403
      @nedim8403 3 года назад

      ​@@SHADABALAM2002 it seems that the calculated values in some publications are different from the avarage. There must be a way other than averaging on non-balance data

    • @nedim8403
      @nedim8403 3 года назад

      The opinion of Noureddin Sadawi is important in this regard.

  • @saisanjaynagarur
    @saisanjaynagarur 5 лет назад

    how to check foolproof metric in a confusion matrix??

  • @anitasahoo4372
    @anitasahoo4372 6 лет назад

    Awesome Explanation Thanks:)

  • @jerryli2436
    @jerryli2436 6 лет назад

    very clear, nice tutorial

  • @malothnaveen3727
    @malothnaveen3727 2 года назад

    Sir, how to find accuracy of each class?

  • @mach9libra
    @mach9libra 3 года назад

    How to calculate, TP, TN, FP, FN overall and not class wise

  • @seemoshaik
    @seemoshaik 5 лет назад

    Really Helped Thank You

  • @fasteinz
    @fasteinz 7 лет назад

    Thank you, great class.

  • @benjamenaremu4110
    @benjamenaremu4110 4 года назад

    Nice tutorial.

  • @lucijeanej
    @lucijeanej 6 лет назад

    Thanks a lot man, great tutorial!

  • @fernandoandrade593
    @fernandoandrade593 9 лет назад

    great tutorial!
    very thanks!

  • @ajinugroho8698
    @ajinugroho8698 7 лет назад

    hontoni arigatou gozaimasu, its make me clear

  • @chetanar4272
    @chetanar4272 2 года назад

    good explaination

  • @pallabchowdhury2544
    @pallabchowdhury2544 4 года назад

    Thanks a lot sir

  • @im_karamo1907
    @im_karamo1907 6 лет назад

    the formula
    for precision, 10:24, 25 twice in the denominator?? can you check it again

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  6 лет назад

      You're right .. it's a typo .. many thanks for pointing it out!

    • @im_karamo1907
      @im_karamo1907 6 лет назад

      Noureddin Sadawi its ok.. Congratulations once again!!

  • @abdulwarissherzad9914
    @abdulwarissherzad9914 3 года назад

    Thank You i got it.

  • @ryunstyle
    @ryunstyle 8 лет назад

    Thank you. Very useful.

  • @apostolosmavropoulos177
    @apostolosmavropoulos177 3 года назад

    perfect video

  • @Appleiphones1
    @Appleiphones1 7 лет назад

    Excellent explanation.Any PPT/Pdf of this tutorial?

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  7 лет назад

      No, but this should be useful: www.saedsayad.com/model_evaluation_c.htm

    • @Appleiphones1
      @Appleiphones1 7 лет назад

      thank you.is there online Confusion matrix calculator for 3 classes?Its tedious to calculate manually every time

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  7 лет назад

      I don't know TBH

  • @im_karamo1907
    @im_karamo1907 6 лет назад

    presicision is the same as the positive predictive value

  • @ujjalkrdutta7854
    @ujjalkrdutta7854 8 лет назад

    nicely explained

  • @ahmedjamel421
    @ahmedjamel421 7 лет назад

    Dear Dr. Sadawi, Thank you for the explanation but when I want to open the link about mukalla notes I cant access it could you please send me that pdf ? جزاك الله خيرا

  • @midos2200
    @midos2200 4 года назад

    pure gold

  • @yellowbeard1
    @yellowbeard1 5 лет назад

    Can you please rerecord this series of videos? The information is great but the mic breathing and other audio issues is painful. I saw some of your more recent videos and you solved this issue with them but it would be nice to spend some time going back over older material.

  • @KonstantinPoddubnyy
    @KonstantinPoddubnyy 8 лет назад

    Nice and clear

  • @gauravbaviskar4778
    @gauravbaviskar4778 5 лет назад

    Great....✓!!!

  • @afsangujarati9427
    @afsangujarati9427 7 лет назад

    Greate video.
    I need help understanding how did they calculate per class accuracy, precision, recall and f1 score in given 10*10 confusion matrix:
    scikit-learn.org/stable/auto_examples/classification/plot_digits_classification.html
    I was trying to calculate the values for class 1:
    Accuracy I understand should be 88/(88+3+1) = 0.956.. [Value from link -> 0.99]
    Precision : 88/(88+1) = 0.9887... [Value from link -> 0.97]
    Recall: 88/(88+3) = 0.9670...
    There is a value mis-match, you help would be really appreciated.

  • @aquinovelascoosorio4078
    @aquinovelascoosorio4078 5 лет назад

    excellent video, could you pass the references?

    • @DrNoureddinSadawi
      @DrNoureddinSadawi  5 лет назад

      Unfortunately I don't have any references .. but my slides are here:
      github.com/nsadawi/DataMiningSlides

  • @MsHappyburger
    @MsHappyburger 5 лет назад

    thank you

  • @7justfun
    @7justfun 3 года назад

    Simply put, 9:55 everything non B is true negative for class b

  • @Priyadharshini-uf7qy
    @Priyadharshini-uf7qy 6 лет назад

    confusion matrix for 7 class please

  • @aymanhamedmoustafa
    @aymanhamedmoustafa 10 лет назад

    very thanks

  • @goldnutter412
    @goldnutter412 3 года назад

    Fascinating.. data robot and Google sent me. Thanks for/from ignorant humans ;)
    fyi 404 on your personal site from desc.

  • @paulwang281
    @paulwang281 8 лет назад

    cool

  • @mehdikerbedj6202
    @mehdikerbedj6202 6 лет назад

    Hi Mr nourredin I'm student in university of Algeria and I work to build project of classification I hope to help me if u can put your email please

  • @darshakramani7196
    @darshakramani7196 2 года назад

    learn english first

  • @nedim8403
    @nedim8403 3 года назад

    Hi. How can we calculate accuracy, percision, recall, specificity and F1 of whole model(not class wise) in multi class classification. Thanks.

    • @zarrintasnim3604
      @zarrintasnim3604 2 года назад +1

      if you found he solution please let me know. i am looking for the same ans.