ROC Curve & Area Under Curve (AUC) with R - Application Example

Поделиться
HTML-код
  • Опубликовано: 1 дек 2024

Комментарии • 499

  • @kaduflutist
    @kaduflutist 3 года назад +3

    Excellent, Sir! Thanks a lot for bringing it so straightforward and consistently. For the first time, I could understand and reproduce the whole thing in r, regarding ROC Curve and AUC.

    • @bkrai
      @bkrai  3 года назад

      You are most welcome!

  • @aaradhyaabhijitshetti3933
    @aaradhyaabhijitshetti3933 4 года назад +4

    Sir... your explanation of ROC and AUC was very simple and easy to understand .Its cleared my all doubts..Thanks a lot...

    • @bkrai
      @bkrai  4 года назад

      You are most welcome!

  • @TH-fe1vs
    @TH-fe1vs 7 лет назад +4

    thank you Sir, very kind of you to send the R code and the data, appreciate it. Your youtube here explains those concepts of ROC and AUC clearly, with a simple example. Well done.

    • @bkrai
      @bkrai  7 лет назад

      Thanks!

  • @johnwilliammeyer6592
    @johnwilliammeyer6592 6 лет назад +5

    Unbelievably helpful video - I've been searching all over internet for this. Thank you.

    • @bkrai
      @bkrai  6 лет назад

      That's good to know!

  • @kalyanasundaramsp8267
    @kalyanasundaramsp8267 6 лет назад +3

    superb sir...phenomenal.....u make tough things look simple....proud of you boss

    • @bkrai
      @bkrai  6 лет назад

      Thanks!

  • @duleepaj
    @duleepaj 6 лет назад +4

    Short, simple and covers everything! Thank you!

    • @bkrai
      @bkrai  6 лет назад

      Thanks for comments!

  • @FOR4MUSIC
    @FOR4MUSIC 7 лет назад +3

    it is very clear you know what you are doing.Thank you for your contribution !

    • @bkrai
      @bkrai  7 лет назад

      Thanks for your feedback!

  • @OrcaChess
    @OrcaChess 6 лет назад +1

    A ROC Curve Tutorial for more than two classes with the 1 vs ALL approach
    would be a very helpful video :).

    • @bkrai
      @bkrai  6 лет назад

      Thanks foe the suggestion, its on my list now.

  • @getugamo7051
    @getugamo7051 3 года назад +3

    Very clear and impressive lecture! Thanks so much!

    • @bkrai
      @bkrai  3 года назад

      You're very welcome!

  • @KR-good
    @KR-good 7 лет назад +4

    This was an amazingly clear approach. Thank you.

    • @bkrai
      @bkrai  7 лет назад

      Great to hear your feedback!

  • @Revboiuk09
    @Revboiuk09 5 лет назад +1

    Thanks a lot sir... for such precise explanation of AU ROC curve. Truly appreciated.!

    • @bkrai
      @bkrai  5 лет назад

      Thanks for comments!

  • @danielsaphir8596
    @danielsaphir8596 4 года назад +5

    Yo that beat in the beginning was fire

    • @bkrai
      @bkrai  4 года назад

      Thanks :)

  • @vijaymore1239
    @vijaymore1239 7 лет назад +2

    Thank you so much for explaining in much much simpler way!!!!!

    • @bkrai
      @bkrai  7 лет назад

      Thanks for your comments!

  • @raghavendras5331
    @raghavendras5331 6 лет назад +1

    Thank you sir...very clear and crisp explanation. In one video I got all the information. From the explanation in the video, I got how to find cutoff for maximum accuracy, by doing this only one class has got more weight in my dataset. but how to find a threshold value of cutoff(which gives maximum of sensitivity and maximum of specificity).

    • @bkrai
      @bkrai  6 лет назад +1

      You can get that using the ROC curve. The color used on the curve changes from 0 to 1. You can identify a point on the curve that is closest to the ideal curve.

    • @raghavendras5331
      @raghavendras5331 6 лет назад +1

      Thank you sir

  • @RoomeyRahman
    @RoomeyRahman 7 лет назад +4

    Dear Sir,
    Your tutorial helps us all to learn about data science. I learn many thing from your tutorial. Now I want to learn how we can make ROC and AUC for multi-class? May you make another video to teach us about multi-class model performance?
    Thank you

    • @bkrai
      @bkrai  7 лет назад

      Thanks for your comments, I'll add your suggestion to my list.

    • @santanumallik6992
      @santanumallik6992 4 года назад

      @@bkrai Sir I have the same query my data set has three class in that case how will I get ROC & AUC curve

  • @freddyflores6608
    @freddyflores6608 7 лет назад +2

    Thank you so much for your explanation, I could run my code and understand better the process.

    • @bkrai
      @bkrai  7 лет назад

      Thanks for the feedback!

  • @sriharshabsathreya
    @sriharshabsathreya 6 лет назад +1

    we an use Deducer package in R to directly run the ROC Curve
    library(Deducer)
    mymodel

    • @bkrai
      @bkrai  6 лет назад

      thanks!

  • @parasrai145
    @parasrai145 6 лет назад +3

    Great video and very well explained!

    • @bkrai
      @bkrai  6 лет назад

      Thanks for comments!

  • @muldon2
    @muldon2 6 лет назад +1

    for those who are looking for the data: stats.idre.ucla.edu/r/dae/logit-regression/

    • @bkrai
      @bkrai  6 лет назад

      Thanks for sharing!

  • @kalyanasundaramsp8267
    @kalyanasundaramsp8267 6 лет назад +2

    Sir, a)for multi-class, how you will will come with false positive, false negative b)how to compute ROC for multiclass

  • @jennykeeping8918
    @jennykeeping8918 6 лет назад +1

    Hi, I'm on R studio v. 1.1.423 now and nnet package isn't available and I can't seem to find an equivalent... any ideas what I can use to get the same results? Thanks.

    • @kamalpada1270
      @kamalpada1270 5 лет назад +1

      Please upgrade RStudio to atleast 1.1.463.. nnet works in this version.. Good luck.

    • @bkrai
      @bkrai  4 года назад

      This is old comment. I guess you must have already updated.

  • @miguelsuarez475
    @miguelsuarez475 7 лет назад +2

    You nailed it teacher..!!

    • @bkrai
      @bkrai  4 года назад

      Thanks!

  • @humbertobarino578
    @humbertobarino578 6 лет назад +2

    very very helpfull !!!
    im sending to some brazilians friends

    • @bkrai
      @bkrai  6 лет назад

      Thanks and hope you find other videos helpful too!

  • @visheshgour
    @visheshgour 5 лет назад +1

    i think m the only one who doesn't able to learn any computer language except R and this all happen just bcs of u sir 🙂

    • @bkrai
      @bkrai  5 лет назад

      Thanks for comments!

  • @avi20009
    @avi20009 7 лет назад +2

    Sir How does ROC curve work when the dependent variable is not binary in nature, in essence more than 2 factors for which we have to model the data(Note: but not continious in nature).

  • @devawratvidhate9093
    @devawratvidhate9093 6 лет назад +2

    How to handle if my all data is categorical my predictor features are subject columns with 1 to 8 grades for each subject and
    response variable is subject where we have to predict response variable grades (1-8 ).
    Before applying model I converted all features and response variable into factors is this right step or should i only covert response variable into factors and keep predictors in numerical format

    • @bkrai
      @bkrai  4 года назад

      I'm seeing this today, but for categorical variables you can try random forest:
      ruclips.net/video/dJclNIN-TPo/видео.html

  • @sengulozdemir418
    @sengulozdemir418 7 месяцев назад +1

    I appreciate your help, excellent video 👏🙏

    • @bkrai
      @bkrai  7 месяцев назад

      You are welcome!

  • @BeKindPlox
    @BeKindPlox 6 лет назад +2

    Great explanation!

    • @bkrai
      @bkrai  6 лет назад

      Thanks!

  • @SachinSingh-uh2xh
    @SachinSingh-uh2xh 7 лет назад +1

    Please do a video on sentiment analysis using R in detail... Deep dive analysis

    • @bkrai
      @bkrai  7 лет назад

      Thanks for the suggestion, I'll probably do it sometime this month.

  • @lianjek5788
    @lianjek5788 4 года назад +1

    hi sir, I have some confusion, please help me to resolve it.
    Your IV (admit) has two levels...0 and 1 and you performed multinomial logit?
    Is it obvious to plot a multiclass ROC rather than a typical ROC curve, when my IV has three levels ( i.e. 1,2, and 3).
    Thanks.

    • @bkrai
      @bkrai  4 года назад +1

      Multinomial logit works for 2 or more levels. However, ROC used here is only for situations where IV has 2 levels.

    • @lianjek5788
      @lianjek5788 4 года назад +1

      @@bkrai Would you please provide a lecture about multiclass ROC? Thanks.

    • @bkrai
      @bkrai  4 года назад +1

      Thanks, I've added it to my list.

  • @omkarthakur2251
    @omkarthakur2251 4 года назад +1

    Very good video sir it is very helpful

    • @bkrai
      @bkrai  4 года назад

      Thanks for comments!

  • @kessiezhang9357
    @kessiezhang9357 5 лет назад +1

    Hi
    Bharatendra, why don't you use glm()? I looked it up and it seems like multinom is used when the dependent has more than 2 levels. In your example, the dependent is admin(no, yes). That's why I'm confused why you chose multinom () instead of glm(). Thank you.

    • @bkrai
      @bkrai  5 лет назад

      multinom works for 2 or more. So when dependent variable has 2 levels, it should work fine.

  • @wanglaoshuwang3831
    @wanglaoshuwang3831 6 лет назад +1

    I have a question about the logistic regression model part. Does the code deal with the whole data? I thought when doing the logistic regression model, you have to divide the data into training set and test set. In the code you've used, does it divide training set and test set automatically?

    • @bkrai
      @bkrai  6 лет назад

      I used full data as focus was more on ROC. But when developing a prediction model it is always good to partition data into training and testing data sets.

    • @wanglaoshuwang3831
      @wanglaoshuwang3831 6 лет назад

      Thank you so much! helped a lot! :)

  • @avinashsingh357
    @avinashsingh357 7 лет назад +2

    Explained very neatly sir, appreciate if you can pls add dataset and code for learning please....

    • @bkrai
      @bkrai  7 лет назад

      email id?

    • @avinashsingh357
      @avinashsingh357 7 лет назад +1

      Thanks for your quick response sir, my emailid is avinashsinghemailid@gmail.com

    • @bkrai
      @bkrai  7 лет назад

      all set.

  • @bulletkip
    @bulletkip 5 лет назад +1

    absolutely excellent explanation. thank you very much.

    • @bkrai
      @bkrai  5 лет назад

      Thanks for comments!

  • @irondia73
    @irondia73 5 лет назад +1

    Hi Dr. Bharatendra Rai, would you be able to make a tutorial on building a logistic regression model using training and validation sets, with performance checking via ROC curve as you have done here? I know you posted one on linear regression, but I thought a logistic model would be very helpful too. Thank you!

    • @bkrai
      @bkrai  4 года назад

      You can get logistic regression from this link:
      ruclips.net/video/AVx7Wc1CQ7Y/видео.html
      ROC steps are already in the current lecture video.

  • @oumaimanassiri5555
    @oumaimanassiri5555 6 лет назад +4

    Hi Professor, i love your videos ,it's very interesting.I'm a PhD student , sometimes i find difficult to have the link between my own variables ( concengrations of elements) and the variables that you work with, that's why ; I wish you have documents well explained concerning the data processing analysis to sends it to me I will be very grateful . Also i want that you sent me the data file.

    • @bkrai
      @bkrai  6 лет назад

      email id?

  • @kpakpomoevi1603
    @kpakpomoevi1603 4 года назад +1

    Im new to ur lecture and find it very interesting and useful.I have one question ?how do you get the cutoff .5 from the classification table @5:39mn of the video.Thanks

    • @kpakpomoevi1603
      @kpakpomoevi1603 4 года назад +1

      I figure out and see it should be a default value.However I'm still having issue with the performance object eval.when I tried to print eval,it is giving me this :
      A performance instance
      'Cutoff' vs. 'Accuracy' (alpha: 'none')
      with 392 data points
      please help me have what you had on your screen.Thanks

    • @bkrai
      @bkrai  4 года назад

      You will see it better after plot.

  • @tanyachichekian9900
    @tanyachichekian9900 4 года назад +1

    Hello Dr. Rai. Would you happen to know what code I can use to compare 2 ROCs and/or AUCs using R? Also, if there is a way to represent 2 ROCs in one graph. Thank you! ~ Tanya

    • @bkrai
      @bkrai  4 года назад

      You can calculate AUC for different models and compare them. It should be higher the better.

    • @sanjibphukan8921
      @sanjibphukan8921 2 года назад

      @@bkrai Sir, in that situation how can we get the p-value of the compared statistic.

  • @nithinmamidala
    @nithinmamidala 6 лет назад +1

    very helpful video sir. thank you so much. I have a doubt how do you fix the threshold value as 0.5.

    • @bkrai
      @bkrai  6 лет назад

      Default threshold is already 0.5, there is no need to do anything for this.

    • @nithinmamidala
      @nithinmamidala 6 лет назад +1

      Ok.. Thankyou sir.

  • @TH-fe1vs
    @TH-fe1vs 7 лет назад +1

    can you tell me the differences below:
    yourmodel

  • @mohsinfayaz8103
    @mohsinfayaz8103 3 года назад +1

    How to generate ROC AUC curve for multi class responsible variable?
    Thank You

    • @bkrai
      @bkrai  3 года назад +1

      You can use this method with two class at a time.

  • @victorhenostroza1871
    @victorhenostroza1871 4 года назад +1

    Sir, based on this miss classification problem for admit =1, how can u change prob to other value in the model? maybe with under 0.45 =0 and over = 1

    • @bkrai
      @bkrai  4 года назад +1

      ROC automatically tries probability values from 0 to 1 and then plots it on the curve.

  • @abhilashiv3599
    @abhilashiv3599 6 лет назад +1

    Thank you so much sir, Just want to ask you whether type='response' is same as type='prob' when I am trying to give type='prob' , R is throwing an error like "Error in match.arg(type) :
    'arg' should be one of “link”, “response”, “terms” ?

    • @bkrai
      @bkrai  6 лет назад

      'Response' usually could be classes such as 'yes' or 'no'. But 'prob' gives probability values. And that could lead to errors that you are getting.

    • @abhilashiv3599
      @abhilashiv3599 6 лет назад

      @@bkrai Thank you Sir

  • @Chirag0729
    @Chirag0729 4 года назад +1

    Nicely explained. Thank you.

    • @bkrai
      @bkrai  4 года назад

      Thanks for comments!

  • @amitgajkal4821
    @amitgajkal4821 5 лет назад +1

    i am getting following error when i use : pred

  • @raj2385
    @raj2385 Год назад +1

    if am using 2 variables only GPA and Admit then what will be the Logistic Regression Model formula

    • @bkrai
      @bkrai  Год назад

      You can refer to this:
      ruclips.net/video/AVx7Wc1CQ7Y/видео.html

  • @sanjeevnair9893
    @sanjeevnair9893 5 лет назад +1

    Hello Sir , i'm an avid viewer of your videos which truly add value to our ML understanding. Have a quick question that once we determine the Best Value of Cut Offs post Model performance evaluation , should we go back and re-run the Model performance with Best Cut Off values and change the cut off of 0.5 that we considered as thumb rule.

    • @bkrai
      @bkrai  5 лет назад

      Yes, that's correct.

  • @francodjo
    @francodjo 7 лет назад +1

    Great video, trying to plot ROC for random fors but it giving me the following error
    Error in prediction(p2, train$Dispute) :
    Number of cross-validation runs must be equal for predictions and labels

    • @bkrai
      @bkrai  7 лет назад

      what are your previous lines before this line where you are getting error?

    • @francodjo
      @francodjo 7 лет назад

      no error. thank you very much for quick response can you forward me the co

    • @francodjo
      @francodjo 7 лет назад

      library(randomForest)
      TD_model

    • @bkrai
      @bkrai  7 лет назад

      i've sent my code file.

    • @bkrai
      @bkrai  7 лет назад

      note that the response variable here has two levels. If your data has more, then the codes shown in the video may not work as it is.

  • @fayarvin2003
    @fayarvin2003 3 года назад +1

    Really helpful!

    • @bkrai
      @bkrai  3 года назад

      Glad it was helpful!

  • @hamsinisankaran2435
    @hamsinisankaran2435 7 лет назад +2

    Thank you for a great tutorial sir. Could you please share the dataset and the code ?

  • @damodharand1519
    @damodharand1519 4 года назад +1

    I need to know how the medical data set are going to use in R studio programming and (example MIMIC, DCOM, etc) which library i have to use... pls if you know anyone inform...

    • @bkrai
      @bkrai  4 года назад

      You can try these:
      ruclips.net/p/PL34t5iLfZddsQ0NzMFszGduj3jE8UFm4O

  • @sanjibphukan8921
    @sanjibphukan8921 2 года назад +1

    Nicely explain. Sir, can you arrange to prepare a video on SVM of binary outcomes

    • @bkrai
      @bkrai  2 года назад +1

      Try this.
      ruclips.net/video/pS5gXENd3a4/видео.html

  • @abhitest1
    @abhitest1 5 лет назад +1

    Sir, can you please also attach dataset files/link along with your videos. This would greatly help us in learning by practicing with same data set. Thanks for great videos sir.

    • @bkrai
      @bkrai  5 лет назад

      I've added a link in the description area below the video.

    • @abhitest1
      @abhitest1 5 лет назад +1

      @@bkrai thanks.

  • @sovon08
    @sovon08 6 лет назад +2

    Thank you so much Sir..the video was really helpful in providing practical knowledge of dealing with predictive modelling problems in R..Can you please tell me how to apply weight of evidence/ fine classing in R - is there any ready made syntax?

    • @bkrai
      @bkrai  6 лет назад

      Thanks for your feedback! I've also added your suggestion to my list.

  • @lewsmash
    @lewsmash 6 лет назад +1

    Hi,
    Firstly, great video this really helped me to understand the ROC curve and implement it with my data in R. I am analysing diagnostic data for a masters degree research project. I wanted to know how to identify the cutoff value from the value that we take from the accuracy versus cutoff curve or the final ROC curve. The scale goes from 0-1 but my independent variable data ranges from 100 to 10^7 . In short, how do I take the best cutoff value that this analysis outputs and relate/convert this to my independent variable and an exact cutoff value?
    Thanks very much.

    • @bkrai
      @bkrai  6 лет назад

      Cutoff value is for the dependent variable. If the dependent has two categories 'yes' and 'no', then default cutoff is probability = 0.5. ROC provides information on how prediction model performance will change if cutoff value changes from 0 to 1.

  • @merumomo
    @merumomo 6 лет назад +1

    Thank you for this great video! And thank you for prompt reply. I have questions.
    If we are doing machine learning, we need to create ROC using predictive model created by test set, correct?
    (in your "Logistic Regression with R" video, you created predictive model using test set. We need to validate the accuracy of the model). Also, if I want to use which.max func to plot the highest values on the eval plot, what code should I use?

    • @bkrai
      @bkrai  3 года назад

      Seeing this today. But roc curves can be done for both train and test data.

  • @ShahbazAhmedENG
    @ShahbazAhmedENG 3 года назад +1

    multinom is not working?? please i upload the csv file but the multinom fuction generate an erro : multinom function require two classes or more.. please help me

    • @ShahbazAhmedENG
      @ShahbazAhmedENG 3 года назад

      Dr. Bharatendra Rai sir please

    • @bkrai
      @bkrai  3 года назад

      Check your response variable. It should have 2 or more classes.

  • @abdulazeez9863
    @abdulazeez9863 6 лет назад

    I have applied the same functions in evaluation of my GAM model where I am not able to produce the confusion matrix. The results shows 2*132 table matrix instead of 2*2 matrix moreover I have 203 'Y" variable in validation data. Why its coming so. Plz help me. Thanking you.

  • @jaituteja88
    @jaituteja88 7 лет назад +2

    Great explanation. Thank you Sir! )

    • @bkrai
      @bkrai  4 года назад

      Welcome!

  • @fatimabadi3335
    @fatimabadi3335 5 лет назад +1

    Thanks for the useful information. I would like to ask you if I can use ROC to measure the effectiveness of the prediction model? And can I use ROC in R software?

    • @bkrai
      @bkrai  5 лет назад +1

      For model effectiveness you can use AUC. Also yes you can do ROC in R, this video gives you all the steps.

    • @fatimabadi3335
      @fatimabadi3335 5 лет назад

      Dr. Bharatendra Rai thanks

  • @balasrm1
    @balasrm1 5 лет назад +1

    Further to my earlier comment, also wanted to ask what software you used to create these videos on data analytics. Thanks

    • @bkrai
      @bkrai  5 лет назад

      I used iMovie.

  • @dorothymartin2477
    @dorothymartin2477 Год назад

    Hi Dr, when i do stacking of ensemble why do i get the roc curve in triangular shape?

  • @juarezantonio656
    @juarezantonio656 Год назад

    Unfortunately unable to proceed.
    An error message appears:
    pred

  • @tapangautam8746
    @tapangautam8746 7 лет назад +2

    Hello sir, Video was excellent but I have small question. Cut off value changed to 0.45 so do we need to again run model. If not then why and if yes then what changed need to be made in code so that pure classification can be made for improving accuracy level of model.

    • @bkrai
      @bkrai  7 лет назад

      You don't need to run model again. The model gives predictions in terms of probabilities. When we use cutoff of 0.45, probabilities that are less than 0.45 are classified into first group and those above 0.45 are classified into 2nd group. So for any cutoff value, probabilities from prediction model are not going to change, they are only used for classification.

    • @tapangautam8746
      @tapangautam8746 7 лет назад

      Thank you sir for your reply and for guiding me
      I am want to know one more thing, as according to your reply I can assume probability less then 0.45 as yes and above 0.45 as no in terms of dependent variable. But where or on which sort of output I will come to know that due to variation in independent variable majority of time dependent variable occur (yes or no).
      Thanks

    • @bkrai
      @bkrai  7 лет назад

      Independent variable will take only one value for each case. If a new student has GRE = 380, GPA = 3.61 and comes from a school ranked 3, then inputting these values in the prediction model will give probability = 0.18. Since p=0.18 is less than cutoff = 0.45, this student will be rejected and will not get admission.

    • @tapangautam8746
      @tapangautam8746 7 лет назад +1

      Good morning, now I understand answer. Thanks for helping me in compression of logistic regression

    • @tapangautam8746
      @tapangautam8746 7 лет назад

      Hello Sir
      I need little help from your side. In code where we create subset out of sample why == TRUE or FALSE is used. What is difference between = and == symbol in Logistic regression.
      Thank you sir

  • @sriharshabsathreya
    @sriharshabsathreya 7 лет назад

    sir, how to identify the optimal cut off value so that TPR (sensitivity) will be high ?

  • @edneideramalho2363
    @edneideramalho2363 5 лет назад +1

    Thanks for the video! Amazing!

    • @bkrai
      @bkrai  5 лет назад

      Thanks for comments!

  • @cameronyi683
    @cameronyi683 7 лет назад

    Hello. This looks like the cutoff would be the probability of a certain student getting admitted based on the multinomial model. If I am working with a dataset with one independent variable contributing to my multinomial model, and am wanting to obtain a cutoff value from that independent variable (ie what is the cutoff value of SAT score that will best tell me if someone is admitted to college), what would I be changing in the code? Thank you.

    • @lewsmash
      @lewsmash 6 лет назад

      I was also looking for this answer to this if you managed to find out?

  • @rafiqulislam1085
    @rafiqulislam1085 2 года назад +1

    Very nice and Excellent explanation. Could you please make another video to draw multiclass (more than 2 class) Roc curve. (one vs rest roc )?

    • @bkrai
      @bkrai  2 года назад

      You can refer to these:
      ruclips.net/video/ftjNuPkPQB4/видео.html
      ruclips.net/video/6SMrjEwFiQY/видео.html

  • @vishnukowndinya
    @vishnukowndinya 5 лет назад

    Hi Sir, i have calculated cut off for accuracy for my data (~0.475). i would like to know where exactly i should replace default 0.5 with this .475 ?

  • @suryagaur7440
    @suryagaur7440 6 лет назад +2

    Hi Sir,
    Thanks for wonderful video.
    Could also make AUC video where dependent variable is continuous.

    • @bkrai
      @bkrai  6 лет назад

      Thanks for the suggestion! I've added this to my list.

  • @subaganesh552
    @subaganesh552 4 года назад +1

    Hello sir.. is possible to calculate auc metric for multiple class prediction

    • @bkrai
      @bkrai  4 года назад

      I'll look into this.

    • @subaganesh552
      @subaganesh552 4 года назад +1

      @@bkrai thank you sir...

    • @bkrai
      @bkrai  4 года назад

      welcome!

  • @sanjibphukan8921
    @sanjibphukan8921 2 года назад +1

    How to compare two ROC? kindly explain, sir.

    • @bkrai
      @bkrai  2 года назад +1

      You can refer to this:
      ruclips.net/video/ftjNuPkPQB4/видео.html

    • @sanjibphukan8921
      @sanjibphukan8921 2 года назад

      @@bkrai Sir can we adjust variables in SVM as we usually do in any MLR analysis. If, possible, how we can proceed. Please suggest.

  • @mashalnabh2747
    @mashalnabh2747 5 лет назад +1

    Hello, This is great video but I am slightly confused with the probability explanation. YOu mentioned that if the prob prediction score is less than .5 then chance are less than average but doesn't it depend on %age of events in the data on which model is based? If in the sample data, the event rate is 1 out of 4 then probability is .25, so any scores above .25 in the final output mean model is saying that this has higher chances. Not necessarily it has to be above .5.

    • @bkrai
      @bkrai  5 лет назад

      In this data there are only two outcomes. Either a student is admitted or not-admitted. If you only calculate % as students admitted and use that as a probability, it will not be a very useful prediction. Because it will remain same for every students irrespective of their gpa, sat score or college they are coming from. The prediction in the form of probability between 0 and 1 here is based on all input variables.

    • @mashalnabh2747
      @mashalnabh2747 5 лет назад +1

      @@bkrai Hello, Let me give an example: Lets say in your data set , there were 100 observations and 10 "events" of passing . So the probability of an event in your data set is 10/100= 0.1 . Now I build the prediction model based on different predictors like Gap, SAT, gender etc, and I get the final predicted probability scores. Lets say for some student, we get a score of .3,. All I want to say is that this student has higher chances than average. Not necessary the probability score has to be above .5. This depends upon the original percentage events in your data on which the model is built - which in this case is 0.1

    • @bkrai
      @bkrai  5 лет назад +1

      For original %, let's consider your own data. Let's say 10 out 100 applicants get accepted giving a rate of 0.1. Now forget about any type of model and simply classify all 100 students as getting accepted. Here without any model, you will be correct 10% of the time, but incorrect 90% of the time. Do not mix-up this overall rate with individual probability. When you develop a prediction model, it should give overall accuracy better than 90% to be of any value.

    • @sachin01663
      @sachin01663 5 лет назад +1

      @@bkrai Thanks. Am I right to say that predicted probability score less than .5 (after building the final model) does not necessarily mean that, that event is less likely to happen.

    • @bkrai
      @bkrai  5 лет назад +1

      That's correct.

  • @dba99999
    @dba99999 7 лет назад +2

    Great Video....

  • @fabioambrosini5265
    @fabioambrosini5265 5 лет назад +1

    what does represent the color from green to red of the ROC?

    • @bkrai
      @bkrai  5 лет назад +1

      It represents cutoff values between 0 and 1.

    • @fabioambrosini5265
      @fabioambrosini5265 5 лет назад

      Dr. Bharatendra Rai thanks

  • @Hehxgdudh
    @Hehxgdudh 6 лет назад +1

    I get a confusion matrix that looks like (201,0; 45,0) with an accuracy of 1... HELP

    • @bkrai
      @bkrai  6 лет назад

      Make a confusion matrix for test data, that will provide more practical accuracy.

  • @veianthanjayaramu2995
    @veianthanjayaramu2995 3 года назад +1

    Thank you very much, sir.

    • @bkrai
      @bkrai  3 года назад +2

      You are welcome!

  • @uhsay1986
    @uhsay1986 6 лет назад +1

    Hi Sir , why did you use multinom function here ? isnt multinom used only if target var have more than 2 categories ? while in this video we have only 2 categories , yes or no ?

    • @bkrai
      @bkrai  6 лет назад

      Multi works for 2 or more, so using it here should be ok.

  • @ahmedbilal1831
    @ahmedbilal1831 3 года назад +1

    Thanks alot man. you helped

    • @bkrai
      @bkrai  3 года назад

      You are welcome!

  • @shubhammishra8550
    @shubhammishra8550 6 лет назад +1

    why u wrote 'y.values' in slot function of AUC?

    • @bkrai
      @bkrai  6 лет назад

      When you run 'eval' that contains that contains accuracy values for various cut offs, you can see different type of information are stored in various slots. And y.values contain data on accuracy.

  • @rishikeshdash12
    @rishikeshdash12 2 года назад

    Sir, I have ran the model using neuralnet package , is it necessary to calculate probability for predicted value or we can directly go with value obtained for test set. One more question sir is it any way to plot roc curve for two models.

  • @prahladbhat9516
    @prahladbhat9516 4 года назад +1

    How do I do this AUC I have NA Values in my dataframe?

    • @bkrai
      @bkrai  4 года назад

      You need to first address missing values. See this link:
      ruclips.net/video/An7nPLJ0fsg/видео.html

  • @ranjitshekdar1720
    @ranjitshekdar1720 7 лет назад +2

    Would you be able to send me the dataset used for this please? awesome job done.

    • @ranjitshekdar1720
      @ranjitshekdar1720 7 лет назад

      rshekdar@gmail.com my email id. sorry accidentally clicked send before could complete.

    • @bkrai
      @bkrai  7 лет назад

      +Ranjit Shekdar all set.

    • @ranjitshekdar1720
      @ranjitshekdar1720 7 лет назад +1

      you sir are very fast, much appreciated. Thanks again. I was not expecting it to be sent out so fast.

  • @tamaraabzhandadze2712
    @tamaraabzhandadze2712 3 года назад +1

    How could we cross validate these results?

    • @bkrai
      @bkrai  3 года назад

      You can refer to this for more detailed coverage:
      ruclips.net/video/ftjNuPkPQB4/видео.html

  • @umairrasool9580
    @umairrasool9580 4 года назад +1

    Hello Sir, Thanks for your every lecture for students like me. Respected Sir, i am working on Groundwater potential mapping, i have compete my analysis using ArcGIS, now i have to validate my results with ROC curve, could you please guide me how could it possible using R. I will be very thankful to you.

    • @bkrai
      @bkrai  4 года назад

      You can follow the steps in this video. If you get struck anywhere, let me know.

    • @umairrasool9580
      @umairrasool9580 4 года назад +1

      @@bkrai Sir i have get problem here, please check, how to make model with my data?
      setwd("C:/Users/Umair/OneDrive/Manuscript Quetta GWP/ANN")
      .libPaths("C:/Users/Umair/OneDrive/Manuscript Quetta GWP/library ANN")
      install.packages("tidyverse")
      library(tidyverse)
      Wells

    • @bkrai
      @bkrai  4 года назад

      Which part has a problem?

    • @umairrasool9580
      @umairrasool9580 4 года назад

      @@bkrai Sir problem to make model for ROC and need to compute the true + and true -.

  • @vairachilai3588
    @vairachilai3588 5 лет назад +1

    True Postive Rate= True Positive/Actual admit ( 29/(29+20) =0.591
    Fales Positive Rate= False Positive/Actual not admit= (98/(253+98)=0.279

    • @bkrai
      @bkrai  5 лет назад

      In the confusion matrix shown in the video, the above formula that you use will not be correct. In the confusion matrix if you have actual class on the LEFT and predicted class on the TOP, then your formula will be correct.

  • @zhongyanxu9047
    @zhongyanxu9047 3 года назад

    Amazing, very usefull, Thanks

    • @bkrai
      @bkrai  3 года назад

      You are very welcome!

  • @francodjo
    @francodjo 7 лет назад

    I need some help please trying to submit my project but cannot get ROC work

  • @rohithebbar722
    @rohithebbar722 6 лет назад

    Hello sir, can you make a video for ploting ROC curve for SVM. I am getting an error in my code. The error is i am getting is format of prediction is invalid. Thank you

  • @sheikhseerat7105
    @sheikhseerat7105 4 года назад +1

    Very nice explanation
    Could u plzz send me dataset and code

    • @bkrai
      @bkrai  4 года назад

      For data, see link below video. For code, see the pinned comment.

  • @abhibhavsharma8706
    @abhibhavsharma8706 4 года назад +1

    In place of "tpr", will a numeric entry work?

    • @bkrai
      @bkrai  4 года назад

      I've not tried, but should work.

  • @1612kanika
    @1612kanika 6 лет назад

    how to check performance for ordinal logistic regression model. By using this method I am getting this error
    'Error in prediction(pedi, heart_A$num) :
    Number of cross-validation runs must be equal for predictions and labels.'

    • @1612kanika
      @1612kanika 6 лет назад

      this is my email id kanikaabap@gmail.com please help me out thank you

    • @bkrai
      @bkrai  6 лет назад

      For ordinal logistic regression you can use this:
      ruclips.net/video/qkivJzjyHoA/видео.html

    • @1612kanika
      @1612kanika 6 лет назад

      I already checked that video but not able to plot roc curve getting error " number of cross-validation runs must be equal for predictions and labels' .

  • @kenzagharouit4041
    @kenzagharouit4041 6 лет назад +1

    thx , very good video.

    • @bkrai
      @bkrai  6 лет назад

      Thanks for comments!

  • @olamidegab2390
    @olamidegab2390 3 года назад +1

    Hello, can this be applied on CNN?

    • @bkrai
      @bkrai  3 года назад

      If there are two classes, then yes.

  • @sateeshkumarojha7414
    @sateeshkumarojha7414 4 года назад +1

    please avail me the codes of ROC Curve & Area Under Curve (AUC) with R - Application Example

    • @bkrai
      @bkrai  4 года назад

      For code, see the pinned comment.

  • @petrusk3185
    @petrusk3185 6 лет назад +1

    I'm a bit late to the party here, but sure You cannot compare total amount emitted vs accuracy of the model?
    In the dataset 68% were admitted, but that data is 100% accurate. If the model is 70% accurate you could get a result between 273 +-30%. Or am I missing something here? You are comparing apples with oranges?

    • @bkrai
      @bkrai  6 лет назад

      Thanks for comments! Probably you missed that "1" represents "admitted" in this data. Number of students admitted should be 31.75% and not 68%.

  • @roopashastri9908
    @roopashastri9908 7 лет назад

    Hi Sir,
    One question here.
    Can ROCR curve be drawn in case of multiple classification.
    Ex:We have to classify the given data in 3 different classes?
    Thanks,
    Roopa

    • @bkrai
      @bkrai  7 лет назад

      It is meant for only two categories. If you have three classes such as 1, 2, and 3. And if your interest lies more in correctly classifying say class "3", then you may still have two classes with 3 and others.

    • @roopashastri9908
      @roopashastri9908 7 лет назад

      ok..got it.Thank you sir!

  • @puneetksidhu
    @puneetksidhu 6 лет назад

    Hello Sir, are precision and recall same as sensitivity and 1-specificity respectively?

    • @puneetksidhu
      @puneetksidhu 6 лет назад

      if we would have to plot a PR curve for the same data that you used to plot ROC curve, how would we do that? Can you please send me the code at puneetkaursidhu@gmail.com