Hi Sir, by far your videos have been the best ever and I have undoubtedly learnt and benefitted by it. I never subscribe to channels but this definitely needs a thumps up :) Keep up the great work and I look forward to your future videos.
Very nice three part video. Any chance you can do a video on anomoly detection techniques? I've been using isolation forest from solitude package in R but there aren't many videos on it for R.
very nice video, im just wondering as to why you put set.seed(222) for 222. what does this actually do? i get quite a large misclassification for my data
Good evening doctor. The package "caret" gives a function "confusionMatrix" with Sensivity, Specifity and other statistics that can be easily applied to predictions and reference. Also, thank you for your videos. I got a question, ¿does multinom accept, like you are using, features/variables with diverse type like, multifactors, binomials or continuous data without a problem? ¡Have a good day!
Sir, for my own data when I followed these steps, misclassification error for training data is around 7.2% but misclassification error for testing data is around 91%. Why such a huge difference, what do we do in such case? Thank you so much for making this series, it has helped me in building my own model.
@@bkrai Yes sir I re-checked if I made some mistakes. My confusion matrix for testing data is not proper, as the classifications (0, 1, 2) on row and column are coming out differently. It is 0 1 2 on row and 2 0 1 on colunm. Perhaps that is why when I run sum(diag(tab1))/sum(tab1) the results are strange. Please guide me on how to fix this issue. Your guidance will help me in completing my model since it is the last step and I have been investing a couple of days in learning about multinomial logistic regression from your channel. Thank you
@@bkrai Yes sir with this suggestion I could fix the issue. All problems are resolved, finally, my model is complete. Thank you so much for helping me out, it really means a lot.
hi sir, followed the same way but when I went for the confusin matrix of the testing data sets its showing .How to change the vector into data frame? How to fix it? Thanks.
Excellent trainning and testing approach! Thanks a lot
You are welcome!
Waiting for more videos like this. Thankyou.
Thanks for comments!
Hi Sir, by far your videos have been the best ever and I have undoubtedly learnt and benefitted by it. I never subscribe to channels but this definitely needs a thumps up :) Keep up the great work and I look forward to your future videos.
Wow, thanks!
Very nice three part video. Any chance you can do a video on anomoly detection techniques? I've been using isolation forest from solitude package in R but there aren't many videos on it for R.
Thanks for the suggestion, I've added it to my list of future videos.
Thank you
You're welcome!
very nice video, im just wondering as to why you put set.seed(222) for 222. what does this actually do? i get quite a large misclassification for my data
You can refer to this for more details on data partitioning:
ruclips.net/video/aS1O8EiGLdg/видео.html
predict() function outputs probability. So how do you convert that value in labels e.g. Label 1, 2, 3 in this example?
Include type = 'class'
You can refer to this for more details:
ruclips.net/video/ftjNuPkPQB4/видео.html
Informative and very much helpful. Please make a tutorial on graph convolution neural network for binary classification for non-image dataset. Thanks.
Thanks for the suggestion, it's on my list of future videos.
Good evening doctor. The package "caret" gives a function "confusionMatrix" with Sensivity, Specifity and other statistics that can be easily applied to predictions and reference.
Also, thank you for your videos. I got a question, ¿does multinom accept, like you are using, features/variables with diverse type like, multifactors, binomials or continuous data without a problem? ¡Have a good day!
Yes, that should work fine.
Can we calculate ROC & Area Under Curve for Multinomial Logistic Regression?
You can try 2 levels at a time.
@@bkrai Sir can u share a video on this?
Sir, for my own data when I followed these steps, misclassification error for training data is around 7.2% but misclassification error for testing data is around 91%. Why such a huge difference, what do we do in such case?
Thank you so much for making this series, it has helped me in building my own model.
I think you are making some calculation error. Check and see if 91% is actually accuracy and not misclassification.
@@bkrai Yes sir I re-checked if I made some mistakes. My confusion matrix for testing data is not proper, as the classifications (0, 1, 2) on row and column are coming out differently. It is 0 1 2 on row and 2 0 1 on colunm. Perhaps that is why when I run sum(diag(tab1))/sum(tab1) the results are strange. Please guide me on how to fix this issue. Your guidance will help me in completing my model since it is the last step and I have been investing a couple of days in learning about multinomial logistic regression from your channel.
Thank you
Try adding one more line after line-16 where you do the same thing for test data.
@@bkrai Yes sir with this suggestion I could fix the issue. All problems are resolved, finally, my model is complete. Thank you so much for helping me out, it really means a lot.
Thanks for the update!
Good morning doctor, how can i get the coefficient of the equation not referred to the reference level?
That's not required. You can see that class-1 is easily predicted using the model.
hi sir, followed the same way but when I went for the confusin matrix of the testing data sets its showing .How to change the vector into data frame? How to fix it? Thanks.
Look at the structure of data using str function and make sure response is a factor variable.
👍
Thanks!