caret package webinar

Presented by Max Kuhn for the Orange County R User Group. Organized and recorded by Ray DiGiacomo, Jr. (President, OC-RUG, [e-mail])

[@ironfrown]

Excellent presentation, I have been using the "caret" package for a while but only now I am digging into its parallel features - this is so exciting!

[@manudotc]

amazing stuff..your book is my the best "go-to" book ..Thanks so much !

[@dubeya01]

Thanks Max. I have been using caret and it's really good! Looking forward to incorporation of multi-class classification in caret.

[@maxkuhn1047]

Thanks for the comment. caret has always been able to do multi-class models, so give it a try.

[@dubeya01]

Max Kuhn Hi Max, thank you so much! You have saved me and humanity countless hours of (unnecessary) coding. Also, I can say with certain degree of confidence that the multi-class classification was not working in caret earlier. I did a re-installation of the package after reading your comment, and now it's working fine. I have also placed an order on Amazon for your book! Best regards, Apurva

[@LuciolaSama]

Many thanks! This webinar seems like a great start for people who are originally not from the field of machine learning. Time for me to do some awesome molecular biology classification with the caret package! :)

[@stanislavmartsenyuk]

Amazing webinar on amazing package!

[@riskyalex8428]

all chunks work great. Thanx a lot for sharing!

[@djangoworldwide7925]

Max you're a legend

[@brotherrain1024]

thank you so much, Max Kuhn

[@rishavbagri2271]

Thank you so much, Max!

[@chrisbaskerville953]

The roc curve is generated by the roc function that's part of the pROC package. It's hard to see in the presentation and it sounds like max is calling it "prc".

[@9906201a]

Max Kuhn starts at 03.45

[@geoffrey_stoel]

Thanks... very nice starter on caret...

[@OscarForero]

Following the steps of the demo I get the following error: Error in train.default(x, y, weights = w, ...) : wrong model type for classification My caret version is caret_6.0-24 Was the demo done with a different version?

[@Sigmaduos]

Hello Max. Thanks for this. I really want to learn one thing: In cross validation within training, is the data sampled randomly into K partitions, or is stratified sampling done? Just like when splitting the data at the begin into training and testing sets with stratified sampling, I was wondering whether the same was done .

[@manasishah2477]

Hi Max, Thanks for a great tutorial. I followed most of the R code presented. When i run the full gbm model with the ctrl and grid parameters specified, I get an error "Error in train.default(x, y, weights = w, ...) : The tuning parameter grid should have columns n.trees, interaction.depth, shrinkage, n.minobsinnode". Even though I have specified all these columns except for n.minobsinnode. Could you explain what this might be? Thank you, Manasi

[@maxkuhn1047]

Manasi Shah We updated the package code so that another parameter (n.minobsinnode) is also tuned. You can add it to the tuning grid. I generally keep it at the default value (n.minobsinnode = 10).

[@manasishah2477]

Max Kuhn Great, I added the parameter and the model worked fine! Thank you! :)

[@laurad1174]

I used this package but the fonction "train" is running during around 1 hour without any result, is it normal ? How to reduce the calculate time ? Thanks a lot.

[@sharp7230]

Max, I am trying to run the CreateDataPartition function after installing the caret library and am getting an error message saying Error: could not find function "createDataPartition". Do you have any idea why this may be happening? Thanks in advance!

[@InfiniteFlash]

Did you load in the package with : library(caret); ? Also, did you download it straight from the CRAN server?

[@viejacalesita]

are the slides available?

[@maxkuhn735]

Yes. See appliedpredictivemodeling.com/blog/2014/2/28/caret-webinar-materials

[@viejacalesita]

Max Kuhn Thanks for the slides and all the work around the package!!

[@zhenhuahe5188]

可惜我的听力不行

[@sa280290]

The pronunciation is unclear... don't understand half of the things you said.