Data Science - Part V - Decision Trees & Random Forests

Рет қаралды 78,244

Күн бұрын

For downloadable versions of these lectures, please go to the following link:
www.slideshare.net/DerekKane/p...
github.com/DerekKane/KZfaq-...
This lecture provides an overview of decision tree machine learning algorithms and random forest ensemble techniques. The practical example includes diagnosing Type II diabetes and evaluating customer churn in the telecommunication industry.

Пікірлер: 33

@graciehan8948 7 жыл бұрын

Learned a lot today. Thank you, Derek!

@nayakool 8 жыл бұрын

Excellent!! It really helps me solidify some of the concepts used in DT and its practical applications !!

@DerekKaneDataScience 8 жыл бұрын

+rakshith nayak Thank you and I am glad that you found some value here. Keep it up!!!

@ShubhamAshokGandhi 7 жыл бұрын

Hi @Derek Kane. Thanks for the video. It was helpful. Can you share the dataset please?

@rjronquillo4488 7 жыл бұрын

Very nice video! Keep up the great work!

@hunhwasong8800 8 жыл бұрын

Thank you for the video! such a useful lecture.

@DerekKaneDataScience 8 жыл бұрын

+hunhwa song You are very welcome.

@bhavanichatrathi7435 7 жыл бұрын

nice explanation,can you please share the links to data sets

@m.anisulkarim7819 7 жыл бұрын

Excellent. Really helpful. It would be much appreciated if you provide a link for data source and code.

@fatmaghorbel6579 6 жыл бұрын

could you apply an implementation of algorithm of CART with python or any other language based on an easy example ??

@jensharbers5620 6 жыл бұрын

Is the green in 29:14 not the diagonal and red the off diagonal? Like in matrices?

@arasuthiru8548 5 жыл бұрын

Very good overview and examples a very useful.

@cragnosliw4685 7 жыл бұрын

hi Derek, can you pls share the dataset and R scripts?

@gregweston20 6 жыл бұрын

Great video

@sourabhiitg1 8 жыл бұрын

Derek, your presentations are very helpful. Would it be possible to give a link for the various datasets that you are using? Some of them are in R but e.g. the phone company 'churning' dataset is not available. If I could practice on the real datasets that would be a better learning experience

@DerekKaneDataScience 8 жыл бұрын

+sourabh sriom Absolutely... Please pm me and I will get you setup on my dropbox account which has the R scripts/ datasets. I will be creating a GitHub account eventually to centralize all of this content but hopefully this will work in the meantime. Good luck and let me know what I can do to help.

@mojazzz284 5 жыл бұрын

While I appreciate all the efforts in putting up this presentation, I hopefully wish that you explained the methodology of all these concepts with some practical explanations / algorithms to help us better understanding on the applications.

@rezadarvish8116 7 жыл бұрын

Thanks a lot very indepth talk

@99chintu 8 жыл бұрын

Can you please share a link to the datasets..Thanks

@DerekKaneDataScience 8 жыл бұрын

+99chintu please pm me and I will get you setup on my dropbox account which has the R scripts/ datasets. I will be creating a GitHub account eventually to centralize all of this content but hopefully this will work in the meantime. Good luck.

@hannachuray4614 8 жыл бұрын

Hi Derek, sorry for asking, but did you create a GitHub account for the content? I am very interesting to see the R script, if it is possible. Thank you

@hannachuray4614 7 жыл бұрын

no, sorry

@LBizKid04 7 жыл бұрын

I just sent him a PM, hopefully he'll get back to me :)

@lariksonfar 8 жыл бұрын

please re-explain at 9:09 information gain...this entropy stuff is highly unclear as to how increase in disorder means its a useful splitting criteria...and what does it mean to just "increase in disorder"? are you transforming the data in some way to test something?... overall great video

@aueret 8 жыл бұрын

first think of entropy being a measure for uncertainty and secondly imagine a decision tree that at some node (representing a specific feature) divides a data set into subsets. The "cleaner" these subsets are (in terms of containing individuals belonging to a certain class when doing a classification problem) the better the feature is. The disorder you are talking about is actually not a disorder but the division ability of a feature in regard to the classes. hopefully understood your problem correctly and that this helps.

@test-mm7bv 6 жыл бұрын

not very helpful. need far more mechanical details. how are purity metrics used? how to combine multiple trees? how to determine variable importance.

@mojazzz284 5 жыл бұрын

couldn't agree more with you... basically he just kinda of narrated the text on the slides most of the time... not much help for newbie like me

@yrnagaraju6685 3 жыл бұрын

Poor quality of audio recording.

@iGatflow 8 жыл бұрын

Somehow he hasn't said a word about actually building a DT... 'We will build a CART model' and the next sentence 'We built a CART model' ? Really? Was it magic? If you're looking for anything practical, skip this video. If you're looking for theory, the video is still too thin

@DerekKaneDataScience 8 жыл бұрын

Hi Kole. I'm sorry that this video lecture did not give you what you are looking for and I appreciate the feedback. It is somewhat difficult to strike a good balance between theory/practical applications/algorithm optimization / code and my intention was to expose the viewer to the basic mechanics where they can build off in the direction that best suits their needs.