PyCaret - Accelerate your Machine Learning Insights Cycle

[@ranjanirajamani7565]

Its a very helpful library. Thank you for the demo.

[@rushikeshbulbule8120]

Very powerful module.... Thanks for this.. . 👌

[@TechnicallyData]

As usual, learnt a lot and excellent video.

[@rampratapa8101]

Super info bro👍🤝... Loved this pycaret😊

[@balinasuryachandra9337]

What can we interpret from the AUC plot and precision plot and classification boundary How to understand the SHAP value 11:06 of the variable. it has two bars on either side of the line What we would say from those. I would be happy if you take one complete video on AUC, precession, recall saying why they matter most in for evaluating the model.

[@AIEngineeringLife]

Balina.. Will do.. That is in plan in one of future video where I will be walking through metrics and explanation of output end to end. This video is more to show feature of pycaret

[@balinasuryachandra9337]

@@AIEngineeringLife Thanks for the Reply

[@vaishalibalaji886]

Hi Srivatsan sir, thank you for the demo. I have one doubt. Is this package useful only for classification models?

[@AIEngineeringLife]

It supports regression as well unsupervised problem. You can check the model supported on their website - pycaret.org/

[@mukeshkund4465]

Hi Sir, Small doubt here..When we are using Pycaret Setup then churn_setup becomes a tuple..If we want to see all the transformed column name which are transformed with onehot encoding then how we can see all the name ??

[@hari141v]

very informative video, thank you

[@saurabhjain507]

Thanks Sri for this quick video. As a Data Scientist, I would be sceptical of using this package. there 's so much abstraction and too little human inference from the plots that the resultant analysis may not be reliable.

[@AIEngineeringLife]

Saurabh.. this is not a replacement for human but an assisstant.. feature engineering and post hoc analysis is still manual. This just automated the model section and hyperparameter cycle . You can check my video here on how automl can help kzbin.info/www/bejne/iKq3ZWqJoLebo8k

[@shwetaredkar734]

Informative and helpful video. I have a request, please make a tutorial on how to perform 10-fold cv and repeated 10 fold cv and calculate all the performance assessment parameters while performing this. Also, how to test an independent dataset on a trained model.

[@AIEngineeringLife]

Shweta..Sure and that is in the plan.. Have some backlogs to complete before I get there

[@shwetaredkar734]

@@AIEngineeringLife okay

[@mukeshkund4465]

Nice Video Sir. Just a small question...Compare model command--> All the models readily avail in pycaret or we need to specify them ??

[@AIEngineeringLife]

It will run with all available models but we can control as well on what to exclude. Also there are methods for individual models to iterate further

[@mukeshkund4465]

@@AIEngineeringLife thank you..i will explore more for further information

[@shaikrasool5520]

sir iam trying to impliment pycart in NLP project.. input--> from pycaret.nlp import * exp_nlp = setup(data='df', target = 'user_suggestion') but iam getting error SystemExit: (Type Error): data passed must be of type pandas.DataFrame or list my df is PANDAS DATAFRAME.. what could be the reasaon

[@AIEngineeringLife]

You have given off in single quotes, remove that one and it must work

[@neemadan]

Great tutorial. I am getting an error "LightGBMError: Do not support non-ASCII characters in feature name" while using pycaret ML library. I found this related GitHub issue and wanted to check if you have across the same issue and got any suggestions. github.com/microsoft/LightGBM/issues/2478. For now, I tried a temporary solution "compare_models(fold=5, blacklist = ['lightgbm'])" and it's working.

[@AIEngineeringLife]

Neeraj.. I did not face this issue. Looks more like LightGBM but that has propagated. Pycaret still uses other libraries

[@sohamkulkarni523]

Very Informative. Could you pls share the .ipynb notebook?

[@AIEngineeringLife]

It is in my repo here github.com/srivatsan88/KZbinLI/

[@balajivs3206]

Can you share the ipynb link of the document

[@AIEngineeringLife]

Check my git repo - github.com/srivatsan88/KZbinLI

[@bharatthakur3642]

Nice video sir. I am getting memory error when using airnb data boston and seattle . Any solution

[@AIEngineeringLife]

Bharat are you running in colab or local?.. see if u can sample if dataset is very large and you dont have lot of memory

[@bharatthakur3642]

@@AIEngineeringLife running at kaggle

[@muhammedsinan3156]

Great content bro.. expecting a flask deployment vedio soon

[@AIEngineeringLife]

I already have a flask video. Are you looking for anything in specific? You can check my playlist for flask video kzbin.info/aero/PL3N9eeOlCrP5PlN1jwOB3jVZE6nYTVswk

[@rajaharshachinta]

This is a good example, can you add the GitHub link for the demo?

[@AIEngineeringLife]

github.com/srivatsan88/KZbinLI/ check pycaret.ipynb

[@madhu1987ful]

where is the github link for code and dataset link?

[@ramchandracheke]

Thank you sir 😊

[@rahatrezasulemani2862]

Hi, How can I create model with my own set of parameters?

[@AIEngineeringLife]

In this you cannot set parameters except for models to run against. I would suggest to use regular hyper parameter search software in your case

[@RajaSekharaReddyKaluri]

For eg: The categorical feature column would be [red, blue, green, yellow, orange], but each sample can belong to multiple categories (such as (red, green)).Delimiter can be , or : or ; Are such columns also taken care by pycaret when using setup method. If yes, what are the arguments that needs to be taken care of.

[@AIEngineeringLife]

You mean it is a list column?.. then you have to handle it externally

[@RajaSekharaReddyKaluri]

@@AIEngineeringLife not a list column. Just image column values to be red, blue and green,red,blue. Instead of single category for an attribute you now have multiple categories for an attribute. Can pycaret handle the above?

[@Thegreatindiaexpedition]

nailed it

[@Shankarpandala]

Few months back I have created a similar python library for myself. Thanks for the video. This is my package: lazypredict.readthedocs.io/en/latest/usage.html

[@AIEngineeringLife]

Nice work Shankar

[@rameshthamizhselvan2458]

Hahahahaa.... ML engineers are not required....

[@AIEngineeringLife]

I would not go to that extent :).. but it does automate some of boring work that happens in ML cycle. We still need to do feature engineering and post hoc analysis manually

[@yogeshgautam4762]

@@AIEngineeringLife sir, means no need to dive into models and hyperparameter tuning, please clear the confusion