compare all models with one line #27641
celestinoxp
started this conversation in
General
Replies: 1 comment 1 reply
-
Are those estimators fine tuned with the right hyperparameters appropriate for the dataset? It doesn't seem like it, and therefore that table is EXTREMELY misleading. Also, a lot of those models require some pre-processing before feeding the data to the model, which I'm not sure if it's done. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
For those who don't know Pycaret, it allows you to compare all scikit-learn models for classification or regression at once in one line of code "compare_models()".
So this seems very interesting because it is also possible to search for several parameters automatically through "tune_model()", this saves many hours of work searching for the best parameters, this is done automatically.
Now that I have drawn your attention to something extremely interesting, I would like to ask for the collaboration of the scikit-learn team to help with development, especially in the matter of support for scikit-learn 1.3 and in the near future for scikit-learn 1.4.
The pycaret team seems to be busy with other things and pycaret development is semi-stagnant, support is at scikit-learn 1.2. I tried to make some pull-requests, but unfortunately I don't have enough knowledge to solve these types of problems (I'm self-taught).
Those interested in collaborating can go to https://github.com/pycaret/pycaret and change scikit-learn to 1.3 in requirements.txt (I believe that many other packages should be changed to newer versions).
So this is a worthwhile challenge, as bugs in scikit-learn itself can be discovered more easily, as well as discovering improvements in scikit-learn itself in relation to other packages, for example: pandas, numpy, etc...
And something that can be more exciting is that new algorithms can be created in scikit-learn and you can easily see if they are better than others through compare_models in pycaret, so everything is intrinsically linked to have better productivity and save a lot of time and work.
What do you think about this?
Beta Was this translation helpful? Give feedback.
All reactions