search

shankarpandala / lazypredict

Lazy Predict help build a lot of basic models without much code and helps understand which models works better without any parameter tuning
MIT License
3.03k stars 345 forks source link

Hyperparameter tuning #309

Open shyamcody opened 3 years ago

shyamcody commented 3 years ago

Is your feature request related to a problem? Please describe. Recently I saw a review of this library where the user suggested that we should include hyperparameter tuning. I do realize it will increase the train time significantly, but it does sound like a good feature.

Describe the solution you'd like Optional parameter to tune the hyperparameters of the models while being fitted lazyclassifier or lazyregressor. We definitely need #114 and #65 to be implemented before this to efficiently run the operation. We will need to look into efficient parallel processing and threading also to reduce the overall time. Additional context Full review of the library which is motivation for this issue can be seen here.

mpmic commented 2 years ago

This sounds like a really good feature to have, since most models have different parameters, and handling hyperparameter tuning within lazypredict would save a lot of time!

BradKML commented 2 years ago

Isn't this issue and https://github.com/shankarpandala/lazypredict/issues/345 related and that tuning can be done through validation nudging? Alternatively what ither forms of non-carpet-search based tuning are there (e.g. Bayesian, Halving, Optuna, HyperOpt)

BradKML commented 1 month ago

Need to self-note that the original author tried to tell himself to add the feature https://github.com/shankarpandala/lazypredict/issues/418 But also other people have tried to fork it to get it working again https://github.com/nityansuman/lazypredict-nightly