search

mljar / mljar-supervised

Python package for AutoML on Tabular Data with Feature Engineering, Hyper-Parameters Tuning, Explanations and Automatic Documentation
https://mljar.com
MIT License
2.98k stars 393 forks source link

Get direction of feature weights for binary classification #730

Open fiaosborne opened 2 weeks ago

fiaosborne commented 2 weeks ago

Hello, I am using autoML to perform a binary classification task. Would someone be able to advise me on how I could create a visual similar to the mean feature importance chart but that also displays weather the feature indicates a weighting towards the positive or negative class. This would be really helpful for my interpretation of my model.

There are packages within scikit learn that do this but I do not know how to extract a particular model from automl to then use in scikit learn.

pplonski commented 2 weeks ago

Hi @fiaosborne,

Thanks for question. What method would you like to use for interpretation? I think the easiest way would be to use AutoML to search for optimal parameters and then train single model with scikit-learn.

fiaosborne commented 2 weeks ago

I am attempting to generate partial dependence plots to understand the directionality of each features influence on my model