2
$\begingroup$

Is a categorical feature that has almost equally distributed in it's category more important or the one which one of it's category is predominant over all other ones? In data prepossessing step for "House Price" competition, I want to decide if Street feature is important or I can drop it from data set in order to avoid over-fitting. So I have plotted a swarm-plot as follows: enter image description here
How should I interpret this plot? Does it show that the Street plot can be dropped, or it says it is valuable for creating a model?

$\endgroup$
2
$\begingroup$

Feature importance is an empirical question. Train a model with the feature in it. Train another model without the feature in it. Then see which model does better when predicting new data, aka the test dataset. The difference between model performance on train and test datasets is one way to measure overfitting.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.