search

rasbt / mlxtend

A library of extension and helper modules for Python's data analysis and machine learning libraries.
https://rasbt.github.io/mlxtend/
Other
4.86k stars 856 forks source link

ci_bound unique value #702

Closed Icron1 closed 4 years ago

Icron1 commented 4 years ago

Hi,

I have a question about returned values from the Sequentiel Feature Selector. When performed with cross-validation, it provides a scoring criterion value and a unique ci_bound value. I read in the doc that is was the 95CI value. But how should we interpret this unique value?

Let's say scoring 0.8, ci_bound 0.2 ... does that mean 0.8 (95CI 0.78-0.82)? or (0.79-0.81)? Or something else? It's not clear to me.

Thanks Loïc

rasbt commented 4 years ago

Hi Loïc,

Let's say scoring 0.8, ci_bound 0.2 ..

in this case, this would mean that the confidence interval (95% by default) would range from 0.6 to 1.0. I hope this clarifies it. Otherwise, please let me know :)

Icron1 commented 4 years ago

Hi rasbt,

Thank you! That clear now.