Disclaimer: Dieser Thread wurde aus dem alten Forum importiert. Daher werden eventuell nicht alle Formatierungen richtig angezeigt. Der ursprüngliche Thread beginnt im zweiten Post dieses Threads.
FYI - Automatic Feature Selection
Just in case you are like me and worry about having too many features but don’t want to manually select the good ones,
you can use one of the following methods
-
Mutual Information: i saw it in the paper uploaded by Prof. Riehle
Luckily Scikit does all the work for you really easily
=> https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.mutual_info_classif.html -
there are other more methods like grid search, Chi-aquared test, Genetic algorithms etc…
but i opted for M.I. since it can be done in scikit in 2-4 lines of code
hope it helps
FYI - Automatic Feature Selection
really interesting, even if I think the chi-square would be more appropriate,
already because we can make it during the bivariate analysis of our different features and then it is visual (we can appreciate the result in a table).
But you’re at least right about the fact that it would be a more complex one to realize.
So thanks in any case for the help.
1 „Gefällt mir“