Using sklearn selectkbest, I'd like to pick the top 20 predictors in my data set, but I'm not quite sure how to determine what makes a predictor effective. I'm using a list of lists(in the form [[price1, price change1], [price2, price change2], up to [price180, price change 180]]) with each list interval of 180 being a predictor.
Should i use f_regression(http://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.f_regression.html#sklearn.feature_selection.f_regression) to select, or mutual_info_regression, or possibly even f_classif? And for whichever function I'm using, what other arrays do i need to create in order to fill the arguments in said function?
My apologies for the novice questions, I'm an undergrad college student who's trying to learn quant finance on his own, and a lot of this material takes some time for me to wrap my head around.